THE US Cybersecurity and Infrastructure Security Agency (CISA) is employing Anthropic's Mythos AI model to audit federal software for security vulnerabilities, aiming to identify and fix potential exploits. The initiative, led by CISA's Attack Surface Evaluation team, has reportedly found numerous vulnerabilities, though details remain undisclosed. Tensions exist between Anthropic and federal officials, particularly after the company's refusal to comply with demands regarding its models' usage.
The National Security Agency (NSA) is also using Mythos, which has been linked to uncovering flaws in sensitive government systems. Furthermore, concerns over access to Mythos' public version led to regulatory issues, including a temporary shutdown of its Fable model.