Cisco and IBM are both preparing to build Anthropic's Claude Mythos into their security products, a move that would put one of the company's most closely guarded models in front of select corporate customers. Cisco announced the plan at its Cisco Live event in Melbourne. IBM said on Monday that it intends to follow suit.
What is red teaming in Cisco's AI Defense tool?
Cisco will add Mythos to AI Defense, a tool that attacks companies' own AI systems to find weaknesses before criminals do. The practice is known as red teaming. In the model's role, Mythos will plan tests and generate attack ideas against AI apps and agents. Cisco researchers already used it in July to push 15 chatbots past their guardrails.
What restrictions apply to Claude Mythos in Cisco's tools?
The company has built in several restrictions. Customers cannot type prompts into Mythos directly. The model works only on a fixed job, and only 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.
Mythos will reach select enterprise customers first, and Cisco has not set a date for wider availability. Michael Moore, Anthropic's cybersecurity lead, said the demand is coming from businesses that are building AI agents into their core operations.
“Enterprises are increasingly building AI agents into their core operations and want to test those agents thoroughly before putting them into production,” Moore said.
How does IBM plan to use Claude Mythos in Concert Protect?
IBM's plan runs along different lines. The company wants to add Mythos to Concert Protect so customers can scan their own code for flaws that nobody has found yet, including ones that standard scanners miss. Concert Protect would then check whether hackers could actually exploit each flaw in that customer's systems. IBM and Red Hat tools would draft fixes, but developers would have to review and approve them before they go in.
IBM cautioned that its plans could change. The company is not relying on a single AI lab. In June, it joined OpenAI's cyber program to offer a similar service.
IBM framed the problem in terms of what happens after a flaw is found. “The bottleneck is no longer finding vulnerabilities. It is determining which can actually be exploited and fixing them before attackers do,” the company wrote.
That gap is wide. Project Glasswing, Anthropic's restricted Mythos program, flagged more than 10,000 possible serious flaws in its first month. By then, only 97 had been patched. IBM said its earlier Mythos work through Glasswing was internal research, and that the Concert Protect plan is its first step toward selling the capability to customers.
Both moves fit Anthropic's strategy of placing Mythos inside partners' products rather than releasing the model on its own. Users of those products would receive findings and fixes without ever touching the underlying model.
Access remains tightly controlled. Anthropic briefly switched Mythos off in June to comply with U.S. export controls, and the White House now reportedly decides who joins its partner list. Anthropic released Mythos 5.2 on Monday, but it is still available only to vetted organizations.
Security was not the only AI news from Cisco. The company also updated its Splunk tools that track what AI agents cost and grade the quality of their work. A cost tool called Tokenomics will enter private preview in December.
This article was written with the assistance of AI.
News Factory APP - agentic news to boost your SEO & AEO.