Policy & safety· 1 min read
EU AI Office sends its first formal information requests to top AI model makers
Under the AI Act, Europe's AI Office asked providers of the most advanced models about security, independent testing and how their models are used.
On October 14, 2026, European Commission Executive Vice-President Henna Virkkunen told an international AI summit in Dublin that the EU's AI Office had sent its first formal requests for information under the AI Act to providers of the world's most advanced AI models.
What the AI Office is asking about
- Security: whether models run in secure, contained environments.
- External evaluation: whether independent testers get deep enough access, and whether companies act on what they find.
- Monitoring: how providers track real-world use of their models and feed it into their risk assessments.
Providers are required to answer these requests. Under the AI Act, breaking the rules for general-purpose AI models can lead to fines of up to €15 million or 3% of worldwide annual turnover.
Why it matters
This is one of the first concrete enforcement steps of the AI Act for powerful general-purpose models, and it targets the very risks that come with AI agents. Read about those risks in Using agents safely.
Read the original at European Commission
Latest AI news
Anthropic files for an initial public offering
The maker of Claude filed a registration statement with the US Securities and Exchange Commission. The number of shares and the price have not been set.
Read moreThe Agent Skills standard moves to the Linux Foundation's Agentic AI Foundation
Anthropic donated Agent Skills, the open format for instructions that AI agents load only when needed, to the neutral foundation that already hosts MCP.
Read more