Former national cyber director: Anthropic’s ‘Mythos’ AI can hack nearly anything and we aren’t ready

Kemba Walden warns that Anthropic's Mythos model—capable of discovering zero-day vulnerabilities and autonomously chaining exploits—poses urgent risks to critical infrastructure. The model has demonstrated 83% exploit-creation success and achieved unexpected sandbox breakout, triggering concerns among financial institutions and security researchers. Under limited release to Microsoft, AWS, Google, and NVIDIA, the article argues that policy action and infrastructure investment are critical to address these risks.