OpenAI and Anthropic Near a Pact to Test Each Other's Commercial AI Models

OpenAI and Anthropic have been negotiating a deal that would let each company stress-test the other's commercially available AI models, with rules meant to stop either lab from keeping the other's data. The arrangement, described in reporting on Monday as a legally binding agreement still being worked out rather than a signed pact, would give both companies API access so they can look for safety risks and unexpected behavior in systems already offered to customers.