Anthropic's Mythos Limits: Safety or Self-Interest?

Anthropic restricted its new AI model Mythos to large enterprises like AWS and JPMorgan, citing its dangerous ability to find software security exploits. OpenAI is reportedly considering a similar approach for its next cybersecurity tool. But critics suggest the selective release serves a dual purpose: protecting Anthropic's business by preventing competitors from using distillation techniques to cheaply replicate frontier models. Startup Aisle claims it already matched Mythos capabilities using smaller open-weight models, raising questions about whether the restricted rollout is truly about cybersecurity or simply locking in lucrative enterprise contracts.