Cisco brings Claude Mythos into AI Defense for agent red teaming, IBM targets early 2027 general availability

Cisco announced on November 11 that it is bringing Claude Mythos into AI Defense to red team AI applications and agents, stress-testing AI systems with the model's frontier adversarial reasoning. Alongside the latest advances in upgraded Splunk Agent Observability and Cisco IQ, Cisco said it is helping enterprises prepare for a trusted era with runtime protection, continuous observability and effective governance for agents.

Cisco global president and chief product officer Jeetu Patel said that without security, governance and observability there can be no scale, and that combining frontier AI runtime protection with machine-speed agent quality control is building a continuously running trusted loop. Since launching AI Defense in January 2025, Cisco said it has helped hundreds of customers identify critical defects before models, applications and agents reach production.

On mechanism, AI Defense's adaptive red team testing lets security teams quickly build custom multi-step attack test plans against target systems, and Claude Mythos strengthens that framework with advanced reasoning, helping plan genuinely meaningful tests and turning results into attack ideas for each system. Cisco built in multiple safeguards, including registered and authorized test targets only, named operator attribution, revocable execution with time-limited sessions, and dual-layer oversight. Adaptive red team testing with Claude Mythos is coming to selected enterprise customers, with Tokenomics opening a private preview in December and a planned general availability through AWS Marketplace in early 2027. On the IBM side, Concert Protect will scan customer-selected code repositories and assess whether weaknesses could be exploited, offering proposed edits that developers may accept or reject, targeting general availability in early 2027.