National Security

PARIS, FRANCE - FEBRUARY 26, 2026: An abstract kaleidoscopic view of the Anthropic logo creating a geometric pyramid shape, representing the complexity of neural networks.

Anthropic withheld its new artificial intelligence model, Mythos, from a broad public release after internal testing showed it could autonomously find and exploit serious software flaws, alarming company researchers and prompting banks and government agencies to move quickly to assess the threat.