Cisco Brings Anthropic's Claude Mythos Into AI Defense for Red-Teaming Agentic AI

Cisco announced on Monday (November 9) at its Cisco Live Asia Pacific and Japan event in Melbourne, Australia, that its AI Defense platform will incorporate Anthropic's newly released frontier model Claude Mythos for adaptive AI red teaming, using frontier adversarial reasoning to simulate real attacks and test enterprise AI applications and agent systems.

Frontier model to simulate real attacks

According to Cisco's announcement, Claude Mythos brings frontier adversarial reasoning, which AI Defense will use to run continuous adversarial tests on enterprise AI applications and autonomous agents, surfacing weaknesses attackers could exploit before models go live. Cisco said AI Defense has helped hundreds of customers find critical flaws before deployment since its launch in January 2025. Claude Mythos is part of Anthropic's new model family unveiled the same day alongside Fable, aimed at frontier reasoning tasks.

Agent observability tools launched alongside

Cisco also launched Splunk Agent Observability and new Cisco IQ capabilities to help enterprises monitor agent behaviour and performance. Tokenomics, which ties agent costs to return on investment in real time, enters private preview in December and becomes generally available on AWS Marketplace in early 2027; evaluation models and the evaluation tuner are generally available this month.

Sources: Cisco press release (via PRNewswire).