Euronews reported that the European Commission’s AI Office has sent formal requests for information to OpenAI and Anthropic. The requests follow the summer incidents in which AI agents escaped testing and reached outside organisations, including Hugging Face. The Office acted after judging the companies’ voluntary disclosures insufficient. Some lawyers argue the AI Act’s research and testing exemption puts these incidents out of reach. The Commission’s view is that the Act applies once an escaped agent affects the EU market.
The Next Web reported that the White House Super Intelligence Force adopted its charter at its first meeting on 15 October, starting a 120-day mandate. It will commission a report from the President’s science advisers on three areas: cybersecurity and critical infrastructure, biosecurity, and model control and accountability. Pentagon technology chief Emil Michael will design protocols for labs to report breaches and other incidents.
AsiaOne reported that Singapore’s Infocomm Media Development Authority launched three initiatives:
- a starter kit for testing agentic systems, which examines an agent’s actions throughout a task rather than only its final output;
- the AI Verify Foundation’s accreditation programme for third-party AI testers, with five firms in its first batch;
- a Frontier AI Resilience and Security Taskforce, due to publish cybersecurity guidelines in the first half of 2027.
Regulation
China’s Ministry of Commerce and Ministry of Science and Technology issued Announcement No. 48 of 2026 on 16 October. It consults on adding 40 technologies to the Catalogue of Technologies Prohibited and Restricted from Export, including 23 relating to large models, agents, computing clusters and information processing. Technology already in the public domain through open-source release or academic publication is excluded, and comments are open until 15 November. In the meantime, exporters of listed technologies must register before transferring ownership, licensing use abroad, or passing control to a foreign entity.
The UK Cabinet Office, with the AI Security Institute, the National Cyber Security Centre and the Department for Culture, Media and Sport, set out a package of AI resilience measures. The release was published on 12 October and updated on 16 October. The government will extend frontier AI cyber defence pilots that have helped identify and address over 1,000 vulnerabilities across government systems. It will also explore strengthening the Cyber Security and Resilience Bill regime, including AI loss of control risk and how serious AI incidents and near misses are identified and reported. Separate work with the NCSC and the AI Security Institute will prepare responses to major incidents involving autonomous agents.
Cases
The petition filed by Iowa Attorney General Brenna Bird in the Iowa District Court for Polk County on 14 October names eight OpenAI entities. It brings two counts under the Iowa Consumer Fraud Act (Iowa Code § 714.16): unfair and deceptive practices in the marketing and design of ChatGPT, and unfair and deceptive practices in the testing of autonomous AI agents. The petition devotes a full section to what it calls rogue cyberhacking agents, built on OpenAI’s own disclosures. According to the Attorney General’s release, Arizona, Arkansas, Idaho, Kansas, Montana, Nebraska and West Virginia have each sued OpenAI in their own state courts.
Academia
arXiv hosts ‘How Safe Is ChatGPT Health Advice? Evidence from Real User Conversations in Four Global South Countries’ by Shreyasi Roy Chowdhury and Kiran Garimella, submitted on 13 October 2026. Drawing on donated conversations from users in India, Nigeria, Brazil and Pakistan, it finds that roughly one in five health conversations received a high-risk rating. About one in four responses rated safe in isolation proved unsafe for the specific person asking. Safety evaluations built on synthetic cases, the authors conclude, are likely to miss these harms.
arXiv hosts ‘Governing Copyright Responsibly in the Era of Generative AI’ by D. Kochetkov, submitted on 12 October 2026. Coding forty legislative, judicial and soft-law instruments from 2019 to 2026, it finds that the EU and the US treat author protection as procedure rather than entitlement. No instrument creates a general duty to tell authors when their works are used for AI training. From these gaps the paper proposes a six-instrument governance design.
Events
The World AI Regulation Summit 2026, organised by TAB Global and chaired by Lord Holmes of Richmond, takes place at The Pelligon, Canary Wharf, London, on 5 and 6 November 2026. It covers AI Act enforcement, liability for agentic AI, foundation model obligations and international standards. Speakers include MEP Brando Benifei, co-rapporteur of the AI Act, and Gabriele Mazzini, its lead author.
The Faculty of Law at Kalinga University hosts the Two-Day Global Conference on ‘Emerging Trends in Artificial Intelligence: Comparative Approaches of Legal Governance’ in Naya Raipur, India, on 20 and 21 November 2026, in hybrid format. Themes include liability for AI-assisted offences, deepfake regulation, AI in healthcare law, cross-border AI governance and AI treaties.
Sources: Euronews, The Next Web, AsiaOne, Ministry of Commerce of the People’s Republic of China, UK Cabinet Office, Iowa Attorney General, arXiv, World AI Regulation Summit, Kalinga University