The EU AI Act Newsletter #97: Safety and Enforcement

The Scientific Panel briefs Virkkunen on recent incidents, the AI Office investigates OpenAI and Anthropic, and the first harmonised AI Act standard is published. Welcome to The EU AI Act Newsletter, a brief biweekly newsletter by the Future of Life Institute, bringing you the latest developments and analyses on the EU's artificial intelligence law. Developments Special meeting of the AI Scientific Panel: On 9 October, Executive Vice-President Henna Virkkunen and Seán Ó hÉigeartaigh, Chair of the AI Scientific Panel, held a special meeting to inform the Commission's actions following recent loss-of-control incidents. The chair and vice-chair set out what is known about the incidents and what is still unknown. The AI Office has already requested and received serious incident reports and sent requests for information on security practices. Building on these efforts, the Commission will send developers further requests combining the panel's and the AI Office's questions. The meeting also covered evidence on recursive self-improvement and on training techniques that make models harder to monitor. Finally, the chair and vice-chair recommended governing the risks of internal models, strengthening EU capabilities to detect, evaluate and investigate such incidents, setting up an international initiative for information-sharing, common requirements and joint [...] --- Outline: (00:40) Developments (06:09) Analyses ---

The Commission backs an international AI safety monitoring body but faces criticism at home over the pace and resourcing of AI Act enforcement. Welcome to The EU AI Act Newsletter, a brief biweekly newsletter by the Future of Life Institute, bringing you the latest developments and analyses on the EU's artificial intelligence law. Developments EU to Trump: We will keep pushing for global AI safety rules: Mathieu Pollet and Pieter Haeck from POLITICO report that EU tech chief Henna Virkkunen has pledged to keep seeking an international agreement on AI security despite US resistance. At the RAID Conference, she acknowledged Washington's opposition to international regulation on innovation grounds, but argued that pro-innovation regulation builds the trust on which innovation depends. She praised the AI Act as a strong regulatory basis for tackling challenges posed by AI bots. Whereas President Donald Trump has dismissed AI existential risks as a "hoax", the EU has backed the initiative led by Finland and Norway and signed by 20 other countries for an international body to monitor AI safety. Pointing to cases of AI agents escaping their environment, inserting malicious code and deceiving humans this summer, Virkkunen noted that talks are under way [...] --- Outline: (00:39) Developments (02:55) Analyses ---

Von der Leyen puts frontier AI risk at the centre of her State of the Union address, while the AI Office confirms OpenAI did not file a report on the RubyGems incident. Welcome to The EU AI Act Newsletter, a brief biweekly newsletter by the Future of Life Institute, bringing you the latest developments and analyses on the EU's artificial intelligence law. Developments 2026 State of the Union address: Ursula von der Leyen, President of the European Commission, describes artificial intelligence as the second tipping point of our times and a foundational layer for Europe's economy and security, arguing that AI's possibilities can only be unlocked if its risks are addressed. Those risks, she notes, have sharpened as frontier models advance, from hacking capabilities soon available to adversaries to the dangers of self-improving models, with agents escaping their environments or inserting malicious code offering only a glimpse. After the Hugging Face incident, developers are ringing the alarm and leading company CEOs are calling for a slowdown on self-recursive models. In response, she presents the AI Act as a crucial guardrail, promises cooperation with Canada, the United Kingdom and others on model evaluation, verification, early warning and AI security [...] --- Outline: (00:39) Developments (04:18) Analyses ---

From first requests for information to crisis preparedness and ChatGPT's designation under the Digital Services Act. Welcome to the EU AI Act Newsletter, a brief biweekly newsletter by the Future of Life Institute providing you with up-to-date developments and analyses of the EU artificial intelligence law. Legislative Process EU orders leading AI labs to detail security practices: Maximilian Henning from Euractiv reports that the European Commission has made first use of its new AI Act enforcement powers, requesting information from leading developers on cybersecurity, safety and copyright compliance. This follows a string of high-profile cybersecurity incidents at several leading AI labs just weeks ago. Tech Commissioner Henna Virkkunen told Euractiv that some providers of the “most advanced” models had received requests, without naming them, adding that others were left out because the Commission has “already had very close dialogue” with them. Anthropic, OpenAI, Google, Meta and Mistral did not immediately respond to Euractiv questions. According to Virkkunen, the Commission wants to know how labs prevent humans or AI from stealing their models, how much access they give external evaluators, and how usage is monitored once a model is public. Separately, more than 30 companies that had not published [...] --- Outline: (00:36) Legislative Process (01:53) Analyses ---

Anthropic sets out how it will mark Claude's text as the new transparency obligations take effect, while experts reject claims that watermarking degrades AI writing. Legislative Process Fourth GPAI Signatory Taskforce meeting: The European Commission's AI Office reports on the Taskforce's 17 July 2026 session, which covered the Safety and Security and Copyright chapters of the GPAI Code of Practice. On the former, the Office explained how analysing model usage can form part of post-market monitoring under Measure 3.5, since information about how a model is used, and its behaviour during usage, complements pre-deployment evaluations and informs several stages of systemic risk assessment. Participants also discussed ‘marginal-risk’ clauses, which indicate that, should competitors begin pushing the frontiers of capabilities by deploying unsafe models, a provider could consider doing likewise. The Office highlighted that such clauses could be invoked only in exceptional circumstances and under appropriate evidentiary and procedural safeguards. Turning to copyright, Measure 1.3(4) commits signatories to publishing information that enables affected rightsholders to learn which web crawlers are used and their robots.txt features, and to notifying them automatically of updates. How are AI agents addressed within the AI Act? The Commission's AI Act Service Desk sets out [...] --- Outline: (00:38) Legislative Process (03:07) Analyses (08:40) Discussion about this post (08:44) Ready for more? ---