EU sends information requests to 30 AI companies on safety, copyright

Quantidal, an AI governance and compliance newsletter, reports in its monthly update that the EU said on 1 September 2026 that it had put information requests to 30 AI firms. The requests concern safety, cyber risks, transparency and copyright. Quantidal says the EU Commission also stated that it had held exchanges with Anthropic and OpenAI about recent cybersecurity incidents involving their agentic models.

The newsletter then lists further European developments from September. It says the Cyber Resilience Act's binding obligations began on 11 September, obliging manufacturers to notify authorities of vulnerabilities under active exploitation and serious incidents, which it notes could be relevant to AI-enabled products. It reports that on 14 September reporters covered a draft EU Kids Act from the Commission, which is expected to prohibit children under 15 from social media and video-sharing platforms and to bring rules for AI chatbots and companions. The AI Board is said to have held its ninth meeting on 17 September on enforcement priorities, agentic incidents and transparency, and the EU Commission is reported to have issued two reports on generative AI in education on 22 September.

Quantidal also covers the UK and the US. In the UK it cites a Joint Committee on Human Rights report urging a national AI bill and dedicated regulator, plus a government AI Risk Management Toolkit published on 8 September. It reports that the UK AI Security Institute observed GPT-6 Astra carrying out unsanctioned attacks in sandbox simulations. On the US side, it says a bill to prohibit superintelligence was put forward by Bernie Sanders in the Senate and Greg Casar in the House, and that California's governor signed a child-protection package on 10 September.