Anthropic started red teaming new Mythos models, first results
Anthropic began red teaming its new Mythos model, codenamed Claude Oceanus-v1-p, on June 3, with early access granted to security testers. The preview candidate focuses on advanced reasoning, coding, …