Anthropic's AI Model Discovers Code Vulnerabilities Faster Than Microsoft Can Fix Them

Anthropic has developed an artificial intelligence model called Mythos that identifies security vulnerabilities in code at an unprecedented pace. The model was provided to selected organizations, including Microsoft, whose engineering team convened in mid-May to address the vulnerabilities being uncovered. The company faces a race against time to patch the flaws before malicious actors and adversarial governments such as China can leverage similar tools for espionage and sabotage. Questions emerged during the meeting about whether Mythos delivered on the capabilities Anthropic had promoted.