Cisco to Add Claude Mythos to AI Defense Testing

Cisco will add Anthropic’s Claude Mythos to its AI Defense platform to help enterprises test their artificial intelligence applications and agents for security weaknesses, the company announced November 9 at Cisco Live APJC in Melbourne. The integration is intended to generate tailored, multistep attacks that probe how a system behaves, including whether it can be manipulated into exposing data, misusing tools or taking unauthorized actions.

Cisco says Mythos will augment, rather than replace, AI Defense’s adaptive red-teaming system. The platform already tests applications for vulnerabilities such as prompt injection, jailbreaks and sensitive-data exposure. Adaptive testing examines a particular system and adjusts simulated attacks in response to its behavior. Cisco says Mythos will add more advanced reasoning to help devise attacks that reflect how an individual application or agent is used.

Access will initially be limited to selected enterprises. Cisco plans to extend access to qualified customers but has not announced a general-release date. The company says customers will not be able to prompt Mythos directly. Tests will run only against systems registered and authorized...