Cisco has integrated Anthropic’s Claude Mythos model into its AI Defense platform to enhance adaptive red teaming capabilities for enterprise AI agents, according to CRN.
The integration, announced at Cisco Live in Melbourne, allows the security tool to utilize frontier-level attacker reasoning while maintaining strict safety protocols. By pairing the Mythos model with an adaptive red teaming engine, Cisco enables users to conduct multi-step, customized attack simulations against their own applications. The platform includes safeguards such as authorized target testing, named operator tracking, and full auditability to ensure exercises mirror real-world threats without compromising customer environments.
Cisco senior vice president DJ Sampath noted that the move from generic to adaptive red teaming is essential for identifying complex security gaps. The system also supports Amazon Bedrock and Amazon CloudWatch, providing visibility into token usage, which is increasingly relevant as attackers employ agent swarms. With approximately 500 customers since its 2025 launch, AI Defense is part of the broader Cisco Cloud Control platform. The company plans to expand these security capabilities to protect agents across laptops, cloud environments, and the edge, offering partners a new way to secure client deployments.
Source: CRN
