Anthropic will give defenders what its strongest model finds, but not the model itself

Anthropic has released Claude Mythos 5 for code-scanning in its Claude Security platform and is embedding it into partners’ defensive tools, delivering only the model’s outputs to users. The company is also allocating $35 million in credits to support open-source security initiatives. This expands access to its most capable cybersecurity model while restricting direct model interaction.