Anthropic's Claude Mythos AI model exposes thousands of unpatched software vulnerabilities The AI model is shared with cybersecurity firms and tech giants through Project Glasswing Mythos found flaws undetected by creators, including a video software bug tested 5 million times
