Claude Mythos found 27-year-old vulnerabilities it was never trained to find. That's the part enterprise AI roadmaps aren't accounting for. reddit-ai_agents · www.reddit.com ·9 replies ↗ ·2d securitymythosagenticanthropicclaude open →