Anthropic’s Nuclear Bomb

A new AI model from Anthropic, called Claude Mythos, can autonomously discover and exploit previously unknown cyber vulnerabilities at an unprecedented scale. This development is compared to the creation of the atomic bomb, representing a transformative capability that its creators may not be able to control. The model's power creates a chaotic asymmetry in cyberspace, giving immense coercive power to a wide range of actors and rendering current defensive doctrines like persistent engagement obsolete. Existing cybersecurity processes are too slow to counter the threat, and the author argues that defensive AI cannot close this offense-defense gap in time as the technology will inevitably proliferate.