Cisco has added Anthropic's Claude Mythos model to its AI Defense security platform, according to detikcom. The integration allows corporate security teams to test AI applications and agents through simulated attacks before those systems reach production.
Jeetu Patel said AI deployment cannot be scaled if the systems involved cannot be secured, managed and monitored.
Drivers of security testing
Global research by Splunk last month found that 81 percent of ITOps professionals and engineers saw more collaboration between security and monitoring teams.
Tokenomics in Splunk Agent Observability enters limited preview in December and becomes broadly available on AWS Marketplace in early 2027. The Eval-G1 evaluation model is claimed to run at up to 98 percent lower cost and answer 2.5 times faster than commercial large language models.
The simulations sit on a platform that hundreds of corporations have used to find critical gaps in AI models and applications since it launched in January 2025. Newer agents go through attack scenarios while they are still being assessed, before anyone puts them into production.
Testing methods and follow-up
The Evaluation Tuner and Splunk's evaluation models became generally available this month. AI Defense already carries standard red teaming that checks thousands of vulnerability categories and refers to the OWASP, MITRE and NIST frameworks, Kompas.com reported.
Liz Centoni said knowing where a company is exposed is only useful if the company can act on the finding.
Who is affected
Cisco IQ watches more than 16,000 customers across almost 20 million devices, whose data flows into corporate security infrastructure. Testing runs only against registered, authorized targets, with operator identities verified cryptographically and sessions that can be stopped at any time.
Adaptive red teaming lets users set test criteria in everyday language, and the system then builds attack scenarios and adjusts its tactics. Red teaming for Claude Mythos will arrive soon for a selected group of enterprise customers.
That group is narrower than the customer base using AI Defense today, so the Claude Mythos capability reaches a chosen set of organizations rather than every buyer.
What is still unclear
Cisco has not disclosed pricing or a licensing scheme for the Claude Mythos integration, and it has not described how the selected enterprise customers will be chosen. No price list or contract terms have been published.
It is also unconfirmed whether results from AI Defense testing can be used outside the Splunk and Cisco ecosystem, and no public schedule shows when the capability moves beyond that first group.