Pentagon AI Breach: Mythos Hacked Classified Systems in Hours

The Pentagon's cyber chief reveals that an advanced AI model breached almost all classified systems in a matter of hours. General Joshua Rudd, who leads the National Security Agency and Cyber Command, shared this finding with Senator Mark Warner. The model, named Mythos, penetrated the bulk of America's classified networks with extraordinary speed. What's interesting is that following this disclosure, the U.S. government abruptly banned foreign users from accessing Mythos 5 and Fable 5, Anthropic's two most powerful AI systems. This ban cut off not only the general public abroad but also close allies, including the Five Eyes intelligence partners: Australia, Britain, Canada, and New Zealand. Permissions that allied governments and major firms had spent months securing vanished overnight. Even Britain's AI Security Institute, which tests AI models for the government, lost access. The restrictions did not extend to Mythos Preview, an earlier version that Anthropic had already limited to a small group of approved customers due to national security concerns. Allied governments still have access to that version. Negotiations between governments are reportedly underway to regain access to Mythos. This situation highlights the growing challenges and security implications of advanced AI for national defense.