Claude Mythos capabilities spark cybersecurity debate and controlled deployment

Anthropic's unreleased Claude Mythos Preview model, capable of identifying and fixing long-standing software vulnerabilities, has triggered industry debate over its power. Instead of public release, it is being deployed only via Project Glasswing with major partners. Covered in Indian Express on April 10. It matters because it demonstrates responsible frontier model governance, balancing breakthrough capabilities with controlled access to mitigate risks in critical infrastructure.