A U.S. official revealed that Anthropic's Mythos model successfully detected vulnerabilities in classified government systems within hours during a testing initiative called Project Glasswing, which aims to enhance cybersecurity. This revelation, mentioned by Democratic Sen. Mark Warner, underscores the potential risks posed by advanced AI technologies to public safety and national security.
Despite the collaboration, tensions have escalated between Anthropic and the Trump administration, particularly regarding the military's use of AI. The administration has imposed restrictions on Anthropic's models, including a recent directive preventing foreign nationals from accessing the latest versions.
Anthropic has complied by disabling these models for all customers, although the company disputes the necessity of such measures.
A coalition of over 100 cybersecurity experts has urged the administration to reconsider its directive, arguing that limiting access to effective cybersecurity tools could inadvertently benefit U.S. adversaries, especially as they advance in their own technological capabilities