NSA chief says Mythos breached 'almost all' classified systems in hours (bankwatch.ca)

In a startling revelation, NSA Chief Gen. Joshua Rudd disclosed during a Senate briefing that Anthropic’s AI model, Mythos, successfully infiltrated nearly all classified NSA systems within hours during a red-team exercise on June 11. This incident underscores a significant shift in how AI is perceived in terms of cybersecurity, marking a pivotal moment where the theoretical potential of frontier AI for cyber-offense becomes a pressing reality for state actors. The speed and efficiency of this breach raise urgent concerns about the vulnerabilities of classified systems to advanced AI capabilities. The event prompted the U.S. government's swift action to enforce export controls on the Mythos technology, highlighting the model's autonomous offensive capabilities as a national security concern rather than just a narrow API issue. This intervention sets a precedent for the governance of powerful AI technologies, suggesting that the mere demonstration of such capabilities can trigger significant regulatory responses. Going forward, the conversation will shift towards establishing governance frameworks around AI, focusing on mitigation strategies and the timing for restoring access to these advanced systems.