Cisco adds Anthropic’s Claude Mythos to AI Defense for red teaming

Cisco will add Anthropic’s Claude Mythos to AI Defense to plan attacks on customers’ own AI apps and agents, a practice called red teaming. Customers get no open prompt access, every test is logged to a named person, and it comes first to select enterprises.

Cisco will add Anthropic’s Claude Mythos to AI Defense, the company announced at its Cisco Live event in Melbourne. AI Defense attacks companies’ own AI systems to find weak spots before hackers do. Mythos will come first to select enterprise customers, with no date set.

Mythos will plan tests and come up with attack ideas against AI apps and agents, a practice known as red teaming. In July, Cisco researchers used it to push 15 chatbots past their guardrails.

Customers cannot type prompts to Mythos directly. It works only on a fixed job, against systems the customer owns and has registered for testing.

Every test is tied to a named person, recorded for audit and can be stopped at any time. Mythos cannot scan the wider internet, Cisco said in a blog post.

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

Security firms are lining up to build 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.

That suits Anthropic, which puts Mythos inside partners’ products rather than handing out the model. It released Mythos 5.2 on Monday, still only for vetted organisations. Tight controls matter: in April, an unauthorised group reached Mythos through a contractor’s access.

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.