Cisco to Add Claude Mythos to AI Defense for Controlled Red Teaming
What happened
On November 9, 2026, Cisco announced at its event in Melbourne that it will integrate Anthropic’s Claude Mythos into its AI Defense adaptive red-teaming framework. This integration will allow enterprises to test AI applications and agents within strict, authorized boundaries. Cisco also introduced updates to Splunk agent observability, linking agent costs to the quality of their output, and expanded its Cisco IQ integrations for infrastructure remediation. The new features are initially available to select enterprises, with broader rollout planned but not yet scheduled.
Key details
- Claude Mythos will only operate on targets registered and authorized by enterprise customers.
- Every red-teaming session is authenticated, time-limited, and logged for auditing.
- Splunk's Enhanced Tokenomics will enter private preview in December 2026, with general availability in early 2027.
- Eval-G1 evaluation models and the evaluation tuner are generally available as of November 2026.
- Cisco IQ serves about 17,000 customers and covers over 22 million connected devices.
Why it matters
Enterprise security teams using Cisco’s AI Defense will be able to conduct controlled red-teaming of their AI systems, with clear boundaries and audit trails, reducing the risk of unauthorized testing or data exposure. The Splunk updates give organizations a way to directly connect the operational cost of AI agents to their effectiveness, helping them manage budgets and improve agent performance. Customers using Cisco IQ can now automate the connection between detected vulnerabilities and remediation workflows, streamlining the process of addressing security issues across large, complex IT and