Cisco will integrate Anthropic’s security-focused model, Claude Mythos 5, into its AI security solution AI Defense, Techzine reported on Nov. 9 local time.

The report said Cisco has been using Mythos through its participation in Anthropic’s Project Glasswing.

Cisco expects the integration to significantly strengthen red team testing aimed at AI used by companies. Red team testing is a security check that identifies system weaknesses from an attacker’s perspective.

Cisco announced AI Defense, an AI security solution, in January 2025. It has two approaches: standard and adaptive. The standard approach looks for common vulnerabilities such as prompt injection, jailbreaking and exposure of sensitive information. But when attacker methods become more sophisticated, the standard approach has limitations. The adaptive approach checks systems based on each company’s different environment and characteristics.

Mythos 5 will be deployed in the adaptive red team function. The company stressed that AI Defense can create multi-step attack paths to find vulnerabilities that are not easily exposed.

It is the same even if infrastructure becomes complex. All work is kept as records for audits.

Customers cannot use Mythos directly. Customers use the AI Defense screen, while AI Defense uses Mythos in the backend. A person provides Mythos with set tasks and goals, and the AI Defense red team engine deploys Mythos according to the task.