Cisco has unveiled new AI security, observability and infrastructure-intelligence capabilities designed to help enterprises secure, govern and measure increasingly autonomous AI agents.
This brings Anthropic’s Claude Mythos into Cisco AI Defense for adaptive red teaming, alongside Splunk Agent Observability enhancements and Cisco IQ integration with Cisco Data Fabric powered by the Splunk Platform.
Cisco said AI agents are becoming more capable and more deeply connected to enterprise tools and sensitive data, increasing the need for continuous testing, governance and visibility. Its AI Defense platform has helped hundreds of customers identify critical flaws in AI models, applications and agents before production deployment since its launch in January 2025.
The addition of Claude Mythos is intended to strengthen Cisco’s adaptive red-teaming framework by applying advanced adversarial reasoning to AI systems. It can generate multi-step, contextual attack paths designed to uncover subtle vulnerabilities across complex architectures, while results can be carried forward into follow-up attack loops.
Cisco said customers can interact only through AI Defense’s structured interface, with no open-ended prompt access. Every engagement is authenticated and cryptographically tied to a named enterprise operator, runs against explicitly registered and authorised targets, and can be halted on demand.
“Enterprises are increasingly building AI agents into their core operations and want to test those agents thoroughly before putting them into production,” said Michael Moore, cybersecurity lead at Anthropic.
“By bringing Claude Mythos to Cisco AI Defense, Cisco can help security teams red team their agents with advanced reasoning that plans tests and generates attack ideas for each system.” Michael Moore
Measuring AI value
Cisco’s enhanced Tokenomics capability is designed to connect AI spending directly with return on investment in real time. Available in private preview in December and expected to reach general availability in early 2027 through AWS Marketplace, it uses usage data from Amazon Bedrock and Amazon CloudWatch to pair each agent’s cost with evaluations of the quality and impact of its work.
Splunk Agent Observability also introduces Eval-G1 evaluation models, which turn agentic traces and expert feedback into custom small language models for evaluations and guardrails. Cisco said these models can run at up to 98% lower cost than a commercial large language model and deliver evaluation response times 2.5 times faster than traditional LLMs.
The Evaluation Tuner enables teams to create fine-tuned evaluator models from agentic traces and subject-matter-expert feedback, helping enforce quality standards and steer agents at machine speed. The evaluation models and tuner are generally available from this month.
Accelerating remediation
Cisco is extending Cisco IQ intelligence into enterprise security environments through Cisco Data Fabric powered by the Splunk Platform. The integration ingests vulnerability, lifecycle and configuration data into Splunk via open APIs, allowing security operations teams and AI agents to correlate live threat detections with underlying network exposure in Splunk Enterprise Security.
New Model Context Protocol and ServiceNow connectors provide structured infrastructure context to supported AI agents and IT service management platforms. They can automatically translate exposure findings into asset-linked ServiceNow problem records, reducing manual data reconciliation and helping coordinate remediation. Cisco introduced an AI-powered resilient infrastructure playbook that surfaces critical lifecycle vulnerabilities, personalised alerts for known defects and quantum-safe cryptographic readiness.
Cisco President and chief product officer Jeetu Patel said: “You cannot scale what you cannot secure, govern, or observe.” He added that the company is building a continuous trust loop that protects agents from compromise while ensuring every token spent delivers measurable business value.