Claude Mythos AI: Cybersecurity Alarms Raised Before Release

Anthropic expects to release its Claude Mythos AI model to customers in the coming weeks. This follows additional testing and the development of safeguards. What's interesting is Claude Mythos has drawn scrutiny. Researchers found it could autonomously identify vulnerabilities and complete complex cyberattacks. The company describes Mythos as its most powerful AI model, surpassing its Opus models. Anthropic has been restricting access through its Project Glasswing program. This program gives select partners controlled access. The company believes Mythos could help defenders find and fix software vulnerabilities. However, some security researchers and government agencies worry its capabilities could also speed up cyberattacks. For example, the U.K.'s AI Security Institute found Mythos could complete a 32-step simulated corporate network attack. Mozilla also said Mythos identified 271 vulnerabilities in Firefox during evaluations. The bottom line is that this technology could significantly impact cybersecurity, for better or worse.