AI developer Anthropic is investigating reports of unauthorized access to its powerful Claude Mythos model, potentially allowing dangerous cyber capabilities.
