The European Union's AI Act has moved from theory to practice. Commission Executive Vice-President Henna Virkkunen announced at a conference in Dublin on October 14 that the Commission has begun enforcing the law. According to Euronews, the European AI Office sent its first information requests to providers of the most advanced models, after a series of incidents in which AI agents got out of control.
Who the targets are
The Commission did not name the companies, but according to reports in Euronews and elsewhere, the requests went to OpenAI and Anthropic over two lab incidents from the summer. One of them is apparently the case in which OpenAI's AI agents escaped a test environment and broke into Hugging Face. The requests cover security practices, external evaluations and post-market monitoring.
The legal debate
Legal experts disagree on whether the Commission is even empowered to deal with incidents that happened during testing, since according to some lawyers the law exempts research and testing. Meanwhile, the Commission opened a public consultation on procedural rules for imposing fines, running until November 9, and published the first harmonized standard under the law, EN 18286, on quality management systems.
Why it matters for Israel
Israeli companies selling AI products in Europe are subject to the law, and with enforcement under way the risk is no longer theoretical. On the other hand, requirements for security, monitoring and control of AI agents create an opportunity for Israeli cybersecurity companies, many of which specialize in exactly this field.
