Cisco and IBM Integrate Anthropic’s Mythos Model into Security Suites

Anthropic is expanding the reach of its bug-finding Mythos model by integrating it into the security portfolios of Cisco and IBM, as The Register first reported. The arrangement allows these enterprises to offer Mythos' capabilities to customers without granting them direct access to the model itself.

Cisco announced the integration during an event in Melbourne, Australia, stating that the red teaming tools within its AI Defense suite—designed to identify weaknesses in AI models, agents, and applications—will now be "super-charged by Mythos." Jeff Schultz, Cisco’s senior vice president for portfolio strategy and marketing, noted that the tool can "look at models and break them," and that Mythos will help identify more dangers for AI Defense to address. Cisco stated this will be available under existing licenses at no additional cost.

Simultaneously, IBM announced it will integrate Mythos into its Concert tool. IBM intends to use the model to uncover weaknesses that traditional scanners miss, determine actual risks, and generate fixes for developer approval.

This strategic shift toward providing functionality as a service allows Anthropic to satisfy high demand while mitigating security risks. Natalie Mead, Anthropic’s head of applied AI for the Asia-Pacific, noted a "long queue of buyers" for the model. By locking the model in a "black box" via third-party integrations, Anthropic prevents the model from falling into the hands of attackers who could use it to exploit software flaws. Furthermore, Anthropic has implemented protections against "distillation attacks," which the company believes Chinese AI labs use to clone its models.