Cisco adds Anthropic's Claude Mythos to AI Defense for red teaming, with no open prompt access

Cisco will add Anthropic's Claude Mythos to AI Defense so the tool can plan attacks on customers' own AI apps and agents, a practice known as red teaming. The company announced the move at its Cisco Live event in Melbourne, according to The Next Web. Mythos will come first to select enterprise customers, and no date has been set.

What Mythos will do inside AI Defense

AI Defense attacks companies' own AI systems to find weak spots before hackers do. Mythos will plan tests and generate attack ideas against AI apps and agents.

Cisco has already put the model to work on this kind of task. In July, Cisco researchers used Mythos to push 15 chatbots past their guardrails.

Access controls

Cisco has built the integration around tight restrictions:

• Customers cannot type prompts to Mythos directly.
• Mythos works only on a fixed job, against systems the customer owns and has registered for testing.
• Every test is tied to a named person and recorded for audit.
• Any test can be stopped at any time.
• Mythos cannot scan the wider internet, Cisco said in a blog post.

Michael Moore, Anthropic's cybersecurity lead, said: "Enterprises are increasingly building AI agents into their core operations and want to test those agents thoroughly before putting them into production."

Pricing for the Mythos-powered capability has not yet been disclosed.

Part of a wider pattern of partner integrations

Other security vendors are building Mythos into their products. Palo Alto Networks launched always-on testing on Mythos and OpenAI's cyber model in September. IBM said on Monday it plans to add Mythos to its Concert Protect tool in early 2027.

The Next Web reports that this suits Anthropic, which puts Mythos inside partners' products rather than handing out the model. Anthropic released Mythos 5.2 on Monday, still only for vetted organisations. The report does not say which Mythos version Cisco will use.

The controls reflect a past incident. In April, an unauthorised group reached Mythos through a contractor's access.

Splunk updates

Cisco also updated its Splunk tools that track what AI agents cost and grade the quality of their work. A cost tool called Tokenomics enters private preview in December.

What this means

For Cisco customers, the integration is a managed red-teaming service rather than direct model access. The design choices (a fixed task, registered target systems, named-person accountability, audit records, a stop control, and no internet scanning) define how Mythos can be used inside AI Defense. Customers get a model-driven attack planner without being able to point it at arbitrary targets.

The April incident, in which an unauthorised group reached Mythos through a contractor's access, explains why those limits matter. Access through a vendor's product with logging and attribution is the route Anthropic is using, rather than distributing the model itself.

The rollout is limited for now: select enterprises only, no date set. Teams building AI agents who want Mythos-driven testing before production deployment will need to wait for Cisco to open access.