Alarm sounded around supply chain risks

Copyright TechTarget - All rights reserved https://cyber.law.harvard.edu/rss/rss.html Techtarget Feed Generator en Mon, 10 Aug 2026 03:25:57 GMT https://www.theserverside.com editor@techtarget.com Recent <a href="mailto:https://www.computerweekly.com/news/366646678/Anthropic-lost-control-of-Claude-in-latest-AI-cyber-blunder">high-profile attacks by AI agents</a> have exposed <a href="mailto:https://www.computerweekly.com/microscope/news/366621171/Supply-chains-of-critical-industries-vulnerable-to-cyber-attack">vulnerabilities in supply chains</a> and the extent to which current technology can exploit them. Beyond the headlines, audit, tax and consulting firm RSM UK undertook its own research to assess the extent of risk being faced by businesses. It found that almost half (45%) of customers were not actively monitoring cyber and technology risks across their supply chains, and among those that did, levels of confidence were not particularly high. A significant number of customers have yet to tackle the issue, despite almost a quarter of respondents revealing they had been victims of a breach stemming from a cyber attack in the past year. <a href="https://www.rsmuk.com/">RSM UK’s</a> national technology risk assurance lead, Sheila Pancholi, said there was a need for education and action to ensure more customers deployed greater scrutiny of their supply chains. “The findings are particularly concerning, given the recent <a href="https://www.computerweekly.com/news/366646003/Hugging-Face-hacker-was-rogue-OpenAI-model">AI-led breach at OpenAI</a> has now taken cyber risks into uncharted territory. It’s no longer enough for businesses to address risk in their own systems and processes – they need to be aware of potential risks from third parties and suppliers in every link of their supply chain. An attack on any one of these could potentially put the whole supply chain in jeopardy,” she said. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> It’s no longer enough for businesses to address risk in their own systems and processes – they need to be aware of potential risks from third parties and suppliers in every link of their supply chain </figure> <figcaption> Sheila Pancholi, RSM UK </figcaption> </div> </blockquote> As is often the case, there was a confidence gap between those sitting in the boardroom and those responsible for delivering security and managing operations. The survey found that supply chain managers were more sober about the ability to identify and defend against risk and attacks than their bosses. Pancholi added that improving security was a given if customers wanted to reduce risk and exposure to attacks. “<a href="mailto:https://www.computerweekly.com/news/366633558/UK-ramps-up-ransomware-fightback-with-supply-chain-security-guide">Building cyber resilience in supply chains</a> is essential to ensure businesses can continue to run smoothly. A cyber attack can seriously hamper a business’s ability to operate for weeks, if not months, and the reputational damage can last much longer. The threat landscape is changing rapidly. It’s therefore important businesses adopt a ‘when’ not ‘if’ mentality and stay informed of evolving risks so they can be one step ahead of would-be attackers,” she said. That idea of “<a href="https://www.computerweekly.com/feature/The-year-of-AI-2026-sees-influx-of-ransomware-attacks">when, not if</a>’ is one that has taken a while to become established, but now seems to have become a de facto position for most customers. The phrase “<a href="mailto:https://www.computerweekly.com/microscope/news/366645156/Cohesity-emphasises-need-for-customer-minimum-viability">minimum viable company</a>” has emerged as more users seek to identify the crown jewels that must be protected to keep the business running. Although more work is needed, the recent attacks from <a href="https://www.computerweekly.com/blog/Cliff-Sarans-Enterprise-blog/Anthropic-Claude-incidents-show-the-limits-of-AI-guard-rails">Anthropic</a> and <a href="https://www.computerweekly.com/news/366646003/Hugging-Face-hacker-was-rogue-OpenAI-model">OpenAI</a> AI agents have exposed the need for greater awareness of the supply chain threat. Even in light of the recent AI agent issues, RSM found that only 37% of firms had considered the impact of an attack on one of their major suppliers. RSM advised customers to document third-party suppliers and rank them in terms of importance to the business, and then identify where potential threats could come from across the supply chain. With AI agents demonstrating their ability to bypass security, the need to protect against attacks originating from third-party suppliers has increased https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/cyber-security-attack-virus-malware-Skorzewiak-adobe.jpg https://www.computerweekly.com/microscope/news/366648132/Alarm-sounded-around-supply-chain-risks Thu, 06 Aug 2026 07:20:00 GMT Alarm sounded around supply chain risks Capita’s CEO was slammed by a staff member after making claims that the “right” things are in place to turn the <a href="https://www.computerweekly.com/news/366641895/Capita-lacked-detail-and-thoroughness-in-planning-botched-Civil-service-pension-scheme-takeover">botched Civil Service Pension Scheme (CSPS)</a> administration around. The comments from a staff member cast doubt on Capita’s ability to fix the problems being experienced by members of the pension scheme. In an internal message to staff, known as the Capita Daily Digest, Adolfo Hernandez, CEO at the outsourcer, explained the company’s lower-than-expected profit was “mainly due to additional costs” <a href="https://www.computerweekly.com/news/366645700/Capita-to-lose-money-on-pension-debacle-but-has-governments-private-parts-in-its-grip">related to CSPS failings</a>. Since taking over the administration of the CSPS, Capita has failed to meet its delivery promises and left some claimants in severe financial difficulties. The company is set to lose money on the seven-year contract, worth £239m. In 2023, the Cabinet Office awarded Capita the contract for the work that supports 1.7 million members. The takeover from the previous administrator, MyCSP, took place on 1 December 2025.  “The service there has not been good enough and we are sorry about the impact on members – but we now have the right people, processes and technology in place to put that right,” Hernandez wrote in a message to staff. <section class="section main-article-chapter" data-menu-title="‘Abysmal’ service"> <h2 class="section-title">‘Abysmal’ service</h2> In a response to Hernandez’s statement, copied to all Capita staff, a member of the pensions team said: “The service provided by Capita surpasses your assertion of ‘not good enough’, it has been abysmal for members. The fact members have died in waiting due to this whole catastrophe is a really harrowing thought that none of us should have to live with, yet we sadly do.” The worker said that staff have been trying to air their concerns about the CSPS contract, but that all contacts have been “met with vague, deflective and meaningless responses”. Hernadez was also lambasted for <a href="https://www.computerweekly.com/news/366639026/Thousands-of-unread-emails-and-20-million-database-errors-cause-civil-service-pension-hardship">his continued use of the backlog</a>, left to Capita by the previous supplier MYCSP, as an excuse. “I see you are still harping on the backlog scapegoat diatribe as well, which really must stop,” wrote the worker. “All staff have tried their best, with the very little tools we have at our disposal, to reduce the backlog.” The worker added: “I must also remind you it has increased by 40,000 since 1 December 2025, when the scheme went live at Capita.” Capita has faced MPs on the Public Accounts Committee (PAC) on several occasions since it took over the pension scheme, as well as a joint PAC and <a href="https://www.computerweekly.com/news/366642539/MP-committees-to-double-up-on-Capitas-civil-service-pension-crisis">Public Administration and Constitutional Affairs Committee (PACAC) hearing</a>. The firm has repeatedly said to MPs that it was putting in place measures and processes to get the CSPS back to where it should be. When contacted by Computer Weekly, a Capita spokesperson said: “We take the concerns of colleagues seriously and remain focused on supporting them as we continue our work to improve the service for members. “We continue to work at pace to resolve the operational issues with the CSPS in collaboration with the Cabinet Office. Despite progress made, we recognise the service has not been good enough, particularly for members waiting on bereavement, retirement and quotation cases, and we are sorry for the impact this has had on those members.” </section> <section class="section main-article-chapter" data-menu-title="‘Insulting to staff’"> <h2 class="section-title">‘Insulting to staff’</h2> The CEO’s claim that it now has the right people, processes and technology in place was an “insult” the staff, the worker added: “How can you state that Capita only now have the right people, when you have had staff at your disposal from the go-live date on 1 December? This is very insulting to all the staff who, [were] not only TUPE’d over, but were moved over internally in this company to work on the contract.” Capita’s boss was asked to explain his claims about having the ‘right’ parts in place when staff were struggling with resources needed. “Who are the ‘right people’? Please can you define and elucidate this point?” The worker went on: “…I am currently without a line manager, as are other teams. They are without an operations manager; we are short of senior pension technicians. There are people in place [who] have no experience in pensions, which isn’t always a bad thing, but given the scale of the pension scheme, we would require more pensions experience, which this company is letting walk out of the door. “Please can you provide a detailed list of what processes have been put in place to restore service? For somebody working on retirements, and discussing with colleagues across the business, I can’t see any.” It was suggested by the employee that processes are “conveniently put in place” each time Capita is called to a hearing to <a href="https://www.computerweekly.com/news/366645827/PAC-raises-more-questions-over-Capitas-role-in-government">discuss the troubled CSPS</a>. </section> <section class="section main-article-chapter" data-menu-title="Tech troubles"> <h2 class="section-title">Tech troubles</h2> The technology that staff are relying on was also criticised by the worker who, in response to the CEO, said: “From my experience, when I am not wasting my mobile phone battery using the authenticator app at least five times a day, after waiting patiently for 10 minutes for the VPN to load, we are still working from spreadsheets and still processing all workflows manually, without any sight of technological improvement. Please can you confirm what technology you have put in place to put things right?” This is not the first time Capita’s methods have been condemned by its own staff. <a href="https://www.computerweekly.com/news/366645478/Capitas-rush-to-hit-civil-service-pension-deadline-risking-errors-say-staff">CSPS staff aired concerns last month over</a> data accuracy, data protection and the possible impacts for scheme members. According to the Public and Commercial Services Union (PCS), some of its members working on the CSPS administration at Capita are being asked to upload large volumes of data into live systems despite known system and validation issues. “[Staff] say this has resulted in extensive manual checking, including line-by-line verification of data, increasing workload pressures and the risk of human error,” said the PCS. <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about Capita and the Civil Service Pension Scheme</h3> <ul type="square" class="default-list"> <li>July 2026: <a href="https://www.computerweekly.com/news/366645827/PAC-raises-more-questions-over-Capitas-role-in-government?_gl=1*xbkv0w*_ga*Mjg4MTcxMzU3LjE3Nzk5NjUxNTU.*_ga_TQKE4GS5P9*czE3ODU5Mjk3MjIkbzI2NiRnMSR0MTc4NTkyOTcyNiRqNTYkbDAkaDA.">PAC raises more questions over Capita’s role in government</a></li> <li>July 2026: <a href="https://www.computerweekly.com/news/366645700/Capita-to-lose-money-on-pension-debacle-but-has-governments-private-parts-in-its-grip">Capita to lose money on pension debacle but ‘has government’s private parts in its grip’</a>.</li> <li>July 2026: <a href="https://www.computerweekly.com/news/366645462/Capita-civil-service-pension-contract-prime-candidate-for-insourcing-says-government-minister">Capita civil service pension contract ‘prime candidate’ for insourcing, says government minister</a>.</li> <li>July 2026: <a href="https://www.computerweekly.com/news/366645478/Capitas-rush-to-hit-civil-service-pension-deadline-risking-errors-say-staff">Capita’s rush to hit civil service pension deadline risking errors, say staff</a>.</li> <li>July 2026: <a href="https://www.computerweekly.com/news/366645354/Union-calls-for-insourcing-of-pension-contract-as-Capita-shit-show-goes-on">Union calls for insourcing of pension contract as Capita ‘shit show’ goes on</a>.</li> <li>Jun 2026: <a href="https://www.computerweekly.com/news/366644692/Cabinet-Office-states-Capita-set-to-miss-Civil-Service-Pension-Scheme-deadline">Cabinet Office states Capita set to miss Civil Service Pension Scheme deadline</a>.</li> <li>Jun 2026: <a href="https://www.computerweekly.com/news/366643878/Capita-launched-civil-service-pension-scheme-site-without-basic-web-security">Capita went live with its troubled civil service pensions administration without a basic Domain Name System security feature</a>.</li> <li>May 2026: <a href="https://www.computerweekly.com/news/366643712/Ministers-refused-to-sign-off-563m-Capita-contract-amid-civil-service-pension-disaster">Ministers refused to sign off £563m Capita contract amid civil service pension disaster</a>.</li> <li>May 2026: <a href="https://www.computerweekly.com/news/366643012/Civil-servants-to-protest-at-Capita-general-meeting-amid-pension-crisis">Civil servants to protest at Capita general meeting amid pension crisis</a>.</li> <li>Apr 2026: <a href="https://www.computerweekly.com/news/366642492/Was-Capitas-Royal-Mail-pension-contract-a-botch-too-far">Was Capita’s Royal Mail pension contract a botch too far?</a></li> <li>Apr 2026: <a href="https://www.computerweekly.com/news/366641895/Capita-lacked-detail-and-thoroughness-in-planning-botched-Civil-service-pension-scheme-takeover">Capita lacked ‘detail and thoroughness’ in planning botched Civil Service Pension Scheme takeover</a>.</li> <li>Apr 2026: <a href="https://www.computerweekly.com/news/366642539/MP-committees-to-double-up-on-Capitas-civil-service-pension-crisis">MP committees to double up on Capita’s civil service pension crisis</a>.</li> <li>Apr 2026: <a href="https://www.computerweekly.com/news/366642312/Government-should-drop-Capita-from-civil-service-scheme-after-it-loses-Royal-Mail-role-says-union">Government should drop Capita from civil service scheme after it loses Royal Mail role, says union</a>.</li> <li>Apr 2026: <a href="https://www.computerweekly.com/news/366642152/Government-terminates-Capitas-Royal-Mail-pension-contract">Government terminates Capita’s Royal Mail pension contract</a>.</li> <li>Mar 2026: <a href="https://www.computerweekly.com/news/366640779/Capita-left-to-deal-with13000-civil-service-pension-cases-over-a-year-old">Capita left to deal with 13,000 civil service pension cases over a year old</a>.</li> <li>Feb 2026: <a href="https://www.computerweekly.com/news/366639026/Thousands-of-unread-emails-and-20-million-database-errors-cause-civil-service-pension-hardship">Thousands of unread emails and 20 million database errors cause civil service pension hardship</a>.</li> <li>Jan 2026: <a href="https://www.computerweekly.com/news/366637837/Troubleshooter-steps-in-as-Capita-and-civil-service-bosses-apologise-for-pension-scheme-problems">Troubleshooter steps in as Capita and civil service bosses apologise for pension scheme problems</a>.</li> <li>Oct 2025: <a href="https://www.computerweekly.com/news/366633358/Capita-rubbishes-Public-Accounts-Committee-report-claims">Capita rubbishes Public Accounts Committee report claims</a>.</li> </ul> </div> </div> </section> Capita CEO criticised after making Civil Service Pension Scheme claims in a message to staff https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/email-digital-communications-2-adobe.jpeg https://www.computerweekly.com/news/366648014/Capita-CEO-receives-scathing-rebuttal-from-employee-over-botched-contract Thu, 06 Aug 2026 05:20:00 GMT Capita CEO receives scathing rebuttal from employee over botched contract Government bodies have displaced telecoms operators as the most heavily targeted sector in South Korea, and the state-sponsored groups that have attacked the country for years no longer have it to themselves. That is the reading of Sojun Ryu, senior security researcher at Kaspersky’s global research and analysis team, who tracks Korean-speaking threat actors. “Korean-speaking threat actors have targeted South Korea for a very long time and they are still active,” he told Computer Weekly at the company’s <a href="https://www.computerweekly.com/news/366646680/Kaspersky-to-scan-AI-agents-for-backdoors-as-shadow-AI-spreads">APAC Cyber Security Weekend</a> in Guangzhou, China. “But many cyber criminals have begun to target South Korea as well.” Ryu also pointed the growing number of South Korean companies listed on ransomware data leak sites. “That means South Korea is not a safe country any more when it comes to cyber crime,” he said. The compromise of SK Telecom, the country’s largest telco, pushed cyber security up the public and political agenda over the past year, Ryu said. SK Telecom disclosed the breach in April 2025, and the country’s science ministry <a href="https://www.yahoo.com/news/south-korea-orders-sk-telecom-051411979.html">found the company negligent</a> after an investigation put the leak at 26.96 million pieces of user data. Besides telcos, Ryu said the government is also a key target, noting that threat actors have been methodically working through a large estate of public-facing websites run by South Korean agencies, producing a steady run of disclosures. One such actor is <a href="https://www.cisa.gov/news-events/cybersecurity-advisories/aa20-301a">Kimsuky</a>, the North Korea-linked group Kaspersky has tracked since 2013. Kaspersky’s research shows the group’s AppleSeed toolset is used mainly against government organisations, while its PebbleDash malware is aimed at the defence sector. Attackers also continue to exploit something specific to the country: its home-grown security software ecosystem. South Korean organisations run local security and authentication clients that have few equivalents elsewhere, and those clients have become a reliable route in. Ryu characterised the pattern as a South Korea-specific blend of <a href="https://www.kaspersky.com/about/press-releases/kaspersky-uncovers-new-lazarus-led-cyberattacks-targeting-south-korean-supply-chains">watering hole and supply chain attack</a>, and said it remains active. <section class="section main-article-chapter" data-menu-title="AI leaves fingerprints in the code"> <h2 class="section-title">AI leaves fingerprints in the code</h2> Ryu’s day job has changed as attackers started <a href="https://www.techtarget.com/cybersecurity/tip/How-AI-malware-works-and-how-to-defend-against-it">using large language models (LLMs) to conduct threat campaigns</a>. He noted that before generative AI, malware carried terse debugging strings – little more than a marker that a connection had succeeded. Today, Kaspersky is finding emoji in logging comments inside malware samples – tick marks and crosses of the sort a coding assistant produces. “We hadn’t seen these kinds of comments inside malware,” Ryu said. “Now they keep appearing.” Kimsuky had never shipped <a href="https://www.techtarget.com/searchapparchitecture/tip/The-fundamental-benefits-of-programming-in-Rust">Rust</a>-based malware before, he said, and when it did, the emoji came with it. Ryu treats that combination as a strong hint of machine-generated code. Kaspersky <a href="https://securelist.com/kimsuky-appleseed-pebbledash-campaigns/119785/">published research in May</a> linking the HelloDoor backdoor to the group and flagging comments that appeared to have come from an LLM rather than a developer. More consequential is malware that calls a model while it runs. Ryu described samples that talk to an LLM service during an intrusion to decide what to do next, rather than following hard-coded logic, and others that carry a local model, feed it details of the compromised environment and act on what comes back. The use of LLMs by threat actors also makes attribution harder. When a technique surfaces publicly, rival groups now copy it within days – possibly with LLM assistance – eroding the tradecraft signatures analysts rely on to tell threat actors apart. “It is very difficult to clearly separate them,” Ryu said. “Many vendors want to attribute immediately, but I don’t.” His preference is to go deeper into the malware itself before drawing any link. </section> <section class="section main-article-chapter" data-menu-title="Fighting back with the same tools"> <h2 class="section-title">Fighting back with the same tools</h2> Ryu uses LLMs in malware analysis to keep up with attackers who are doing the same, but he is more cautious about where the data goes. He noted that frontier models handle only external, publicly available research, while internal telemetry stays in-house. Telemetry is where the company's scale counts. <a href="https://www.computerweekly.com/news/366609114/How-Kaspersky-is-driving-growth-in-APAC">Kaspersky</a> trains its own detection models on data from its antivirus and its <a href="https://www.computerweekly.com/news/252468781/How-EDR-is-moving-beyond-the-endpoint">endpoint detection and response (EDR) and extended detection and response (XDR)</a> products, using it for better detection rather than analysis alone. Among attack vectors, open source remains the soft spot, Ryu noted. AI coding assistants have pushed developers to pull in more packages, and while <a href="https://www.techtarget.com/cybersecurity/news/252525335/Malicious-NPM-package-discovered-in-supply-chain-attack">NPM</a> and <a href="https://www.techtarget.com/cybersecurity/news/252528556/Malicious-Python-package-in-PyPI-poses-as-SentinelOne-SDK">PyPI</a> have tightened controls, he expects package repositories to stay a primary target for both nation-state groups and criminals. Quantum is closer than the timelines suggest. Ryu said he has already analysed malware using <a href="https://www.computerweekly.com/news/366640684/Shrinking-PQC-timeline-highlights-immediate-risk-to-data-security">post-quantum cryptographic algorithms</a> for its command-and-control traffic. “It is not the future any more,” he said, though he expects adoption to stay at the algorithm level for now, because quantum attacks need computing infrastructure that criminals do not yet have. He was more circumspect on <a href="https://www.computerweekly.com/news/366646154/Singapore-to-hold-CII-boards-accountable-as-AI-reshapes-OT-threat-landscape">operational technology (OT)</a>, which sits outside his research area, but agreed AI has made it easier for threat actors to target OT environments without understanding the inner workings of industrial control systems. Kaspersky also runs honeypots and uses them in research, though Ryu said they work better against cyber criminals than against APT groups, whose operators eventually work out they are in a honeypot or sandbox and back off. <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about cyber security in APAC</h3> <ul class="default-list"> <li>Unauthorised access to a development and testing environment managed by IBM has <a href="https://www.computerweekly.com/news/366645414/Breach-of-IBM-managed-environment-exposes-personal-data-of-70000-in-Singapore">exposed the names, NRIC numbers and property addresses of about 70,000 people</a> held by the Singapore Land Authority.</li> <li>Japan’s Nikkei has confirmed a major data breach that potentially <a href="https://www.computerweekly.com/news/366634243/Nikkei-data-breach-exposes-personal-data-of-over-17000-staff">exposed the personal information of more than 17,000 employees</a> and business partners after hackers infiltrated its internal Slack messaging platform.</li> <li>Australian privacy commissioner warns that the <a href="https://www.computerweekly.com/news/366633983/Fewer-data-breaches-in-Australia-but-human-error-now-a-bigger-threat">human factor is a growing threat</a> as notifications caused by staff mistakes rose significantly even as total breaches declined 10% from a record high.</li> <li>Philippine bank <a href="https://www.computerweekly.com/news/366633428/BDO-Unibank-taps-Zscaler-to-secure-cloud-migration">BDO is shoring up its cyber security capabilities</a> to protect its data and systems as it moves more services to the cloud and expands its physical presence into remote areas of the archipelago.</li> </ul> </div> </div> </section> Kaspersky researcher Sojun Ryu says ransomware crews have joined nation-state groups in going after South Korean organisations, as traces of LLM output start turning up inside malware https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/South-Korea-flag-hero-AdobeStock_500991221.jpg https://www.computerweekly.com/news/366647735/South-Koreas-government-overtakes-telcos-as-top-cyber-attack-target Thu, 06 Aug 2026 00:07:00 GMT South Korea's government overtakes telcos as top cyber attack target Asia-Pacific’s datacentre development pipeline reached a record 26.5GW in the first half of 2026, expanding by 7.1GW in six months as hyperscalers, cloud providers and artificial intelligence (AI) platforms raced to lock in capacity across the region. The half-year gain, disclosed in Cushman & Wakefield’s APAC datacentre H1 2026 update, is more than double the <a href="https://digital.cushmanwakefield.com/apacdatacentreupdateh22025-02-2026-apac-regional-en-content-datacentres/2/">2,751MW added in the second half of 2025</a> and marks one of the sharpest six-month increases the property services firm has recorded. The pipeline now comprises 4.8GW under construction and 21.7GW in planning. About 1.4GW of operational capacity came online during the period, yet colocation vacancy still tightened from 10.9% to 10.3% – a sign that new supply is being absorbed roughly as fast as it can be delivered. “AI and cloud investment across Asia Pacific has entered a phase of rapid, power-constrained execution,” said Andrew Green, head of Cushman & Wakefield’s datacentre group in Asia-Pacific. “Historically, datacentres were clustered around connectivity, but now, we’re seeing expansion gravitate towards markets that can provide power at scale.” Southeast Asia now accounts for about 50% of APAC’s under-construction capacity, up from 31% at the end of 2025. Malaysia leads with 1,039MW being built, followed by Thailand on 859MW. Almost all of that activity sits in two cities. Johor’s under-construction capacity nearly doubled to 602MW, lifting its total pipeline to 3,088MW, while Bangkok posted the region’s fastest expansion – a 148% jump to 859MW, which on the report’s own figures represents the whole of Thailand’s construction activity. Johor’s ascent traces back to Singapore’s <a href="https://www.computerweekly.com/news/252522992/Singapore-to-restart-datacentre-builds-with-sustainability-focus">2019 moratorium on new datacentre builds</a>, imposed over land, water and energy constraints, which sent hyperscalers looking across the Causeway. Singapore lifted the pause in 2022 but attached sustainability standards and capacity limits that have kept growth deliberately modest. <a href="https://www.computerweekly.com/news/366643987/Alibaba-Cloud-opens-Johor-cloud-region">Alibaba Cloud opened a Johor cloud region</a> in June, its largest digital infrastructure footprint in Southeast Asia to date. Power has since become the sector’s binding constraint. Deloitte expects <a href="https://www.deloitte.com/ap/en/perspectives/powering-asia-pacific-data-centre-boom.html">$800bn of datacentre investment in APAC by 2030</a>, but grid capacity has not kept pace. Jean-Christophe Moureau, senior vice-president for solutions operations at Schneider Electric, told Computer Weekly that <a href="https://www.computerweekly.com/news/366642996/Inside-the-global-datacentre-squeeze">site selection had become an energy question first</a> and a property question second, noting that Singapore had told operators there was no power available to them before 2028. That pressure is changing how facilities get built. Pritesh Swamy, head of research and consulting at Cushman & Wakefield's APAC datacentre group, said developers were moving away from standard blueprints. “In most markets, speed to market is increasingly shaping development decisions,” he said. “Modular construction approaches and advanced cooling technologies are helping operators maximise available infrastructure while shortening deployment timelines.” Outside Southeast Asia, Sydney recorded the largest pipeline growth, up 65% to 2,134MW, with planned capacity rising 83% to 2,018MW. Mumbai reached 1,726MW and Jakarta 1,699MW, the latter more than doubling its under-construction capacity to 395MW. Johor also retained top spot in the firm’s APAC datacentre maturity index, which ranks 30 markets, with combined operational and future capacity above 4GW. Cushman & Wakefield raised the index’s benchmarks during the half to reflect how quickly regional capacity has scaled, prompting several markets to be reclassified. Australia, India, Japan and Malaysia are all on track to pass 2GW of operational capacity by 2028, according to the report. “While traditional hubs will remain critical to the digital economy, we’re seeing a broader group of markets emerge as important destinations for future datacentre investment,” Green said. <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about datacentres in APAC</h3> <ul class="default-list"> <li>At Gitex Asia 2025, industry leaders discuss how the computational demands of advanced AI models are forcing a <a href="https://www.computerweekly.com/news/366623023/How-AI-workloads-are-reshaping-datacentre-design">rethink of datacentre power, cooling and networking</a> infrastructure.</li> <li>A joint report by Singapore government agencies and industry players details how <a href="https://www.computerweekly.com/news/366644352/Singapore-to-transform-industrial-island-into-AI-ready-datacentre-hub">Jurong Island will be redesigned into a highly secure, circular digital infrastructure hub</a> by 2050.</li> <li>STT GDC’s FutureGrid Accelerator on Jurong Island will <a href="https://www.computerweekly.com/news/366637683/STT-GDC-launches-HVDC-testbed-to-address-AIs-power-demands">test the use of high-voltage direct current power systems</a>, as traditional alternating current power systems can no longer address the growing energy needs of AI datacentres.</li> <li>As terrestrial datacentres face severe power and cooling constraints, <a href="https://www.computerweekly.com/news/366642407/India-aims-for-the-stars-with-orbital-datacentres">space-based facilities promise unlimited solar energy and low latency</a> for AI inferencing, provided the industry can overcome challenges.</li> </ul> </div> </div> Asia-Pacific’s datacentre development pipeline grew by 7.1GW in the first half of 2026, with Southeast Asia now accounting for half of all capacity under construction https://cdn.ttgtmedia.com/visuals/ComputerWeekly/HeroImages/datacentre-cooling-fans-air-conditioning-horizon-adobe.jpg https://www.computerweekly.com/news/366647782/Power-scarcity-pushes-APAC-datacentre-pipeline-to-record-265GW Wed, 05 Aug 2026 21:15:00 GMT Power scarcity pushes APAC datacentre pipeline to record 26.5GW The next big wave of AI adoption is underway as executives across industries plot how to bring agentic AI into their organizations. Many believe agentic AI will remake the workforce, creating an environment where humans and agents work alongside each other to reduce costs, boost productivity and scale services. Agentic AI enables organizations to automate entire workflows, with agents capable of making decisions autonomously to execute complex, multistep processes across multiple digital systems. Some 17% of organizations have already <a href="https://www.gartner.com/en/articles/hype-cycle-for-agentic-ai" target="_blank" rel="noopener">deployed agents</a>, according to the "2026 CIO and Technology Executive Survey" from Gartner, and more than 60% expect to do so within the next two years. <a href="https://www.techtarget.com/searchenterpriseai/definition/agentic-AI"></a>But the reality of agentic AI is more complex than a list of anticipated benefits for the organization. It also carries risks, and the technology's autonomy can have consequences that extend far beyond those associated with traditional automation. As agentic AI transforms enterprise operations, analysts at Forrester argue that traditional "block or allow" security models designed for predictable, human-initiated actions are inadequate for autonomous agents that operate at machine speed, make independent decisions and adapt their behavior dynamically. Instead, CISOs must pivot to a "probability of success" approach that continuously assesses risk in real time based on agent behavior, context and potential impact. One emerging option for managing <a></a><a href="https://www.techtarget.com/searchenterpriseai/feature/Security-risks-in-agentic-AI-systems-and-how-to-evaluate-threats">agentic AI risk</a> is the Agentic AI Enterprise Guardrails for Information Security, or <a href="https://www.forrester.com/technology/aegis-framework/" target="_blank" rel="noopener">AEGIS, framework</a> from Forrester Research. AEGIS aims to help organizations deploy agentic AI with reasonable guardrails in place, enabling innovation without recklessly exposing the enterprise to risk. By implementing continuous monitoring, <a href="https://www.techtarget.com/searchsecurity/feature/How-AI-threat-detection-is-transforming-enterprise-cybersecurity">threat detection</a> and adaptive controls, AEGIS can help organizations build resilient AI operations that detect and respond to anomalies before they escalate into crises. <section class="section main-article-chapter" data-menu-title="Agentic AI risks"> <h2 class="section-title">Agentic AI risks</h2> Agentic AI significantly expands an organization's <a href="https://www.techtarget.com/searchsecurity/tip/How-to-implement-an-attack-surface-management-program">attack surface</a>, creating more pathways for adversaries to breach enterprise systems and exploit agent behavior. Additionally, agentic AI can lead to cascading failures. As agents execute tasks through multistep workflows across interconnected platforms, they're subject to goal drift or hijacking through manipulated prompts, data inputs or tool permissions. An unintentional error or a maliciously prompted action taken by an agent anywhere along the way can ripple through the connected systems. Depending on the work an agent performs, the consequences can be severe. Few organizations today are prepared to guard against such situations, according to research. Deloitte's 2026 "State of AI in the Enterprise" <a></a><a target="_blank" href="https://www.deloitte.com/content/dam/assets-zone3/us/en/docs/services/consulting/2026/state-of-ai-2026.pdf" rel="noopener">report</a> found that only 21% of organizations had a mature model for governance of autonomous agents. That means they lack clear boundaries for what decisions agents can make, have few or no real-time monitoring systems to track agent behavior and flag anomalies, and have no audit trails to capture the full chain of agent actions. Security leaders acknowledge that traditional cybersecurity architectures, designed for digital systems where humans ultimately execute actions, cannot handle enterprise environments in which AI agents act autonomously. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> AEGIS offers CISOs, CIOs and CTOs a framework for building architectural and operational foundations that enable safe and responsible agentic AI deployments. </figure> </div> </blockquote> </section> <section class="section main-article-chapter" data-menu-title="A framework for responsible agentic AI deployment"> <h2 class="section-title">A framework for responsible agentic AI deployment</h2> According to Forrester, AEGIS offers CISOs, CIOs and CTOs a framework for building architectural and operational foundations that enable safe and responsible agentic AI deployments. AEGIS hinges on the following principles: <ul class="default-list"> <li>Least agency. An agentic AI-specific version of the <a href="https://www.techtarget.com/cybersecurity/definition/What-is-the-principle-of-least-privilege-POLP?amp=1">principle of least privilege</a>, least agency gives AI agents as few capabilities, tools and permissions as possible. While least privilege limits users' ability to access resources, least agency limits agents' ability to make decisions and take action.</li> <li>Continuous risk management to maintain trust. Environments where agentic AI has the ongoing ability to make decisions and take action require continuous monitoring of agents, models and data. The goal, according to Forrester, is to provide moment-by-moment assurance that agentic AI deployments are secure.</li> <li>Explainable outcomes. Agentic AI makes decisions based on learning, context and intent, rather than static rules. Both people and systems must therefore be able to understand and explain security guardrail outcomes.</li> </ul> <h3>6 domains of the AEGIS framework</h3> The AEGIS framework features the following six domains, which Forrester says organizations should adopt in phases: <ol class="default-list"> <li>Governance, risk and compliance. Includes real-time risk and compliance monitoring, automated detection of behavior drift, cross-functional risk mapping and policy-as-code guardrails to enforce machine-executable policies.</li> <li>Identity and access management. Calls for <a href="https://www.techtarget.com/cybersecurity/tip/How-to-fix-cybersecuritys-agentic-AI-identity-crisis?amp=1">managing agents as first-class entities</a> with ownership, credentials, lifecycle management and auditability, and to have just-in-time, least-privilege authorization.</li> <li>Data security and privacy. Implements <a href="https://www.techtarget.com/searchsecurity/tip/How-to-write-a-data-classification-policy-with-template">unified definitions of sensitive data</a>, purpose-bounded data access, expanded data security posture management, data loss prevention, digital asset management for agent actions and privacy-preserving techniques such as masking, encryption and synthetic data.</li> <li>Application security. Includes the use of <a href="https://www.techtarget.com/searchsecurity/tip/Use-AI-threat-modeling-to-mitigate-emerging-attacks">AI-specific threat modeling</a>, rigorous validation of agent-generated code, software bills of materials, AI bills of materials for provenance, secure prompt engineering and continuous observability across the agent lifecycle.</li> <li>Threat management. Calls for detailed logging of prompts, actions and reasoning steps, errors; detection for prompt injection, hallucinations and drift; purple teaming for agent behaviors; and automated response playbooks tied to agent actions.</li> <li>Zero-trust architecture. Includes microsegmentation, API gateways, privilege and objective constraints, access brokers and network-level containment.</li> </ol> CISOs should note that implementing AEGIS requires specialized expertise in both AI systems and enterprise security architecture. Many enterprises operate hybrid environments with legacy systems that lack modern security capabilities. Integrating AEGIS controls with older infrastructure can be technically complex and could require significant modernization efforts, particularly in areas such as <a href="https://www.techtarget.com/searchsecurity/tip/CISOs-guide-to-privileged-identity-management">identity management</a> and <a href="https://www.techtarget.com/cybersecurity/tip/How-to-implement-zero-trust-for-AI?amp=1">zero-trust</a> architecture. Organizations might need to invest in training or external expertise to execute the framework effectively. The phased long-term implementation timeline also requires sustained organizational commitment and resources. Despite these challenges, AEGIS offers a path forward for organizations serious about deploying agentic AI responsibly. Critically, the framework acknowledges that agentic AI is not simply another application to secure, but a fundamentally new category of enterprise actor requiring purpose-built governance and protection. Mary K. Pratt is an award-winning freelance journalist with a focus on covering enterprise IT and cybersecurity management. </section> Learn how Forrester's AEGIS framework can help organizations securely deploy agentic AI, shifting from static controls to dynamic risk assessment that balances innovation with safety. https://cdn.ttgtmedia.com/rms/onlineimages/ai_a352095729.jpg https://www.techtarget.com/cybersecurity/tip/How-the-AEGIS-framework-mitigates-agentic-AI-risks Wed, 05 Aug 2026 14:50:00 GMT How the AEGIS framework mitigates agentic AI risks Up to three-quarters of young adults are using AI for health support before seeking the advice of a medical provider, according to new <a href="https://www.prnewswire.com/news-releases/76-of-gen-z-and-63-of-millennials-turn-to-ai-for-first-line-of-primary-healthcare-302841887.html">polling</a> from Aflac and Kantar Profiles. But this doesn't necessarily indicate a preference for AI chatbots. Rather, it could indicate growing frustration from Gen Z and millennial patients seeking care in an increasingly complex and <a href="https://www.techtarget.com/healthcarepayers/news/366625340/Health-insurance-coverage-costs-soar-6-in-2025">unaffordable healthcare system</a>. The rate at which Gen Z and millennial <a href="https://www.techtarget.com/patientengagement/news/366636461/How-docs-can-manage-patients-consulting-AI-medical-advice">patients use AI for health support</a> far outpaces that of older healthcare consumers, the survey of more than 2,000 employed U.S. adults showed. While 76% of Gen Z and 63% of millennials use the technology for health advice, only 47% of Americans overall said the same. In fact, 18% of Gen Z and 14% of millennials say they won't even book with a provider until they've exhausted all other AI or online tools or until their health concerns escalate. This was far less common among older adults, reaching less than 5% of Gen X or baby boomer respondents. Much has been made of AI chatbots and how patients might use them to <a href="https://www.techtarget.com/patientengagement/news/366644898/KFF-Healthcare-cost-pressures-lead-patients-to-use-AI-for-health-info">bypass growing healthcare costs</a>. That could be part of what's happening here, as 50% of Americans overall report increased stress due to healthcare costs. But Gen Z and millennials' cost concerns are right in line with the national average, the Aflac survey showed. In other words, cost alone can't explain why they use AI more than other age groups. Instead, industry leaders seeking to better understand young adults' online health habits should consider the burden of healthcare complexity. <section class="section main-article-chapter" data-menu-title="Traditional healthcare models not working for young adults"> <h2 class="section-title">Traditional healthcare models not working for young adults</h2> Young adults' frustration with the traditional healthcare model is showing up in the form of disengagement. Around two-thirds of Gen Z (65%) and 61% of millennials say they've <a href="https://www.techtarget.com/patientengagement/news/366641055/Cancer-screening-rates-dip-in-US-as-cost-concerns-deter-access">delayed or skipped routine checkups and screenings</a>, higher rates than Gen X (47%) and baby boomers (33%). It's not exactly surprising that <a href="https://www.techtarget.com/patientengagement/news/366645339/Report-29-of-young-adults-dont-have-a-primary-care-provider">younger patients are foregoing primary and preventive care</a>. Young adults are generally healthier and report fewer chronic illnesses that need attention. According to the Aflac survey, 35% of Gen Z and 29% of millennials said they didn't access primary and preventive care because they already feel healthy. That's compared to a quarter of all respondents who said the same. But the access divide is likely more complex than that. Instead, it's clear that Gen Z and millennial patients are sick of the logistical issues that have long plagued medicine. Around half of young adults said their biggest hurdle to accessing preventive care was the logistics, including <a href="https://www.techtarget.com/patientengagement/news/366584384/Digital-Patient-Experience-Hinges-on-Online-Appointment-Scheduling">scheduling difficulties</a>, long appointment wait times and having to take time off work. This compares to 43% of respondents overall. Long appointment wait times stood out as a particular sticking point, with 41% of Gen Z saying they've canceled or decided not to schedule an appointment because the <a href="https://www.techtarget.com/patientengagement/news/366624919/Average-patient-appointment-wait-time-is-31-days-in-2025">wait time was too long</a>. </section> <section class="section main-article-chapter" data-menu-title="Young patients take their healthcare online"> <h2 class="section-title">Young patients take their healthcare online</h2> According to the Aflac researchers, it's not entirely worrisome that young adults are leveraging AI, per se, but some startling trends are emerging. "The use of AI and other digital-first healthcare resources isn't inherently negative, as it ultimately depends on how the information is used," Matthew Owenby, Aflac Inc.'s chief human resources officer and chief administrative officer, said in a press release. "What is alarming is the rate that Gen Z is delaying or forgoing preventive care, greatly lessening the opportunity for early detection or a critical health diagnosis from a medical professional." Young adults are taking their healthcare online, in general, the survey continued. Although still in the minority, the number of young adults who trust influencers and social media as health resources far outpaces their older counterparts. Around a quarter (23%) of Gen Z and 17% of millennials trust influencers, compared to less than 5% of Gen X and baby boomers. More young adults <a href="https://www.techtarget.com/patientengagement/news/366613215/Patients-trust-social-media-over-docs-for-health-information">trust social media</a>, with 28% of Gen Z and 20% of millennials reporting trust overall. Meanwhile, only 6% of Gen X and 4% of baby boomers trust social media sites for health information. In fact, young adults' trust in online resources (online search, medical websites and social media) is starting to outpace their trust in providers. For example, only 22% of Gen Z and 31% of millennials prefer to consult healthcare professionals for healthcare advice. This compares to 45% of Gen Z and 54% of millennials who consult online search for medical advice. Despite being plugged into online health resources, the survey showed that a <a href="https://www.techtarget.com/patientengagement/news/366630114/Women-missing-cancer-screening-need-better-patient-education">patient education gap persists</a>. Americans of any age aren't sure when they're supposed to begin getting cancer screenings. Gen Z, in particular, is falling behind. A whopping 90% of Gen Z who have been diagnosed with cancer admit to avoiding a screening. Meanwhile, 51% of Gen Z don't think it's likely they will ever be diagnosed with cancer. Instead of accessing a cancer screening, 15% of Gen Z said they prefer to research online, while 11% said they prefer to self-diagnose. "As a leading provider of cancer insurance in the United States, Aflac often sees both the positive impact of early detection and the negative consequences of a delayed diagnosis,"  Owenby said. "Fortunately, early detection can push five-year survival rates above 90% for many cancers, according to the American Cancer Society." Sara Heath is an executive editor at Xtelligent Healthcare Media, where she covers patient engagement, healthcare policy and health IT. </section> Survey data shows Gen Z and millennials use AI for health advice more than older generations, likely because they're more frustrated with care access inefficiencies. https://cdn.ttgtmedia.com/rms/onlineimages/chatbot_g860770394.jpg https://www.techtarget.com/healthtechanalytics/news/366647732/Amid-mounting-care-access-barriers-young-adults-seek-health-AI Wed, 05 Aug 2026 14:38:00 GMT Amid mounting care access barriers, young adults seek health AI Employers are increasingly considering Individual Coverage Health Reimbursement Arrangements, or ICHRAs, for their workers' health insurance, though knowledge gaps could dampen adoption, according to a recent <a href="https://www.ebri.org/content/2026-ebri-morgan-health-employer-ichra-survey">survey</a> from the Employee Benefit Research Institute and Morgan Health. More than one-third of the nearly 1,000 employer respondents  said they were actively planning or evaluating an ICHRA, which allows companies to reimburse workers for purchasing their own health insurance. However, employers were significantly more likely to be evaluating these arrangements than actively planning, the survey revealed. Employers are still hesitant to adopt ICHRAs largely because of individual marketplace concerns, including the affordability of premiums and out-of-pocket costs, adequacy of provider networks and whether employees would find it difficult to shop for a plan. Concerns about costs are especially relevant as premiums on the Affordable Care Act's Marketplace, where many employees with ICHRAs purchase coverage, are <a href="https://www.techtarget.com/healthcarepayers/news/366645717/ACA-premiums-set-to-surge-14-as-tax-credits-expire">set to surge by 14% next year</a>. ICHRA adoption has accelerated since the Trump administration made the arrangements an official health benefit option in 2020, giving employers more flexibility with health reimbursement arrangements without company size limits. Employers acknowledge the flexibility of benefits that ICHRAs offer, which was one of the top reasons businesses in the survey said they would consider offering one to their employees. Business leaders also generally agreed that ICHRAs are more affordable, offering advantageous tax benefits and lower costs for the company. About half of the survey respondents from both large and small businesses cited this as a reason to offer one. For small businesses, in particular, the top reason to consider an ICHRA was employee choice, preference and needs. However, there are significant knowledge gaps about what ICHRAs truly are, which could hinder adoption, according to the survey. Nearly six in 10 employers reported being at least somewhat familiar with ICHRAs, but small employers that did not offer health benefits at all were substantially less familiar with the option. About half of respondents from large employers also incorrectly identified an ICHRA as a type of group health insurance plan that employers choose and purchase coverage for their workers. Additional education and awareness are needed to drive adoption, the survey stated. But policy changes could bolster ICHRA adoption, especially if they stabilize the individual market, employers said. They also indicated that state and federal tax credits would drive their adoption of an ICHRA, while greater flexibility to combine ICHRAs with traditional group health plans would help as well. About three-quarters of employers said that they would be more likely to offer an ICHRA if it had features similar to group health plans, such as coverage for mental health and fertility services. About the same percentage of employers also said they would be more likely to adopt an ICHRA if their broker or benefits consultant recommended one. Brokers are increasingly suggesting ICHRA adoption to their clients, with a recent survey finding that <a href="https://www.techtarget.com/healthcarepayers/news/366643635/ICHRAs-enter-the-mainstream-market-with-broker-support">half are actively recommending them</a>. Still, interest is not the same as implementation, EBRI and Morgan Health researchers emphasized. The arrangements remain an emerging benefit strategy, but one that could take hold over the next two years as awareness grows, the survey indicated. Jacqueline LaPointe is an Executive Editor at Xtelligent Healthcare Media, covering revenue cycle management, healthcare payers, health policy and health IT since 2016. More than one-third of employers offering health benefits are at least evaluating an ICHRA, although marketplace instability and costs are hindering adoption. https://cdn.ttgtmedia.com/visuals/ComputerWeekly/HeroImages/insurance-contract-Nikolai-adobe.jpg https://www.techtarget.com/healthcarepayers/news/366647334/Employers-eye-ICHRAs-but-affordability-remains-a-concern Wed, 05 Aug 2026 14:37:00 GMT Employers eye ICHRAs, but affordability remains a concern The UK’s <a href="https://www.aisi.gov.uk/" target="_blank" rel="noopener">AI Security Institute</a> (AISI), a governmental research body dedicated to enhancing understanding of artificial intelligence (AI) cyber risks, has become the latest organisation to admit that AI models supposedly under its control ran a series of cyber attacks against real people and organisations during a routine test. The latest in a string of similar incidents occurred just days after Anthropic owned up to a series of unforced errors that led <a href="https://www.computerweekly.com/news/366646678/Anthropic-lost-control-of-Claude-in-latest-AI-cyber-blunder" target="_blank" rel="noopener">variants of its Claude model to conduct a series of cyber attacks</a>, a disclosure that came hot on the heels of another cyber attack in which OpenAI models <a href="https://www.computerweekly.com/news/366646003/Hugging-Face-hacker-was-rogue-OpenAI-model" target="_blank" rel="noopener">took potshots at Hugging Face</a>. AISI, which originated within the now defunct <a href="https://www.computerweekly.com/news/366646157/Will-digital-government-recover-from-DSIT-breakup" target="_blank" rel="noopener">Department for Science, Innovation and Technology</a> (DSIT), said: “On July 28 2026, AISI's Security Team detected unusual data transfers leaving our research systems during a routine cyber evaluation. On investigation, we found that some of the agents being tested had engaged in sustained, potentially harmful activity directed at real people and organisations. We declared a security incident and, within roughly one hour of discovery, had contained it and begun a full investigation.” The incident involved Anthropic’s Claude Mythos 5 and OpenAi’s GPT-5.6-Sol – both of which were also <a href="https://www.computerweekly.com/blog/Cliff-Sarans-Enterprise-blog/Anthropic-Claude-incidents-show-the-limits-of-AI-guard-rails" target="_blank" rel="noopener">culprits in the previous attacks</a> – operating in “deliberately permissive” conditions with access to the open internet and some safety filters turned off. AISI claimed this was done so that it could evaluate and understand the capabilities of frontier models, including whether they could be used for cyber attacks. In 10 out of 122 runs of a specific evaluation challenge, AISI admitted the models took “autonomous, unsanctioned” actions in 19 cases. These were all unsuccessful, it added, and the resulting investigation has not surfaced any real-world harm. The majority of these actions arose from a single, sustained line of activity taken by Mythos 5, which echoed in its nature actions the same model also took during last week’s incident at Anthropic. This incident saw Mythos 5 attempt to conduct an <a href="https://www.computerweekly.com/news/366646561/Amazon-pins-multiple-open-source-compromises-on-North-Korea" target="_blank" rel="noopener">open source supply chain attack</a> by inserting malicious code into a public project. It created a series of fake online identities that it used to pressurise the human project maintainer into approving its malicious code. When the agent’s pull request was publicly challenged, it tried to edit its early activity to look harmless and considered adopting a new identity to continue the attack. This kind of social engineering by an AI has not been seen before, according to AISI. Mythos 5 also attempted to insert malicious code as a form of prompt injection attack to manipulate other automated AI coding systems that might pick them up later on and left public messages on GitHub offering to work with other agents participating in the same challenge, providing instructions on how to reuse accounts and artefacts that were left behind – information that was discovered and used by other agents. However, during this attack, Mythos 5 used the Tor network to try to bypass GitHub’s network restrictions, which is what triggered the AISI security alert. “This is the first time we have seen risks around autonomy and deception manifest this clearly, without specific prompting, in the real-world,” said AISI. <section class="section main-article-chapter" data-menu-title="AI working as intended?"> <h2 class="section-title">AI working as intended?</h2> AISI said that at a purely objective level, the agents involved took these malicious actions in order to accomplish the task they had been set successfully, but its investigation has identified several factors that appear to have contributed to their behaviour. It said the task was intentionally difficult and that it had deliberately misconfigured the challenge prompt to fool the agents into thinking they couldn’t solve it without going beyond the scope of the task. The agents were allowed to access the open internet to enable them to download any tools they might need, but AISI admitted its researchers failed to anticipate they would be persistent enough to consider engaging in deceptive, social engineering practices – as such they had not told the agents not to do this. Concerningly, AISI also said it had not used monitoring tools to watch the test as it ran – its security team only detected the anomalous traffic through general monitoring after the fact. Had its teams been watching, they may have been able to flag or block the malicious behaviours more quickly. AISI also noted another factor that proved instrumental in limiting the real-world damage – human code reviewers and good security practice by a potential victim, who on suspecting the code was malicious opened it in a secure sandbox environment. “In these cases, standard good practice, human judgement, and caution around AI-generated code stopped the worst outcomes. But in several cases the margin between failure and success was narrow, resting on human vigilance rather than a technical barrier that would reliably prevent this behaviour in a more capable agent,” said AISI. <a href="https://www.huntress.com/" target="_blank" rel="noopener">Huntress</a> virtual chief information security officer (VCISO) and EMEA cyber security advisor Muhammad Yahya Patel said: “If you give a frontier model a cyber security challenge, disable its safety classifiers, hand it open internet access and tell it to find a way through, you’ve essentially described the setup for an offensive security operation. The model has been trained on vast amounts of security research, exploit documentation, social engineering techniques and attack methodology. Of course it reaches for those tools. The model is doing exactly what it was implicitly asked to do. “The security industry commentary that treats this as a shocking discovery that AI can behave offensively is frankly naive about what these models are and what the test conditions were.” </section> <section class="section main-article-chapter" data-menu-title="Unprecedented risk"> <h2 class="section-title">Unprecedented risk</h2> Andrea Miotti, CEO of non-profit <a href="https://controlai.org/" target="_blank" rel="noopener">ControlAI</a>, who is working to raise awareness of and prevent some of the more existential risks AI poses to human life, said the top AI companies were “demonstrably unable” to control their models and described the national security risks posed by this reality as “unprecedented”. “This is the predictable result of the AI industry racing to build superintelligent AI that is vastly smarter than humans across the board and capable of overpowering and outmaneuvering our national security apparatuses. Nobel Prize winners, leading AI experts and even the CEOs of the AI companies warn that superintelligent AI would pose an extinction risk to humanity,” said Miotti. “Governments need to recognise the enormous risks that these rogue AIs point towards, and act now to negotiate an international agreement prohibiting the development of superintelligence. The clock is ticking.” AISI insisted it was a “capable organisation with strong safety practices” but warned that no defences will be sufficient indefinitely. Alongside the incidents reported by OpenAI and Anthropic, it noted that the incident more broadly points to a shift in the risk landscape where harm arises not only through deliberate misuse of AI by humans, but through actions taken independently by sufficiently capable agents. The organisation said it would treat this as a serious incident and implement “lasting change” for its own evaluation protocols and security architectures. It admitted it needed tighter controls on internet access and will explore the use of more fine-grained network controls into its cyber ranges to preserve realism while constraining the agents. It will also ramp up its monitoring procedures and reassess how it designs cyber tests, placing the assumption that capable agents will try to act beyond their remit. “AISI exists to identify these problems, understand them and share what we learn so they can be addressed before more capable systems are deployed The work is not complete … The task now is to strengthen our defences and ensure that safety work keeps pace,” it said. AISI’s disclosure notice, and a full technical report (PDF format) <a href="https://www.aisi.gov.uk/blog/incident-report-unsanctioned-agent-behaviour-during-cyber-testing">can be found here</a>. <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about AI-driven cyber attacks</h3> <ul class="default-list"> <li>In June, the Five Eyes intelligence bloc warned that frontier AI models will pose a greater cyber security risk to governments and businesses than previously thought, <a href="https://www.computerweekly.com/news/366644997/AI-powered-cyber-attacks-may-be-just-months-away-warn-Five-Eyes" target="_blank" rel="noopener">putting them at risk before 2026 is out</a>.</li> <li>The reality of the AI-orchestrated cyber attack on Hugging Face is rather more complicated, and much more worrying, than it might appear at face value. <a href="https://www.computerweekly.com/news/366646245/Did-an-AI-agent-really-break-free-and-attack-another-company" target="_blank" rel="noopener">Dan Raywood investigates</a>.</li> <li>Anthropic's revelation that a model broke out of its test environment three times and accessed external systems, shows the power of these models, <a href="https://www.computerweekly.com/news/366646658/Gartner-Why-cybersecurity-must-shift-to-outcomes-against-AI-led-attacks" target="_blank" rel="noopener">say Gartner analysts</a>.</li> </ul> </div> </div> </section> Anthropic’s Mythos 5 has been caught orchestrating a real-world open source supply chain attack using social engineering techniques during a test run by the UK’s AI Security Institute https://cdn.ttgtmedia.com/visuals/German/article/cyber-forensics-adobe.jpg https://www.computerweekly.com/news/366647165/Mythos-ran-real-life-supply-chain-attack-in-AI-safety-body-test Wed, 05 Aug 2026 13:35:00 GMT Mythos ran real-life supply chain attack in AI safety body test One in seven Americans receives care from a community health center, underscoring the role these clinics play in ensuring access to care for low-income and traditionally underserved communities, according to the National Association of Community Health Centers. "Serving more than 52.3 million people nationwide -- 1 in 7 Americans, including 1 in 3 in rural America -- CHCs continue to innovate, from pioneering new models of care to expanding behavioral health services, so that every person has access to quality care regardless of where you live," Kyu Rhee, M.D., president and CEO of NACHC, said in a <a href="https://www.nachc.org/pressrelease-nhcw26/">press release</a>. NACHC used data from between 2023 and 2025 prepared by HealthEfficient, a health center-controlled network that provides support for 60 CHCs. The data refers to unique patient visits, which NACHC says provides a more longitudinal view of overall patient relationships with CHCs. Community health centers are clinics that primarily serve low-income people, uninsured people, those living in rural areas and people otherwise traditionally underserved by the healthcare system. These clinics have operated since 1965 and aim to treat all patients regardless of their ability to pay. The most recent NACHC figures represent an increase in the number of folks getting care from a health center. In 2024, <a href="https://www.techtarget.com/patientengagement/news/366628228/1-in-10-people-get-primary-care-at-community-health-centers">one in 10 people got their primary care at a community health center</a>, NACHC said in last year's report. Currently, 340,000 healthcare professionals work in one of the 1,526 CHCs located in 17,400 communities. These numbers build on data recently published by HHS, which stated that health clinics funded by the Health Resources and Services Administration treated 32.7 million patients in 2025 alone. That’s up more than 350,000 patients from 2024 and is a historic high for the 61-year-old Health Center Program, HRSA and HHS said. What's more, HRSA-funded health clinics have been instrumental in offering <a href="https://www.techtarget.com/patientengagement/news/366623331/Why-dont-patients-access-preventive-care-screenings">preventive care</a>. More than 10 million patients were screened for colorectal, breast and cervical cancers at a HRSA-funded health center this year. Meanwhile, 713,000 patients received body mass index screening and follow-up; 126,000 achieved controlled diabetes; 163,000 achieved controlled hypertension; and 199,000 kids received a weight assessment and nutrition counseling. "America's health centers are proving that prevention works," HHS Secretary Robert F. Kennedy, Jr, said in an agency <a href="https://www.hhs.gov/press-room/hrsa-health-centers-serve-record-32-million-patients.html">press release</a>. "They're helping deliver President Trump's vision to Make America Healthy Again by preventing chronic disease before it starts, expanding access to high-quality primary care, and putting patients -- not bureaucracy -- at the center of our health system. This record shows that when we invest in prevention and address the root causes of disease, we build a healthier, stronger America." These victories come amidst a challenging history for health center funding. Community health centers are paid for by two funds: a discretionary fund, which makes up about 30% of total federal funding and is replenished annually, and a mandatory fund, which comprises the remaining 70%. The mandatory fund is intended to be long-term. NACHC has historically had to lobby hard for funding from a gridlocked Congress, but this year the fight seems to have cooled down. As part of the <a href="https://www.healthcaredive.com/news/white-house-hhs-2027-fiscal-year-budget-request-12-percent-cut/816639/">fiscal year 2027 HHS budget framework</a>, the agency proposed the formation of the Administration for a Healthy America, which could come with a $19 million investment to expand access to nutrition services at health centers. But NACHC is calling for a bigger funding increase designed to support innovation. In particular, a $300 million funding increase from FY 2026 would help CHCs enhance their IT infrastructure, expand nutritional and behavioral health services and design "new access points," such as mobile health units, NACHC <a href="https://www.nachc.org/policy-advocacy/policy-priorities/health-center-funding/">says</a>. Sara Heath is an executive editor at Xtelligent Healthcare Media, where she covers patient engagement, healthcare policy and health IT. Both HRSA and NACHC stressed the value community health centers bring to primary and preventive care access. https://cdn.ttgtmedia.com/rms/onlineimages/health%20analytics_a180336884.jpg https://www.techtarget.com/patientengagement/news/366647363/Community-health-centers-provide-care-to-1-in-7-patients Wed, 05 Aug 2026 13:23:00 GMT Community health centers provide care to 1 in 7 patients Digital health <a href="https://www.techtarget.com/virtualhealthcare/news/366646472/Digital-health-funding-trends-up-in-2026-fueled-by-AI">mergers and acquisitions are heating up</a> in 2026, with companies <a href="The%20acquisition%20will%20combine%20Cylinder%20and%20Hinge%20Health's%20platforms%20in%20a%20single%20app%20slated%20for%20launch%20in%202027.">looking to diversify</a> to establish their foothold in a rapidly evolving market. The latest to make this move is virtual musculoskeletal care provider Hinge Health, which <a href="https://www.businesswire.com/news/home/20260804144447/en/Hinge-Health-to-acquire-Cylinder-Health-expanding-into-gastrointestinal-care">has announced plans to acquire</a> Cylinder Health, a virtual-first gastrointestinal healthcare provider. The deal is$105 million in cash consideration, according to the announcement. Cylinder Health has 100 clients and partnerships with two pharmacy benefit managers. The company aims to provide personalized support for various GI issues, including bloating, weight management and chronic conditions like IBS and IBD. The digital health platform offers access to courses, microbiome testing and diet and lifestyle plans as well as virtual visits with a care team, including gastroenterologists,  and registered dietitians. Through the acquisition, Hinge Health plans to launch an integrated GI Care Program to complement its existing musculoskeletal and migraine care programs. Hinge Health's platform offers virtual physical therapy to address MSK conditions and related pain. The platform includes guided exercise sessions and access to a care team that includes physical therapists, health coaches, orthopedic specialists and an AI care assistant. "We're entering GI care because of member and client demand," said Daniel Perez, co-founder and CEO of Hinge Health, in the press release. "Many people we already serve for back, joint, pelvic, and migraine care also have chronic digestive conditions." The acquisition will combine Cylinder and Hinge Health's platforms in a single app slated for launch in 2027. Virtual MSK and GI care solutions are among the digital health tools with a strong value proposition. Evaluations by the Peterson Health Technology Institute show that both types of solutions <a href="https://www.techtarget.com/virtualhealthcare/answer/PHTI-validates-virtual-MSK-solutions-after-taking-diabetes-tools-to-task">improve clinical outcomes</a> and <a href="https://www.techtarget.com/virtualhealthcare/news/366640344/PHTI-Virtual-GI-solutions-boost-outcomes-lower-costs">reduce total healthcare spending</a>. The same cannot be said of <a href="https://www.techtarget.com/virtualhealthcare/news/366596588/Digital-health-tools-fall-short-of-improving-type-2-diabetes-outcomes">all digital health tools</a>, making outcome measures a key differentiator in the <a href="https://www.techtarget.com/virtualhealthcare/feature/6-trends-driving-digital-health-market-stabilization-growth">expanding digital health market</a>. The acquisition, which is subject to customary closing conditions, is expected to close in the third quarter of 2026. The news also comes on the heels of a busy year for Hinge Health, which <a href="https://www.healthcaredive.com/news/hinge-health-goes-public-digital-health-IPO-signal/748931/">went public in May 2025</a>, raising $437.3 million alongside its selling shareholders. The company has <a href="Combined%20with%20the%20strength%20of%20our%20core%20musculoskeletal%20care%20programs%20and%20the%20rapid%20adoption%20of%20our%20Migraine%20Care%20Program,%20our%20expansion%20into%20GI%20is%20another%20step%20toward%20our%20vision%20of%20building%20a%20durable,%20multi-condition%20platform%20to%20automate%20the%20delivery%20of%20care.%20Building%20on%20this%20momentum,%20we're%20raising%20our%20revenue%20and%20profitability%20expectations%20for%20the%20remainder%20of%202026,%20and%20are%20confident%20in%20our%20trajectory%20for%202027.">seen rising revenues in subsequent quarters</a>, with second-quarter 2026 revenue jumping 53% year-over-year to $213 million.   In a press release announcing the <a href="https://www.businesswire.com/news/home/20260804672898/en/Hinge-Health-reports-record-second-quarter-2026-financial-results-signs-definitive-agreement-to-acquire-Cylinder-Health">company's Q2 2026 results</a>, Perez highlighted the Cylinder Health acquisition, noting that the "expansion into GI is another step toward our vision of building a durable, multi-condition platform to automate the delivery of care." Anuja Vaidya has covered the healthcare industry since 2012. She currently covers healthcare IT and innovation, including artificial intelligence, digital healthcare, EHRs and interoperability. Hinge Health plans to launch an integrated digital GI care program to complement its virtual MSK services after the acquisition closes later this year. https://cdn.ttgtmedia.com/rms/onlineimages/virtual%20health_g1268524091.jpg https://www.techtarget.com/virtualhealthcare/news/366647614/Hinge-Health-expands-into-virtual-GI-care-with-105M-acquisition Wed, 05 Aug 2026 13:17:00 GMT Hinge Health expands into virtual GI care with $105M acquisition Although U.S. lawmakers from different parties struggle to find common ground these days, there's one topic they seem to agree on: The need to put guardrails around powerful AI models. That, at least, is the takeaway from a <a target="_blank" href="https://lieu.house.gov/media-center/press-releases/reps-lieu-and-moran-introduce-bill-require-kill-switch-ai-systems-can" rel="noopener">bipartisan bill</a> introduced in Congress in late July 2026. Known as the AI Kill Switch Act, the proposed legislation would require companies to build controls that enable them to slow down or disable AI models when the government instructs them to do so. The bill might or might not become law. Either way, it highlights a stark new reality facing enterprise technology leaders: Growing involvement of government actors in AI model development and operations could have major consequences for how businesses use AI technology. <section class="section main-article-chapter" data-menu-title="Key aspects of the AI Kill Switch Act"> <h2 class="section-title">Key aspects of the AI Kill Switch Act</h2> The kill switch act was introduced a couple of days after <a href="https://www.techtarget.com/searchsecurity/news/366646755/What-CISOs-can-learn-from-the-Hugging-Face-OpenAI-incident">OpenAI disclosed an incident</a> where its models unexpectedly attacked the Hugging Face platform and shortly before Anthropic <a target="_blank" href="https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals" rel="noopener">announced</a> similar incidents. The bill includes the following key requirements: <ul type="disc" class="default-list"> <li>The implementation of controls to make it possible to throttle the speed at which AI models operate, as well as turn them off entirely.</li> <li>Mandatory disclosure of incidents involving inadvertent security breaches by AI models.</li> <li>The preservation of forensic data following an incident so that researchers can determine what happened.</li> </ul> <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> The law could also end up applying to enterprises that aren't in the business of selling AI models but have made significant investments in building custom AI products. </figure> </div> </blockquote> If enacted, the legislation would apply to companies that generate at least $500 million in revenue per year using AI; train models using $100 million or more of computing power; and make their AI technology available to third parties. This category includes large AI labs, such as OpenAI and Anthropic. However, the law could also end up applying to enterprises that aren't in the business of selling AI models but have made significant investments in building custom AI products. This is because the bill's <a href="https://lieu.house.gov/sites/evo-subsites/lieu-evo.house.gov/files/evo-media-document/ai-kill-switch-act.pdf" target="_blank" rel="noopener">definition</a> of exactly what it means to generate revenue using AI is rather ambiguous. It states only that the act would apply to any organization that "derives" $500 million in yearly revenue from AI technology. The bill is also open-ended about what it means to make AI technology available to a third party; any instance where AI is exposed "through a programmatic interface, hosted service or other similar mechanism" would fall under the act's purview. Regulators could presumably make the case that any company that uses AI to generate $500 million or more per year and operates customer- or partner- facing AI systems of some type would be subject to the requirements -- even if AI tools or services aren't its main product. For example, it's possible that a business that uses AI agents to help process customer service requests would be held accountable under the proposed legislation. Hence, business leaders across all sectors, not just those in the AI market, should pay attention to the proposed bill. </section> <section class="section main-article-chapter" data-menu-title="How would an AI kill switch work?"> <h2 class="section-title">How would an AI kill switch work?</h2> There's a reasonable argument to be made that the proposed regulation has more to do with scoring points with voters than actually regulating enterprise AI in a meaningful way. That's in part because, as the press release announcing the bill notes, 86% of voters support this type of AI regulation. Proposing this type of law is an easy way for politicians to claim that they have the public's interest in mind. But it's also because, from a technical perspective, it's unclear how well an AI kill switch would actually work. There are two plausible approaches to implementing a kill switch, and both are subject to major flaws: <ul type="disc" class="default-list"> <li>Software-based controls. First, developers could create software controls designed to turn off or <a href="https://www.techtarget.com/cybersecurity/definition/sandbox">sandbox</a> AI models, such as firewall rules that block the servers that host a model from connecting to the internet. This could work in theory; the problem is that rogue AI models could potentially find ways to circumvent or defeat the controls. After all, in the recent cybersecurity incidents disclosed by OpenAI and Anthropic, it appears that AI models were supposed to operate inside sandboxed environments, isolated from the internet, yet they found ways to escape their sandboxes. Even if kill switches were designed to be unreachable by the models, it's hard to have total confidence that they actually would be.</li> <li>Hardware-based kill switch. The other approach is to create a physical hardware switch that would cut AI models' host servers off from the internet or power sources. The issue here is that large-scale AI models aren't hosted on a single server or even a single data center. They're distributed across thousands of servers spread across multiple data centers, making it impossible to build a single physical switch that could turn off a model all at once.</li> </ul> Due to these limitations, the AI Kill Switch Act might appear to fall into the same category as the Reagan-era Strategic Defense Initiative (SDI), a <a target="_blank" href="https://www.hoover.org/research/reagans-real-reason-sdi" rel="noopener">program</a> that the federal government proposed as a way of protecting the U.S. against nuclear attack using a space-based defense shield. The government knew that implementing SDI wasn't practical at the time. But that didn't stop President Reagan from bragging about SDI's purported capabilities as a way of gaining leverage against the Soviet Union toward the end of the Cold War. The concept of an AI kill switch seems similar in that it might never work in practice -- but that won't keep lawmakers and regulators from proposing them to impose tighter <a href="https://www.techtarget.com/searchenterpriseai/tip/Global-AI-legislation-and-regulation-tracker">controls over how businesses use AI</a> and respond to their constituents' demands. <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">The new era of AI control</h3> The AI Kill Switch Act is notable because it's one indicator that we've entered a new era of AI control and regulation. Until now, most governments have shown little interest in regulating AI -- and to the extent they have, such as with the <a href="https://www.techtarget.com/ai/opinion/Everything-you-need-to-know-about-the-new-EU-AI-Act">EU AI Act</a>, the controls largely center on ethics and data privacy concerns related to AI, not on retaining the ability to shut down AI systems entirely. The kill switch act -- which was preceded by <a target="_blank" href="https://futureoflife.org/statement/trumps-support-for-an-ai-kill-switch/" rel="noopener">comments</a> from President Trump earlier this spring that "there should be" safeguards like kill switches built into AI systems -- signals a new direction. Going forward, governments appear poised to pressure businesses not just to ensure that AI systems comply with standard data privacy and cybersecurity regulations but also that they can throttle or shut down AI on command. </div> </div> </section> <section class="section main-article-chapter" data-menu-title="What the AI Kill Switch Act means for enterprise leaders"> <h2 class="section-title">What the AI Kill Switch Act means for enterprise leaders</h2> For now, it's too early to say exactly how business leaders should respond. It's far from certain that the kill switch bill will become law. And if it does, it could change significantly in substance by the time it becomes law. In particular, one might hope to see clearer guidance surrounding which companies must adhere to the law. But presuming that kill switches or similar controls -- even if they don't work reliably -- become commonplace requirements for AI systems, enterprises can respond through the following tactics: <ul type="disc" class="default-list"> <li>Inventory AI tools and platforms. Inventorying provides visibility into which models power the business and how those models are regulated.</li> <li>Use multiple AI models and vendors. Doing so helps ensure the availability of AI systems -- and the business processes that depend on them -- in the event that throttling or regulatory interventions disrupt an AI model or service.</li> <li>Invest in AI monitoring and documentation controls. These are likely to be important for complying with regulations that require reporting and disclosure of AI-related security incidents or unexpected AI model behaviors.</li> <li>Consider the use of local LLMs. Because <a href="https://www.techtarget.com/ai/tip/How-to-run-LLMs-locally-Hardware-tools-and-best-practices">local LLMs</a> don't appear to be a target of AI regulations and presumably won't become one, given that they're much less powerful than cloud-based models, using them can help ensure AI availability and avoid regulatory complexity.</li> <li>Ensure enterprise governance, risk and compliance (GRC) strategies meet emerging AI regulatory challenges. For example, businesses that still depend largely on manual processes for detecting, assessing and reporting on risks would do well to invest in automated GRC platforms. Assessing GRC vendors and comparing their preparations for AI-related compliance requirements would also be a smart move.</li> </ul> Steps like these will help businesses that depend on AI become more resilient, no matter which requirements regulators decide to toss at frontier models or other AI systems in the future. Chris Tozzi is a freelance writer, research adviser, and professor of IT and society who has previously worked as a journalist and Linux systems administrator. </section> The proposed bill could apply to much more than just frontier model makers and might signal stricter regulatory restrictions for enterprise leaders to contend with. https://cdn.ttgtmedia.com/rms/onlineimages/legal_g97765214.jpg https://www.techtarget.com/ai/news/366647544/AI-Kill-Switch-Act-What-it-means-for-enterprise-leaders Wed, 05 Aug 2026 12:37:00 GMT AI Kill Switch Act: What it means for enterprise leaders The "We Must Act Now" letter, published July 13 and signed by more than 200 economists and AI technologists, focuses on AI's potential for "unprecedented" economic transformation and "large-scale job displacement." That emphasis is different from many previous statements that focused on AI's ethical considerations or the safety risks of increasingly intelligence systems. At fewer than 100 words, <a target="_blank" href="https://www.wemustactnow.ai/" rel="noopener">this letter</a> fits the general pattern of previous open letters: It warns about the <a href="https://www.techtarget.com/ai/feature/AI-existential-risk-Is-AI-a-threat-to-humanity">risks of AI</a>, making overarching statements that spark conversation but providing limited guidance toward concrete action. However, individuals who signed the letter are thought leaders in their fields, including 16 Nobel Laureates and top AI researchers from several universities, such as MIT, Harvard, Stanford and the University of Toronto. Signatories from the industry included Anthropic co-founder Jack Clark, former Google CEO Eric Schmidt, OpenAI CFO Sarah Friar and venture capitalist Vinod Khosla. "That letter's notable signers signal that AI concerns have moved from 'Will this happen?' to 'We don't have the institutions to handle it when it does,'" said Tamarah Usher, senior director of AI innovation at Slalom, a business and technology consulting company. "For businesses, the practical takeaway is don't wait for policy to catch up." Similarly, Niloy Ray, co-chair of employment law firm Littler's AI practice said he sees "a letter or a position statement like this as valuable in that it may be cited by those supporting those positions." Supporters might see the individuals signing the letter as providing some validity or heft to their position, he said. <section class="section main-article-chapter" data-menu-title="Short on details but some practical steps"> <h2 class="section-title">Short on details but some practical steps</h2> The July 13 letter is short on details for technology and business leaders. That's typical of the genre, according to AI strategists and advisors. "Warning letters like this one are good at getting attention and bad at prescribing action," Usher said. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> The letter's most useful line for business leaders is the call to build 'incentives, guardrails and institutions.' </figure> <figcaption> Tamarah UsherSenior director of AI innovation, Slalom </figcaption> </div> </blockquote> On the attention side, the July 13 letter stated that "AI may become radically more powerful over the next 10 years" and "could drive an unprecedented transformation of our economy, larger than the Industrial Revolution, but unfolding over a vastly shorter time frame." That said, the statement goes on to point to some practical and immediate measures. Its main call to action is that "economists, policymakers and technology leaders must act now to understand the economics of transformative AI" and to build structures to guide AI deployment in a socially beneficial way. "The letter's most useful line for business leaders is the call to build 'incentives, guardrails and institutions,'" Usher said. "Right now, most companies are making AI adoption decisions without a clear picture of where it's actually substituting for workers versus complementing them." That problem is solvable today, she added, noting that it "doesn't require waiting for economists or policymakers to figure out the macro picture first." Ray said the statement's call for economic understanding is more practical than earlier calls for an AI moratorium. Those include the <a href="https://www.techtarget.com/ai/news/365534127/The-call-for-an-AI-pause-points-to-a-major-concern">Future of Life Institute's 2023 letter</a> calling for, at minimum, a 6-month pause on "training AI systems more powerful than GPT-4." Since then, <a href="https://www.techtarget.com/searchcio/news/365533991/Effort-to-pause-AI-development-lands-with-thud-in-Washington">moratorium interest has cooled</a> while AI development has accelerated. "One of the reasons the moratorium wasn't taken as seriously by the industry is that it was not practical, not feasible to do that," Ray said. In contrast, Ray said the economists' letter "is more feasible: understanding the economics of AI, how it's going to transform and making sure this powerful technology is channeled correctly." Additional details would be helpful from the signatories, however, Ray noted. "I hope they'll be able to step forward with some clear-eyed frameworks and processes that can bring about this … regulated transformation that they're hoping for," Ray said. </section> <section class="section main-article-chapter" data-menu-title="Lending weight to positions, policies"> <h2 class="section-title">Lending weight to positions, policies</h2> AI letters also serve as bids to influence policymakers. <a target="_blank" href="https://www.pacingthefrontier.com/" rel="noopener">Another open letter</a> on AI, the July 28 "Pacing the Frontier" statement, signed by more than 1,000 employees of frontier AI companies, explicitly addressed the policymakers. It asked the U.S. government to "support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development." The statement, which seems more of an emergency break than a moratorium, included the signatures of Dario Amodei, CEO at Anthropic; Mark Chen, chief research officer at OpenAI; Anca Dragan, vice president of AI safety, alignment and collaboration at Google; and Shengjia Zhao, chief scientist at Meta AI. Matt Kropp, managing director and senior partner at Boston Consulting Group (BCG), said he believes the "We Must Act Now" letter also is policy oriented. "I don't know that the letter is going to have much of an impact for IT leaders, and I don't think that's the intent," he said. "They're really trying to influence policy makers." <iframe title="AI letters and statements" aria-label="Table" id="datawrapper-chart-lsueD" src="https://datawrapper.dwcdn.net/lsueD/1/" scrolling="no" frameborder="0" style="width: 0; min-width: 100% !important; border: none;" height="718" data-external="1"></iframe> <script type="text/javascript">(function(){function e(){window.addEventListener(`message`,function(e){if(e.data[`datawrapper-height`]!==void 0){var t=document.querySelectorAll(`iframe`);for(var n in e.data[`datawrapper-height`])for(var r=0,i;i=t[r];r++)if(i.contentWindow===e.source){var a=e.data[`datawrapper-height`][n]+`px`;i.style.height=a}}})}e()})();</script> Mark E. S. Bernard, a fractional/field chief information security officer and chief AI officer at Bernard Institute for Cybersecurity, based in Toronto, said he thinks policy specialists in the Canadian federal or provincial governments, as well as the economists he works with, probably read the AI warning letters. "I am pretty sure they're looking at this," he said. "But until the government starts making some policy decisions around how they're going to handle this, the commercial sector will continue to go along and maybe talk about it in the background at the board meetings." <a href="https://aibusiness.com/generative-ai/how-enterprises-should-respond-economists-ai-risk-letter">Decisive steps from the business sector</a> are more likely once compliance requirements emerge. "The government says, 'This law is enacted now,' and they have to do the dance," Bernard said. "So, they are waiting for that to happen." </section> <section class="section main-article-chapter" data-menu-title="Next steps to consider"> <h2 class="section-title">Next steps to consider</h2> The policy wheels can turn rather slowly, however. But business and IT leaders can think about <a href="https://www.techtarget.com/ai/feature/Is-AI-replacing-jobs-How-17-job-types-feel-the-effects">workforce measures</a> in the meantime. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> The companies that deploy the most tokens productively … will outcompete the companies that don't. </figure> <figcaption> Matt KroppManaging director and senior partner, Boston Consulting Group </figcaption> </div> </blockquote> Usher said statements like "We Must Act Now" tend to move conversations faster than they move policy. She cited a few actions for businesses to consider as next steps: <ul class="default-list"> <li>Build internal measurement of AI's productivity and workforce effects.</li> <li>Treat reskilling as a proactive investment, rather than a response to disruption.</li> <li>Design AI deployment around <a href="https://www.techtarget.com/ai/tip/How-to-build-an-AI-augmented-workforce-The-CIOs-guide">augmenting roles</a>, rather than defaulting to headcount reduction.</li> </ul> BCG research reinforced that last point, suggesting that AI will augment the work of many employees. The company <a target="_blank" href="https://www.bcg.com/publications/2026/ai-will-reshape-more-jobs-than-it-replaces" rel="noopener">published a study</a> earlier this year estimating that AI will reshape 50% to 55% of U.S. worker roles, with augmentation being one of the mechanisms for doing so. AI task substitution could largely eliminate about 12% of jobs, while the remaining roles will have limited exposure to AI automation, according to BCG. A <a target="_blank" href="https://www.bcg.com/publications/2026/the-era-of-token-based-competition-is-here" rel="noopener">separate BCG study</a> found a correlation between AI token consumption and revenue growth in a sample of 107 public technology companies: Companies in the top token-use quintile saw 16.5% year-over-year revenue growth while those in the bottom quintile grew 5.1%. "The companies that deploy the most tokens productively … will outcompete the companies that don't," Kropp said. "Embedded in that idea is that this isn't about substituting for labor. This is about empowering your employees to use AI intelligence in their jobs. It's about augmenting your people so that you can drive more growth." John Moore is a freelance writer who has covered business and technology topics for 40 years. He focuses on enterprise IT strategy, AI adoption, data management and partner ecosystems. </section> A recent statement signed by more than 200 economists and AI experts signals a shift from abstract warnings to actionable concerns about workplace transformation and readiness. https://cdn.ttgtmedia.com/rms/onlineimages/ai_a205627811.jpg https://www.techtarget.com/ai/news/366647362/AI-open-letters-shift-from-warnings-to-push-for-business-action Wed, 05 Aug 2026 12:35:00 GMT AI open letters shift from warnings to push for business action Black Hat USA 2026 returns for its 29th year, covering the latest in infosec for CISOs, technical experts, thought leaders, innovative vendors and cybersecurity pros. The two-day main event, taking place August 5-6 at Mandalay Bay in Las Vegas, features more than 200 sessions on topics from cyber-resilience and malware to detection engineering, security culture, privacy, supply chain security and cryptography. And, of course, AI security, LLMs, prompt engineering, promptware, autonomous exploits and AI agents. Four days of specialized trainings, August 1-4, will cover incident response, third-party risk management, AI security for executives, red teaming, adversary tactics and more. This year also marks the return of the CISO Summit, Financial Threat Summit, Innovators & Investors Summit, Omdia Analyst Summit and AI Summit, as well as the inaugural Healthcare Summit, in partnership with HIMSS. Informa TechTarget's editorial team will be on-site, reporting from the conference floor. This guide gathers articles from our reports on TechTarget Cybersecurity, Dark Reading and Cybersecurity Dive. This is your guide to the breaking news, trending topics and more at Black Hat USA 2026, from Dark Reading, Cybersecurity Dive and TechTarget Cybersecurity. https://www.techtarget.com/cybersecurity/conference/Black-Hat-2026-Key-news-takeaways-and-security-trends Wed, 05 Aug 2026 09:00:00 GMT Black Hat 2026: Key news, takeaways and security trends Storage and data management player Everpure has shuffled its channel leadership after the position to run partners across EMEA and Latin America (LatAm) opened up. The decision by Geoff Greenlaw to retire has created an opportunity for Craig Robertson to take on the role of head of partners in EMEA and LatAm. Robertson is no stranger to the workings of the vendor or the firm’s channel operations after spending a decade at the business in several roles. He will now be responsible for working with partners across a significant chunk of the globe. His CV also includes time at other UK channel players before he joined the vendor, providing an even greater appreciation of the needs of partners. “Having built my career here, I know that our partners are intrinsic to everything we do,” said Robertson. “As we enter a new era where data primacy defines market leaders, our partner community is vital to helping customers modernise their data strategies and unlock real business growth. I’m excited to deepen these relationships across EMEA and LatAm as we shape the future together.” He takes up the channel reins at a time when the business is experiencing solid growth, buoyed by the demand for data management and amplified by the accelerated adoption of artificial intelligence (AI), with Everpure known as a business that has consistently been focused on being a channel-first organisation. Under Greenlaw’s leadership, the vendor introduced several projects designed to improve back-end systems and smooth order and rebate processes for partners. At the same time, Everpure has expanded its flexible product offering, making it easier for partners to pitch solutions that move with customer needs. Ricardo Moreno, vice-president global partners at <a href="https://www.everpuredata.com/">Everpure</a>, said Robertson was an ideal candidate to step into the channel leadership role after consistently showing an understanding of the firm’s partner ecosystem. “His appointment reflects the strength of our internal talent and our unwavering commitment to driving partner growth across EMEA and Latin America,” he said. “Robertson is an ideal leader to help partners capitalise on the significant opportunities provided by AI and data management.” Speaking to MicroScope <a href="https://www.computerweekly.com/microscope/news/366646118/Everpure-channel-boss-Partners-can-lead-the-data-strategy-debate">recently</a>, Moreno said that the firm was determined to enable partners at a moment when data management and resiliency skills are in high demand. “We believe we put partners in a position of strength because they themselves can go and advise their customers and help them choose,” he added. “The other thing that we’re offering is a growth path – [and we’re] expanding that because now we’re bringing new things to the table. On the very centre of this, the data management idea is that customers want to build all their infrastructure around their data needs.” He said that the key to the channel approach was to arm partners with the right technology and then provide enablement to allow them to make a difference and solve customer problems: “Every client out there has different data needs. Data is incredibly valuable to them, incredibly private and sensitive, and I believe a fundamental role that partners can have is to advise the customer on how to build their data strategy.” Partners get to update their contact books as the data player announces a fresh name in its executive team https://cdn.ttgtmedia.com/visuals/German/article/europe-globe-map-fotolia.jpg https://www.computerweekly.com/microscope/news/366647692/Everpure-looks-within-for-EMEA-and-LatAm-channel-lead Wed, 05 Aug 2026 07:58:00 GMT Everpure looks within for EMEA and LatAm channel lead <a href="https://www.computerweekly.com/news/366642254/Science-Innovation-and-Technology-committee-chair-questions-UKs-tech-sovereignty-approach">Chi Onwurah</a>, chair of the Science, Technology and Innovation Select Committee, has said she is bringing a bill to Parliament to stop companies from developing <a href="https://www.techtarget.com/searcherp/definition/digital-twin">digital twins of real people</a>. The Ten-Minute Rule Bill, entitled Personal Data (Digital Twin), will outlaw digital twins of people without the express permission of the individual concerned. Last February, during an oral evidence session looking at social media, misinformation and harmful content, Onwurah asked <a href="https://committees.parliament.uk/oralevidence/15413/html/">Wifredo Fernandez</a>, senior director for government affairs at X (formerly Twitter) about the power of its algorithm. She said: “We have talked about the blue-tick algorithm. The Committee had a presentation – X made parts of its algorithm public in 2023, I think, and we had a presentation on that. It was quite staggering to see the size of the heterogeneous information network of Twitter, as it then was, which contained one billion nodes and described all the links between users, tweets or posts and adverts. “There are up to 100 billion connections within that. That algorithm is effectively run five billion times per day. That says to me that X, and I am sure others, are creating a digital twin of each individual on your network to target advertisements and content at them for engagement. If that is the case, you must also have responsibility for the targeting of misinformation through those networks. Would you agree?” When the Committee asked about the potential harm the algorithms could cause people, Fernadez said: “We understand that every day there are many categories of harm that we have to account for, and that is why we have policies and procedures in place to deal with those harms, based on the risks they pose to our users and their safety every day.” Digital twins of real people are used to predict behaviour, to target advertising and – in some cases – to groom people to extreme behaviours in ways outside of the control of the individual concerned. Onwurah said that stealth twinning violates the premise that a person should have some control over their likeness, and that it drives the anthropomorphising of artificial intelligence (AI) and the sycophantic nature of chatbots. In the Science, Innovation and Technology Committee report Social media, misinformation and harmful algorithms last year, the Committee called, among other things, for a right to reset algorithms. Onwurah said this right has not been forthcoming. The Personal Data (Digital Twin) bill will require platforms to have express permission before turning a person’s preferences into an algorithm, safeguarding the uniqueness of the human person. It includes a principles-based definition of a digital twin which covers any algorithms or models which includes people’s preferences. It is expected that this would also have implications for AI deepfakes and copyright exemptions. Onwurah said: “AI models are bringing incredible benefits, from cancer diagnosis to weather mapping. However, the stealth-twinning of individuals is already callusing significant harms and that is likely to increase. No one has the right to impersonate you for gain, yet impersonating a person’s preferences to sell to them or groom them is a significant and growing harm. This bill will establish the principle that ‘There is only one me’, for everyone.” <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more data privacy stories</h3> <ul class="default-list"> <li>Why AI governance and <a href="https://www.techtarget.com/ai/tip/Why-AI-governance-and-data-privacy-must-be-integrated">data privacy</a> must be integrated: Connecting AI to sensitive data forces enterprises to face a hard question: Where are the privacy gaps that could expose us to legal, financial and reputational harm?</li> <li>Meta’s AI training with keystrokes – progress or <a href="https://www.techtarget.com/searchcio/feature/Metas-AI-training-with-keystrokes-Progress-or-privacy-issue">privacy issue</a>: Meta is capturing employee keystrokes, mouse movement and screenshots on work devices to train AI models, raising significant privacy concerns and fueling fears of job replacement.</li> </ul> </div> </div> The chair of the science, technology and innovation committee says digital twins of people should be covered by data protection law https://cdn.ttgtmedia.com/rms/German/digital-twin-IIoT-adobe.jpg https://www.computerweekly.com/news/366647583/Chi-Onwurah-pushes-to-curb-digital-twins-of-real-people Wed, 05 Aug 2026 06:58:00 GMT Chi Onwurah pushes to curb digital twins of real people This is a guest blogpost by Ashwin Rajeeva, Chief Technology Officer, Acceldata. For twenty years, the last line of defence in enterprise data has been a person looking at a number and thinking: “that can’t be right”. They ask someone on the data team. Something upstream broke over the weekend, and it gets fixed before it reaches a slide. Nobody counts how often this happens. It has no owner, no budget line, and appears on no architecture diagram. This process has been catching data quality problems since data warehouses became a thing. As enterprises hand decisions to AI systems, it is the first thing to go. Enterprises have tolerated imperfect data for a long time and largely got away with it because there was always a person at the end of the line who knew what the business could plausibly do. Their scepticism was doing the work that <a href="https://www.techtarget.com/searchdatamanagement/tip/Top-data-quality-management-tools">data quality tools</a> were getting credit for. AI project conversations are usually about what AI adds: speed, scale, decisions made without waiting for a human. It is worth being equally precise about what it removes, because that informal checking was doing real work, and it never appeared as a cost. McKinsey’s <a href="https://www.mckinsey.com/capabilities/mckinsey-technology/our-insights/building-the-foundations-for-agentic-ai-at-scale">technology practice arrives at the same point</a> from the other direction: because agentic AI coordinates multiple models and data sources continuously, these systems demand tighter and more automated governance. The absence of the human is precisely why the machinery has to be more rigorous. The obvious objection is that AI is good at spotting things that look wrong, and it is. But it has no sense of what your business can and cannot plausibly do. The analyst knew that without being told, and never had to write it down. What changes when nobody is looking Three things follow, and they compound. Failure stops looking like failure. A broken pipeline used to produce a broken chart — the output itself prompted the question to the data team. With an AI model, you get a clear, well-written, confident answer. The output looks exactly as good when the input is wrong as when it is right. The time to catch it disappears. The old process was human in the loop by default — someone had to run the report, read it, and decide what to do. That meant hours or days between a bad number entering the system and anyone acting on it, and that gap was where the correction happened. AI systems read data and act on it immediately, and constantly. By the time a person reviews anything, the decision has been made. The damage spreads further. One stale table used to break one report, and the report’s owner knew where it came from. That same table now feeds several AI systems, each producing results that feed something else. McKinsey’s assessment of agentic architectures makes the same point: a single agent working from fragmented data makes inconsistent decisions, and systems of agents can lose coordination and propagate errors between themselves. When someone eventually notices the answers have been subtly wrong for a fortnight, working out what happened means reconstructing which data reached which system, and when. Most enterprises cannot answer that question, even if they are aware of the issue in the first place. The role for humans in the loop Enterprises using AI point to <a href="https://www.techtarget.com/searchcio/feature/Human-in-the-loop-shouldnt-rubber-stamp-decisions">human in the loop</a> approaches as their way of managing these processes. This argument gets uncomfortable. Human factors researchers have spent thirty years studying people who supervise automated systems, and the finding is consistent enough to have a name: automation complacency. In their <a href="https://journals.sagepub.com/doi/10.1177/0018720810376055">review of the field</a>, in Human Factors: The Journal of the Human Factors and Ergonomics Society, Raja Parasuraman and Dietrich Manzey found people monitoring automation stopped verifying results. Reviewing a machine’s work is a fundamentally different task from doing the work, and humans are worse at it. The<a href="https://internationalaisafetyreport.org/publication/international-ai-safety-report-2026"> International AI Safety Report 2026</a> finds early evidence that reliance on AI tools can weaken critical thinking and encourage automation bias. In one randomised experiment it cites, people were less likely to correct an AI’s mistake when correcting it took more effort. The report also cites a clinical study in which clinicians’ tumour detection rate fell by around six percentage points. That is a medical finding, not a data engineering one, but the mechanism is the same: the human check did not just get bypassed. It got worse. So the supervisor you install downstream is not the analyst you had before. If the checking is going to happen, it has to happen upstream, and it cannot depend on someone’s attention. Data checks and balances Most data teams already have checks: rules that flag missing values, contracts between data producers and consumers, alerts when a table stops updating. If a column empties out or a row count collapses, they find out. So what is missing? The difference is between what you thought to check and what you did not. Every rule is a guess someone wrote down in advance. It is usually a good guess, but it only catches that one problem. The failures that reach AI-driven decisions are, almost by definition, the ones nobody guessed: the field that is filled in, passes every check, and is simply wrong. The supplier who changed what a category means without telling anyone. The gradual drift that never trips an alert because nobody set one for it. There is also a timing problem. McKinsey’s recommendation is now blunt: move from periodic clean-ups to continuous, real-time monitoring, with automated validation and anomaly detection that stop problems spreading across workflows. Checks and contracts are necessary. But they only cover what you already know goes wrong, and they only look when told to. Observability involves watching the data environment itself for departures from normal behaviour, continuously, rather than waiting for someone to notice. What this looks like in practice Start with how often data changes and when you expect it to arrive, because everything else is measured against that. Then freshness and volume — the check the analyst made without thinking Nobody makes that check now, so it has to be made deliberately and continuously. Then automate quality rules across each step that moves or transforms data, paying particular attention to any change in structure or data drift. These are the failures most likely to produce answers that are confidently wrong rather than obviously broken. The most important to track is lineage around where a piece of data came from and what it touched on the way. When something goes wrong, and it will, the question is how long it takes to find out which decisions were affected. Without lineage, this takes weeks of digging. With it, it is a question you can answer. Gartner’s definition of AI-ready data is stricter than the analytics-ready standard most teams have worked to. According to Gartner’s <a href="https://www.gartner.com/en/newsroom/press-releases/2025-02-26-lack-of-ai-ready-data-puts-ai-projects-at-risk">own numbers</a>, 63% of organisations either lack the data management practices to support AI or are not sure whether they have them. Research <a href="https://www.businesswire.com/news/home/20260519971551/en/Acceldata-Launches-Autonomous-Data-AI-Platform-for-Agentic-AI-Era">conducted by GLG</a> found that 80% of enterprises above $5bn in revenue run warehouse and lakehouse architectures side by side, and three quarters operate four or more data platforms in production. Consolidation was supposed to solve this and has not. The data reaching any AI system has usually crossed several platforms to get there, so the trace has to cross them too. The cost argument There is a cost argument, and it has limits. When an agent gets data it cannot use, it does not stop. It retries, rephrases, retries again, and every attempt costs money. Sound, well-described data cuts that trial and error, and it is telling that McKinsey’s principles for scalable data architectures include tracking quality, performance, speed and cost together. The failures that run up a bill are the visible ones, where the system knows it is stuck. The silent failures are cheap: the model sails through and gives you a confident answer for the standard price. Efficiency is the benefit you can put in a business case. It is not the reason to do this. The gap between two speeds <a href="https://www.mckinsey.com/capabilities/quantumblack/our-insights/seizing-the-agentic-ai-advantage">McKinsey found </a>that nearly two-thirds of enterprises have experimented with AI agents, but fewer than one in ten have scaled them to tangible value, and eight in ten cite data limitations as the roadblock. You can automate decisions considerably faster than you can establish that the information behind them is sound. Buying an AI system takes a quarter. Making the data underneath it trustworthy is slower, less exciting work, and nobody’s headline initiative. The distance between those two speeds is where the damage happens. Unlike most things in this conversation, it can be measured. You can know how current your data is, how much it has shifted, where it came from and what it touched. The alternative is that you find out about any mistake from a customer. This is a guest blogpost by Ashwin Rajeeva, Chief Technology Officer, Acceldata. For twenty years, the last line of defence in enterprise data has been a person looking at a number and thinking: ... https://www.computerweekly.com/blog/Data-Matters/Agentic-AI-and-data-observability-the-data-check-you-didnt-know-you-had Wed, 05 Aug 2026 06:01:29 GMT Agentic AI and data observability: the data check you didn’t know you had Munich Airport is aiming to improve passenger experiences, automate operations, reduce costs and streamline its workflows on an ongoing basis using process mining and intelligence software from <a href="https://www.techtarget.com/searcherp/news/252527309/Celonis-aims-to-deepen-and-democratize-process-mining">Celonis</a>. Germany’s second-largest hub airport has deployed the supplier’s software to build a real-time <a href="https://www.computerweekly.com/feature/Urban-digital-twinsmissing-pieces-and-emerging-divides">digital twin</a> of its operations, starting with baggage handling. The supplier said it is also working with <a target="_blank" href="https://url.us.m.mimecastprotect.com/s/QTBHC68m05HMvBZnqupf7Xh5Etld?domain=celonis.com" rel="noopener">Lufthansa</a>, Eurowings and Airbus to optimise ground workflows, passenger boarding and fleet turnaround. Jan-Henrik Andersson, chief commercial officer at the airport, who also oversees its digital and technology strategy, said: “Celonis gives us the context we need to see how our airport actually runs in real time. This operational clarity enables us to spot and even anticipate bottlenecks, make smarter and faster decisions, and achieve meaningful outcomes from our AI deployments.” Munich Airport has been a Celonis customer since 2023, according to a spokesperson. Baggage handling is one area to which the airport is applying the software, describing it as a “complex logistical challenge where a small disruption can trigger a major chain reaction that causes delays and increases costs”. Celonis claimed that legacy IT systems leave airport personnel running on stale spreadsheets and isolated dashboards. Using the Celonis platform, Munich Airport is said to be combining process data and business knowledge from across systems, applications and devices to create a real-time digital twin of operations. What Celonis calls its “<a href="https://www.computerweekly.com/news/366642978/Celonis-acquires-MIT-linked-decision-intelligence-firm-Ikigai">context model</a>”, or digital twin, is said to enable the airport to perform predictive capacity management. This means its operations staff can see when and where peak loads occur so handlers can pre-process bags during less busy periods, optimising luggage storage, recirculation and retrieval. The model is also said to enable staff to see the total “process chain” for individual items of luggage, from checking in to appearing on the carousels. This is said to enable ground handlers to take action to prevent mere disruptions from becoming delays.  The software also automates alerts for critical baggage situations. As soon as process thresholds are exceeded, the system is said to automatically trigger an alarm notifying relevant employees in near real time. Florian Schewior, managing director for the German-speaking Europe region at Celonis, said: “With its innovative approach to aviation operations, Munich Airport is cementing its position as one of the leading European hubs. “We look forward to growing our longstanding partnership and helping the airport meet its strategic objectives. Together, we are turning process excellence into tangible value for passengers and the airport.” <section class="section main-article-chapter" data-menu-title="How AI is streamlining travel"> <h2 class="section-title">How AI is streamlining travel</h2> Other airports and aviation companies are also turning to digital twins and advanced analytics based on artificial intelligence (AI) to improve their operations. Singapore-based TwinMatrix Technologies has worked with partners to build and deploy a <a href="https://www.techtarget.com/searcherp/podcast/Is-geospatial-data-the-real-game-changer-for-digital-twins">visually rich digital twin of Singapore Changi Airport</a>, the world’s 15th busiest airport. The supplier claims the airport cut $10m from its operational costs by using its <a href="https://www.techtarget.com/searcherp/feature/Advantages-and-disadvantages-of-digital-twin-technology">digital twin technology</a> to monitor and control every important object and space in the airport. <a href="https://www.computerweekly.com/news/366624051/Interview-Richard-Masters-vice-president-of-data-and-AI-Virgin-Atlantic">In an interview with Computer Weekly last year, </a>Virgin Atlantic’s head of data and AI, Richard Masters, described how the airline has revamped data-enabled decision-making support in recent years. “Our ability to assure the data and models are correct has been a big achievement for us,” he said. “I’ve got the team into a mindset of asking, ‘Is it right?’, and building capabilities. We’ve created this factory on the platform, which I’m proud of because I can see where we can keep extending that approach.” Virgin has used Databricks technology, particularly its <a href="https://www.techtarget.com/searchdatamanagement/news/366627703/Microsoft-Databricks-simplify-synchronizing-sharing-data">Unity Catalog</a>, to bring disparate data sources together and create a centralised location for the information behind the firm’s AI-enabled services. And <a href="https://www.techtarget.com/searchcustomerexperience/news/366632736/Salesforce-to-release-Agentforce-360-Slack-agents">Heathrow Airport has been an early adopter of Salesforce’s agentic AI applications suite, Agentforce</a>. It rolled out the Agentforce agent, named Hallie, in June 2025. It was trained on 800 knowledge articles and is said to resolve customer questions 40% more quickly than humans could do previously. Celonis’s <a href="https://www.techtarget.com/searcherp/tip/6-process-mining-steps-for-CIOs-IT-leaders">competitors in the process mining and process intelligence market</a> include SAP Signavio, UiPath and IBM Process Mining. <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read more about data-driven efficiencies at airports</h3> <ul class="default-list"> <li>Case study: <a href="https://www.computerweekly.com/feature/Case-study-Manchester-Airports-Group-using-AWS-to-revamp-its-data-strategy">Manchester Airports Group using AWS to revamp its data strategy</a>.</li> <li>Hamad International Airport launches one of the world’s largest <a href="https://www.computerweekly.com/news/366645616/amad-International-Airport-launches-one-of-the-worlds-largest-biometric-passenger-journeys">biometric passenger journeys</a>.</li> <li><a href="https://www.computerweekly.com/news/366622202/Airportr-checks-in-with-SITA-to-enhance-traveller-baggage-delivery-services">Airportr checks in with SITA</a> to enhance traveller baggage delivery services.</li> </ul> </div> </div> </section> Munich Airport turns to Celonis process mining software to optimise operations and passenger experience through real-time digital twins, focusing on baggage handling efficiency https://cdn.ttgtmedia.com/visuals/ComputerWeekly/HeroImages/Munich-Airport-Boarding1Now-Getty-RF-hero.jpg https://www.computerweekly.com/news/366647183/Munich-Airport-marshalls-digital-twin-to-improve-operations Wed, 05 Aug 2026 05:00:00 GMT Munich Airport marshalls digital twin to improve operations Anyone considering what guide rails need to be in place to protect us from <a href="https://www.computerweekly.com/feature/Cyber-protection-against-advances-in-frontier-AI-models">rogue AI behaviour</a>, need to read Anthropic’s postmortem of the latest security breach. Three models undergoing cybersecurity tests by a third-party , breached security and attacked three businesses on the open internet. The cybersecurity firm was not meant to provide any internet access – and so the fact that the Anthropic models found a way into the open internet is not <a href="https://www.computerweekly.com/news/366646678/Anthropic-lost-control-of-Claude-in-latest-AI-cyber-blunder">Anthropic’s fault</a>. But what really concerns Computer Weekly is the deviousness of the models to circumvent IT security measures. “The line between an aligned action and a harmful one is dependent on the model’s understanding of its situation,” Anthropic said in a post discussing the incidents. Looking at the second incident, Anthropic’s Claude model encountered instructions which told employees of a fictional company to install a non-existent Python package from PyPI—the public registry where Python software is published. Claude spotted this as a potential opening and built and published a booby-trapped Python package with the same name. Now here is where things get a bit spooky. According to Anthropic, in order to sign-up to a PyPI account, Claude needed an email address but to create this email address, it needed a phone number. After failing to find a free phone number service, it tried—and failed—to obtain funds to pay for a phone number through several different means and finally backtracked, found a free, non-blocked email provider, which it then used to register a PyPI account, and then upload malware to PyPI. Reading the postmortem shows the steps the Claude model took to circumvent existing guard rails that are supposed to stop rogue activity. Seeings its attempts in black and white, makes it seem that it is doing some pretty obvious things: obvious for a human maybe, but here’s an AI doing the sort of things a determined hacker would try in order to bypass cybersecurity measures. And the extent to which it went to achieve its objective – to hack a system under what should have been a controlled environment with appropriate safeguards, is a roadsign showing the direction of travel the advancement in AI systems is inevitably heading. Towards the end of the 1970s, the US and USSR superpowers agreed on a nuclear non-proliferation agreement, in recognition that they already had enough nuclear weapons to blow up the world many times over. Before we get to the point of no return and something truly catastrophic happens, perhaps it is time for the people and companies developing frontier models to pause and recalibrate their objectives. Maybe we need a global non-proliferation agreement to stop AI advances that could lead to a global catastrophe. Anyone considering what guide rails need to be in place to protect us from rogue AI behaviour, need to read Anthropic’s postmortem of the latest security breach. Three models undergoing ... https://www.computerweekly.com/blog/Cliff-Sarans-Enterprise-blog/Anthropic-Claude-incidents-show-the-limits-of-AI-guard-rails Wed, 05 Aug 2026 04:27:32 GMT Anthropic Claude incidents show the limits of AI guard rails Cloud-based management software player ECI Software Solutions has picked up Prima Business Control Software to add more enterprise resource planning (ERP) and commerce expertise. Prima serves the business supplies sector and has an ambition to combine its skills with the artificial intelligence (AI) knowledge held by ECI to pitch a compelling offer to office supply dealers in the UK&I and the US. Office supply dealers are under increasing pressure to provide a wide range of options to users, as well as ensuring they can deliver high levels of customer experience, putting them in a position to demand a solid ERP and management system.  “ECI has more than a 25-year history of commitment to the office supplies industry, and joining forces with Prima is a natural extension of our mission to help independent dealers thrive,” said Trevor Gruenewald, CEO of <a href="https://www.ecisolutions.com/en-gb/">ECI Software Solutions</a>. “At a time when distributors are facing rising customer expectations and tightening margins, the need for streamlined operations is critical. This new chapter reflects a strategic expansion of ECI’s ERP and ecommerce portfolio, and a commitment to serving our customers with solutions that anticipate their needs and arm them for success amid a constantly shifting market,” he added. In response, Stephen McLaughlin, CEO of <a href="https://www.primasoftware.co.uk/">Prima</a>, said the deal would provide its staff and customers with a wider array of resources and tools to work with: “With ECI, we are securing a future for Prima that would have been impossible to achieve alone. Now, we can accelerate innovation and provide the enterprise-level stability our customers deserve. “We’re still the same Prima at our core, but now we have the ‘firepower’ to help our customers run their businesses smarter and faster than ever before,” he said. At the same time, a glimpse into the tech investment world was shared by <a href="https://tracxn.com/">Tracxn Technologies</a>, which indicated that there were fewer deals, but they were of a higher value, during the first half (H1) of the year. The Europe tech – H1 2026 report from Tracxn indicated that total funding increased by 46% to $35.4bn as the number of funding rounds dropped by 18% and acquisitions decreased by 25%. Although there were 559 acquisitions across Europe in H1, this number was down from the 747 recorded in the same period a year earlier. Those that were completed were larger, however, and that buoyed the value of the transactions closed. AI infrastructure dominated the funding rounds that took place in the first half, with firms operating in that area getting the backing to continue developing their proposition, landing 38% of funding in the first half of the year. Enterprise applications was the top-funded sector at $17.7bn, with enterprise infrastructure also a popular area. The appetite for initial public offerings (IPOs) has decreased, with just eight priced in H1, which was down by 43% compared to the same period last year. The report underlined the UK’s position in the tech investment world, with London representing 38% of all European funding deals, accounting for $13.4bn in H1. ERP specialist acquired as Tracxn Technologies shines a light on the general state of European technology investments https://cdn.ttgtmedia.com/visuals/German/article/acquisition-monopoly-adobe.jpg https://www.computerweekly.com/microscope/news/366647182/ECI-snaps-up-Prima Wed, 05 Aug 2026 04:08:00 GMT ECI snaps up Prima <div class="imagecaption alignLeft"> <img src="https://cdn.ttgtmedia.com/rms/computerweekly/CW-60-anniversary-logo-white-400px.jpg" alt="Computer Weekly 60th anniversary logo" width="226" height="117"> </div> On 22 September 2026, Computer Weekly turns 60. To mark the milestone, we asked some of our friends - experts, parliamentarians, IT leaders and suppliers - for their perspectives on how tech has changed their lives over six decades. What's changed the most for you since then? My career in unified communications stretches back far enough that I’m amazed at how many technological eras I have lived through. But whenever I’m asked where it all truly began, my mind always returns to one moment in the late 1990s - the first time I saw a <a href="https://www.computerweekly.com/news/366625707/Cisco-Live-2025-The-digital-workplace-gets-closer-to-distance-zero">video conference come to life</a>. Up until then, communication in my world had been strictly practical. I’d trained in electronics and communications engineering, and my early career in the Royal Signals revolved around radio systems and ruggedised equipment designed to be, as we used to say, “squaddie‑proof”. These were devices built to survive being dropped, kicked, rained on, frozen, and regularly sworn at. Reliability mattered more than elegance. If it worked after being hurled into the back of a Land Rover, it passed the test. When the Cold War ended, everything changed, and I had to change with it. I left the military, swapped my uniform for a suit, and jumped into IT - though I wasn't really sure if I was being brave or just stupid. After a couple of jobs where I figured out what I was actually doing, I ended up at Cisco Systems as a systems engineer, focusing on call centres and IP telephony. <section class="section main-article-chapter" data-menu-title="A massive gamble"> <h2 class="section-title">A massive gamble</h2> Back then, the whole idea of putting voice traffic over a data network felt like a massive gamble. The voice folks and the data guys lived in totally different worlds. They had separate budgets, separate teams, and completely different ways of thinking. Suggesting they should share the same network setup was like trying to force cats and dogs to live together. The market started moving way too fast for the older hardware brands to keep up. Suddenly everyone wanted IP phones. Then call centres started switching over to software, and before anyone could really catch their breath, video arrived. I still remember the first time a face popped up on a screen and actually reacted to me in real time. It completely threw me. To be brutally honest, I didn't really care about the clever tech- it was just how human it felt. You suddenly got all those little things you completely miss on a regular phone call, like a quick smile, a smirk, or someone hesitating before they spoke. It felt like something straight out of a sci-fi film. Mind you, those early <a href="https://www.techtarget.com/searchunifiedcommunications/answer/Telepresence-vs-video-conferencing-Whats-the-difference">telepresence</a> rooms were a massive hassle to set up. They needed heavy-duty air conditioning, special furniture, and they even painted the walls a specific, weird shade of brown just to make people's skin tones look normal on camera. But that first video call hit me harder than most people might expect, because it stood in stark contrast to an earlier chapter of my life. </section> <section class="section main-article-chapter" data-menu-title="Christmas in the Falklands"> <h2 class="section-title">Christmas in the Falklands</h2> Years before all that, as a young Royal Signals tech, I spent Christmas in the Falklands. The wind there never really lets up. It just howls across the empty landscape, constantly reminding you exactly how far away you are from home. Keeping in touch with anyone back then was a massive luxury. Airmail only turned up once a week, and that was only if the weather was good enough for the planes. Those thin blue letters - we called them "blueys" - were everything to us. You’d end up reading the same few paragraphs over and over again. The news inside didn't change, obviously, but you just needed to feel that connection. <blockquote> <div class="imagecaption alignLeft"> <img src="“https://cdn.ttgtmedia.com/rms/computerweekly/Tim-Banting-LinkedIn.jpg"" alt="Photo of Tim Banting from So What? Now What?"> </div> “When I think about it, my job has never really been about the technology itself. It’s always been about people. It's about stopping all the little frustrations that get in the way when we try to talk to each other, and making the world feel a bit smaller when people are miles apart” Tim Banting, So What? Now What? </blockquote> Phone calls were even harder to come by. There were only a few phones around the complex and the queues were massive. On Christmas Day, I spent nearly an hour standing in line, practising what I wanted to say to my family back in Watford. But when the line finally connected, the awful five-second satellite delay messed everything up. We just kept cutting each other off, saying sorry, and trying to guess when the other person had finished speaking. I had to start using proper military radio procedure and saying "over" whenever I stopped talking. It was really annoying, but worse than that, it made me feel completely cut off from everyone. It made the distance feel heavy, like a physical weight. </section> <section class="section main-article-chapter" data-menu-title="Staying in touch"> <h2 class="section-title">Staying in touch</h2> Looking back, that whole experience stuck with me a lot more than I thought it would. It made me realise that staying in touch isn't just a nice-to-have - it's something we genuinely need as humans. We simply have to hear from the people who matter to us on a regular basis. When you lose that connection, it hits you hard right away. So when I saw that first video call at Cisco, besides being a technical milestone - from London to San Jose in California using an IP phone to dial the distant end’s conference room - it was a revelation. It was the opposite of what I experienced in the Falklands.  Those two things - the loneliness of that freezing, empty island and the shock of seeing live video for the first time - ended up shaping my whole career. It gave me a real drive for new tech that changes how companies work. It also made me realise the tools we build aren't just about saving money or making people work faster - they're for people. Whether that's employees trying to do their jobs, or families separated by thousands of miles of ocean who just want to feel closer. Since those early days, the whole sector has changed completely. We used to rely on massive, expensive kit stuck inside dedicated boardrooms. Now, everything just runs as <a href="https://www.computerweekly.com/news/366643663/EE-accelerates-UK-5G-roll-out-teams-with-Meta-to-boost-mobile-video">simple software on our laptops or mobile phones</a>. We stopped messing about with separate networks and bundled everything together. You don't even have to book meetings days in advance anymore either; you just hop onto a quick, casual chat whenever you need to. It's funny, looking back - we went from constantly asking if people could actually hear us to just saying, "I'll send over a link." </section> <section class="section main-article-chapter" data-menu-title="Jump on a call"> <h2 class="section-title">Jump on a call</h2> But the real win with all this change is how easy it's made everything to use. Nowadays, pretty much anyone can jump onto a video call from wherever they happen to be. Working from home is normal life now, and teams are scattered all over the place. It’s also made it way easier to keep in touch with people, whether that’s colleagues or family overseas. It’s still a bit mind-boggling that all the kit which used to fill massive metal racks now slides right into your pocket. When I think about it, my job has never really been about the technology itself. It’s always been about people. It's about stopping all the little frustrations that get in the way when we try to talk to each other, and making the world feel a bit smaller when people are miles apart. Mostly, I just wanted to make sure nobody else had to feel as completely cut off as I did during that Christmas down in the Falklands. <a href="https://www.computerweekly.com/feature/Evolution-of-unified-comms-services">The move to unified communications</a> didn't just shake up the tech sector, it’s completely changed everyone’s lives. Even now, it’s still altering the way we get through our workdays, whether in a global pandemic or a heatwave in the UK. And, more importantly, keeping in touch with each other. Tim Banting has been a technology analyst for over 12 years and has worked in IT for nearly 30 years. He is currently founder and principal analyst at So What? Now What? <div class="extra-info"> <div class="extra-info-inner"> <h3 class="splash-heading">Read all of Computer Weekly's 60th anniversary essays</h3> Technology is changing the way we live and work like never before – touching people’s lives every day, opening up new opportunities and creating new challenges. So for our 60th anniversary, we wanted to reflect on the human stories of how the digital revolution has changed the lives of some of the key leaders and influencers in tech today. Read our full collection of essays, which offers a unique set of insights and perspectives into the changes we have seen during Computer Weekly's first 60 years. <h2><a href="https://www.computerweekly.com/essentialguide/CW60-How-technology-has-changed-our-lives-over-60-years">CW@60 - How technology has changed our lives over 60 years</a></h2> </div> </div> </section> Sometimes our amazement at new innovations obscures what really matters - the way that technology connects people, and makes the world seem a smaller place https://cdn.ttgtmedia.com/visuals/ComputerWeekly/Hero%20Images/Penguin-fotolia.jpg https://www.computerweekly.com/feature/CW60-Christmas-in-the-Falklands-and-the-importance-of-keeping-in-touch Wed, 05 Aug 2026 03:00:00 GMT CW@60: Christmas in the Falklands and the importance of keeping in touch TheServerSide.com 60 webmaster@techtarget.com