Anthropic adds 150 new organizations to its Project Glasswing initiative, bringing its highly restricted Claude Mythos Preview model to critical infrastructure sectors. The AI model has already uncovered over 10,000 severe software vulnerabilities for initial partners.