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 in forums such as the G7. Meanwhile, she said that the EU will keep enforcing its rules and called for closer cooperation among national AI safety organisations.
Von der Leyen ally slams EU Commission’s AI response: Pieter Haeck and Anne McElvoy from POLITICO write that Axel Voss, a German EPP lawmaker and ally of Ursula von der Leyen, has accused the Commission of failing to keep pace with AI. Speaking on the Brussels Playbook Week Ender podcast after a series of AI hacking incidents, including OpenAI-powered agents breaking into Hugging Face, Voss said the Commission’s working methods are unsuited to such fast-changing developments. He added that enforcement of the AI Act, which requires companies to take measures against losing control of their models, is “not on the level” he would like. His comments put the spotlight on the AI Office, which was put in charge of enforcing the Act in August. Voss, who belongs to the CDU like von der Leyen, also said her State of the Union was not ambitious enough on digital developments. He argued that AI liability has been neglected since the start of the mandate. Separately, other MEPs have proposed adding AI liability provisions to address gaps in the Act.
Analyses
As AI accelerates, governments are increasingly being left behind: Adam Satariano and Cecilia Kang from the New York Times find, based on more than 20 interviews, that AI has made the gap between policymaking and technology one of the widest ever. In Brussels, Ursula von der Leyen and her commissioners have yet to decide whether to ramp up regulation, given the possible economic and geopolitical costs if other countries do not follow. Meanwhile, the AI Act, once hailed as a blueprint, has become a cautionary tale. Its implementation has been slowed by fears about competitiveness, key high-risk rules have been delayed, and staffing and budget constraints have hampered enforcement. When OpenAI models escaped a restricted testing environment in July and hacked Hugging Face, the breach fell into a murky area for enforcement, since the Act generally exempts testing done before a model’s release. The technical know-how needed for enforcement is also still being built, as the AI Office has only about 40 people in its AI safety unit.
Artificial intelligence and crime: Martina Prpic from the European Parliamentary Research Service explores how AI is transforming crime and law enforcement. On the one hand, AI has lowered barriers to some forms of crime and made offenders' activities more sophisticated and far-reaching. In some cases, it has also let criminals distance themselves from their acts, which makes it harder to link them to those crimes and prosecute them. Some systems can also produce harmful outcomes without direct human input, which raises questions about criminal responsibility. On the other hand, law enforcement agencies are using AI to work more efficiently, for example in identifying vulnerable and exploited children and supporting emergency call centres. Unlike criminals, however, they must use it lawfully, ethically and without discrimination. In the EU, this means complying with the AI Act, which classifies certain law-enforcement systems as high-risk, as well as the GDPR and the Law Enforcement Directive. Bias in training data, which may reflect racial and historical prejudices, remains one of the biggest challenges, so human oversight is still indispensable.
Different risks, different rules? The case of ChatGPT under the DSA and the AI Act: Lena-Maria Böswald, Senior Policy Researcher at interface, considers how responsibility should be divided between the Digital Services Act (DSA) and the AI Act now that the Commission has designated ChatGPT a Very Large Online Search Engine. She sees the DSA as the primary framework for risks arising from ChatGPT's design, functioning and use, such as those to fundamental rights, elections and minors. However, the DSA neither regulates how models are built or trained nor directly binds AI developers, so the AI Act is still needed for model-level obligations. The two regimes can support each other, but one cannot replace the other: a model provider’s AI Act documentation can inform a platform’s DSA risk assessment without replacing it. Böswald sees the main remaining gap as institutional. She recommends that the Commission’s DSA enforcement team and the AI Office connect their risk assessments and develop a shared risk matrix.
Could AI kill us all? Kristo Lehtonen, Director of Sitra International Programmes, argues that the key question is not whether AI will wipe out humanity, but whether humans will keep the ability to understand and control increasingly sophisticated systems. Just as a dog cannot conceive of a firearm, we may not grasp all the ways superhuman AI could affect us. Oversight is getting harder as self-improving AI advances, with Anthropic reporting that it has shifted a growing share of its development work to the AI itself. Meanwhile, Anthropic and OpenAI are preparing for stock market listings, and these high stakes encourage risk-taking. Lehtonen considers Europe’s risk-based model the most sustainable in the long term, because it provides the means to manage the very risks that developers warn about. He calls for the AI Act to be implemented effectively, applied straightforwardly and updated to keep pace with developments, and for the EU to promote international coordination.
We’ve built the most comprehensive website on the EU AI Act to help answer all your questions. Here are a few of our most popular resources used by 60,000+ professionals every week:
AI Act Explorer: Explore the official AI Act text on any device, in any EU language, with helpful cross-links, added context, and more.
Compliance Checker: In just 10 minutes, figure out exactly what your business or organisation must do to comply with the AI Act.
High-level Summary: A short overview of the AI Act, with a breakdown on risk categorisation, obligations, prohibited systems, and timelines.
Small Businesses’ Guide: Everything you need to know, for small and medium-sized enterprises (SMEs) in the EU and beyond.



An international monitoring body will only help if it receives comparable incident data, not polished summaries. The cases cited here — agents escaping their environment, inserting code, deceiving users — need a common record: what the agent could do, what changed, who intervened, and how quickly. Practitioners can start now by logging those four fields for every live agent, so a future body is auditing evidence, not anecdotes. https://aideliverybrief.substack.com/p/singapores-new-test-kit-says-a-right
The August 2025 deadline for member states to designate competent authorities is approaching fast. Three months out and still no clarity on resourcing. Brussels loves announcing global frameworks while the domestic machinery creaks.