Summary
The proposed assessments cover safety cases, safeguards, capability thresholds and misalignment incidents. OpenAI says evaluators should have access proportionate to their claims, disclose methods and retain editorial independence, while the company may request sensitive redactions.
Why it matters
This is a concrete governance mechanism for moving frontier-model oversight beyond internal testing, although partners and access terms remain unresolved.
Related coverage
- Twenty-one countries call for mandatory frontier-AI testing
- Microsoft disrupts AI-assisted EvilTokens phishing service
- OpenAI extends Daybreak cyber-defence access to Ukraine