Cisco releases open-source language for governing agentic AI behavior
Cisco has launched an open-source domain-specific language to define policies for autonomous AI agents. The move aims to provide the security confidence required for enterprises to deploy agentic systems at scale.
Cisco is launching an open-source domain-specific language (DSL) designed to prescribe and govern the behavior of agentic AI systems. The tool allows administrators to define specific policies that agents must adhere to within their allocated token budgets. This initiative complements Cisco AI Defense, a product announced in 2024 that scans AI supply chains for vulnerabilities and provides runtime guardrails against threats such as prompt injection and tool poisoning.
To build this security stack, Cisco acquired Galileo for evaluation capabilities, Robust Intelligence for model red-teaming, and Asterix Security for non-human identity management. The company also open-sourced Defense Claw, a tool for agentic security, and collaborated with NIST, MITRE, and OWASP to create a public AI safety taxonomy. Cisco Secure Access is used to block unauthorized 'shadow agents' that employees might connect outside the established security boundary.
Sampath noted that Cisco AI Defense projects have collected over 5,000 GitHub stars. He argued that security should be viewed as an accelerator for AI adoption rather than a roadblock, providing the confidence needed to deploy autonomous agents. "We basically said there needs to be a domain-specific language that describes these agentic behaviors," Sampath said.