Anthropic is the most open of the major labs; RSP, actual model cards, and they publish their red teaming results. C.f. openAI who released o3 and didn’t publi

Anthropic is the most open of the major labs; RSP, actual model cards, and they publish their red teaming results. C.f. openAI who released o3 and didn’t publish any model card or safety eval at all, their justification being “it’s not GA, only the top paid subscription can use it”. That’s not how safety works.