OpenAI and Anthropic have been discussing a legally binding agreement to conduct stress tests on each other's AI models. Earlier this year, the two companies and their legal teams explored a plan to run a series of evaluations on their respective models, aiming to uncover vulnerabilities or potential hazards.

The proposed arrangement would grant each company application programming interface (API) access to the other's commercially available AI models, while excluding any models that have not yet been released. Both OpenAI and Anthropic have assured that they will not retain each other's data during the testing process, and it remains unclear whether the two parties ultimately finalized the deal.