News
OpenAI and Anthropic Negotiated a Legally Binding Mutual Stress-Test Pact, Then the Talks Stalled
The Information reports the two labs came close to a binding agreement under which each would get API access to the other's commercially available models, excluding unreleased ones, with both sides barred from retaining the other's data, specifically to probe for vulnerabilities and security risks. The talks stalled, and it is unclear whether anything was signed before OpenAI's July agent-driven intrusion into Hugging Face. A formalized cross-lab red-team channel is the single most concrete governance mechanism the industry has floated that does not require legislation, and it did not close.
↳ Follow the thread