Anthropic limits release of its Mythos model to large enterprises, citing cybersecurity risks and blocking distillation by rivals

Anthropic restricted its newest model, Mythos, to a select group of large companies and organizations operating critical online infrastructure, including AWS and JPMorgan Chase, saying the model is too capable at finding software security exploits to release publicly. Critics suggest cybersecurity concerns may not be the only motivation, with some analysts arguing the selective release strategy also protects Anthropic's enterprise revenue by preventing competitors from using distillation, a technique that leverages frontier models to train cheaper rival LLMs. AI cybersecurity startup Aisle claims it replicated much of what Anthropic says Mythos accomplished using smaller, open-weight models, raising questions about whether the model's capabilities are as singular as the restricted release implies.