Regulate to protect. Representatives of the twenty-seven member states of the European Union are meeting on Thursday in Brussels within the AI Council. According to Politico, the Commission will present an update on European and international policies as well as a technical briefing on recent incidents linked to artificial intelligence. Global cooperation is also among the topics of discussion, in a context of disagreement with Washington on how to supervise the most advanced models.
Key Points
EU member states are discussing recent AI-related incidents and international cooperation in Brussels on Thursday, according to Politico.
The AI Act already imposes obligations on providers of general-purpose AI models, reinforced in the event of systemic risk.
The main obligations regarding high-risk AI systems have been postponed to December 2027 or August 2028 depending on their category.
Like MiCA for crypto, the European framework serves as a basis for international exchanges, without guaranteeing the adoption of global rules.
AI: Brussels seeks rules beyond Europe
The AI Council brings together representatives of member states to monitor the application of the AI Act. According to the agenda consulted by Politico, Thursday's meeting aims neither to adopt a new regulation nor to adopt a definitive common position. The discussions will focus in particular on the work undertaken at the G7, the G20 and the United Nations.
The first UN Global Dialogue on AI governance has already been held in Geneva on July 6 and 7, 2026. The European Union therefore has a forum in which to defend its approach, without being able to impose international rules alone. Commission Vice-President Henna Virkkunen recently called for global cooperation: the risks associated with advanced models do not stop at European borders.
This move comes after calls from leaders like Dario Amodei, at Anthropic, and Sam Altman, at OpenAI, to slow down the development of AI. Donald Trump rejected these warnings, believing that such a slowdown would favor China. The Brussels meeting allows Member States to compare their analyses, but it does not resolve this disagreement.
AI Act and MiCA: A regulatory advance to be qualified
Europe already has a binding framework for developers of general-purpose AI models. Since August 2025, the AI Act has imposed on them obligations of transparency and respect for copyright. Model providers that may present systemic risk must also assess and mitigate that risk.
In contrast, the timeline for high-risk AI systems has changed. A regulation that entered into force in July 2026 postpones their main obligations until December 2, 2027 for certain uses, notably in employment, education or critical infrastructure, and until August 2, 2028 for systems integrated into certain products. The transparency rules applicable since August 2026 are not affected by this postponement.
The crypto sector offers a point of comparison, without constituting an exact precedent. With MiCA, the Union has set common rules for issuers of certain cryptoassets and platforms that provide services to European customers. This harmonization makes the obligations easier to read, but does not guarantee either the emergence of European leaders or the adoption of similar rules elsewhere. The same goes for AI: having a European framework provides a basis for international discussions, not a global agreement.
On Thursday, European representatives will review recent incidents and compare international initiatives. Donald Trump's rejection of a slowdown in AI, however, highlights the difficulty of obtaining global agreement on the risks of the most advanced models.
