Anthropic launches cybersecurity AI models, experts raise concerns

Anthropic has launched two new AI models, Fable 5 and Mythos 5, enhancing their Project Glasswing AI program focused on cybersecurity. Mythos 5 is touted as the most capable model for cybersecurity, to be initially deployed with US government collaboration before broader availability. Fable 5 includes added safeguards to prevent misuse in sensitive areas. Both models significantly undercut previous pricing, with Mythos 5 and Fable 5 priced at $10 and $50 per million tokens, respectively.

However, cybersecurity experts express concerns over these new tools, emphasizing the need for improved governance and accountability, particularly as their capabilities grow. Industry leaders warn of potential risks if guardrails fail, underscoring that enhanced AI tools must be deployed responsibly to avoid compromising public systems.