When Access Control Fails: Powerful Zero-Day Discovery AI Exposed to Unauthorized Users

A highly restricted AI tool (Mythos) designed to discover zero-day vulnerabilities was inadvertently accessed by unauthorized users through contractor access loopholes and public data scraping, not sophisticated hacking. This incident underscores a critical lesson: the security of even the most advanced AI systems depends less on technical safeguards and more on organizational access control and human accountability. As AI tools grow more powerful, securing who can access them becomes as critical as the technology itself.