The EU wants to regulate AI but needs OpenAI and Anthropic to let regulators through the door

The European Commission is negotiating access to frontier AI models from OpenAI and Anthropic to strengthen oversight under the EU AI Act and cybersecurity rules.

OpenAI has offered access to its GPT-5.5 Cyber model, with talks ongoing over which EU bodies can review it.

Anthropic has held multiple meetings with the Commission about its Mythos model, but no access agreement has been reached.

The episode shows EU regulators still depend on voluntary cooperation from AI firms, limiting enforcement and security oversight.