How is the European Commission accessing new AI models?

EU seeks access to latest AI models from OpenAI and Anthropic

The European Commission said it is in ongoing discussions with OpenAI and Anthropic to obtain access to their latest AI models. OpenAI, in parallel, is described as “proactively offering” access.

The move matters because EU policy discussions increasingly hinge on being able to evaluate frontier systems—not just hear high-level descriptions. Access to current model capabilities is what enables regulators and researchers to test behavior, robustness, and limitations in ways that can inform oversight and compliance.

This is also occurring alongside broader regulatory efforts aimed at governing how advanced AI systems are built, deployed, and audited. For the Commission, model access conversations indicate a practical approach: getting the ability to work with the same-generation models that companies are offering commercially.

What we know from the coverage

• The Commission is talking with OpenAI and Anthropic about accessing their latest models.
• OpenAI is said to be proactively offering access.
• No specifics were provided on the technical mechanism, scope, or whether access would be through APIs, downloadable weights, or some other evaluation setup.

Why this could be significant

• Regulation without testing is harder: Access supports direct assessment.
• Fast-moving tech: “Latest models” suggests the Commission is trying to avoid evaluating AI that’s already out of date.

Overall, the reported discussions point to a regulatory strategy that depends on real model access rather than relying solely on documentation or older benchmarks.