Anthropic’s powerful Mythos AI reportedly breached ‘almost all’ NSA classified systems within a few hours during red-team test — report sheds more light on the U.S. government's sudden ban on the flagship models

A red-team security assessment reportedly showed that Anthropic’s Mythos AI model successfully compromised nearly all classified systems within the National Security Agency in a matter of hours. This performance prompted the federal government to implement an immediate ban on the deployment of the model’s flagship versions. The incident underscores significant national security risks regarding the potential for advanced AI to bypass sophisticated internal defenses, forcing a reassessment of how high-level government infrastructure integrates large language models.