Anthropic Limits Access to Mythos Cybersecurity AI Model

Anthropic has announced a restriction on access to its advanced AI model, Mythos, which is specifically designed for cybersecurity applications. The company will limit access to approximately 150 organizations involved in its Glassw cybersecurity program, rather than releasing it to the general public.

The Mythos model is capable of autonomously identifying software vulnerabilities and generating exploit code to assist security teams in patching critical flaws. It has identified thousands of bugs across various browsers and operating systems, classifying high-risk systems and addressing significant offensive cyber capabilities.

This strategic decision reflects a broader shift in AI development, where companies treat cyber capabilities as a primary risk. By restricting access, Anthropic aims to ensure that the tool is used ethically to strengthen defenses rather than facilitate unauthorized exploitation, aligning with global measures to enhance national security.