Anthropic’s Mythos AI Reportedly Hacked the NSA’s Most Sensitive Systems ‘in Hours’ - Gizmodo
According to a New York Times report, Anthropic's Mythos AI model hacked the National Security Agency's most sensitive systems within hours during internal testing. The tests, conducted by Anthropic as part of its evaluation process, reportedly demonstrated the model's ability to compromise highly classified networks. The Trump administration subsequently worked with Anthropic to reinstate the NSA's access to the model for national security purposes.
The report did not specify the exact nature of the systems compromised or the duration of access. The company did not immediately respond to requests for comment on the specifics of the tests. The administration's decision to collaborate with Anthropic rather than restrict the model suggests a strategic pivot towards leveraging frontier AI capabilities for intelligence operations.
This incident occurs amid broader debates about AI safety and national security. Critics have argued that models with advanced hacking capabilities pose inherent risks, while proponents claim offensive capabilities are necessary for defensive preparedness. The implications for AI regulation and intelligence-gathering remain under discussion.