Cisco offers Cisco AI Defense, launched in January 2025, which uses AI to help enterprise security teams discover unauthorized AI applications in use across their organization, automatically test AI models for hundreds of safety and security vulnerabilities, and protect running AI applications from attacks like prompt injection and data leakage.

Cisco AI Defense operates at the network level, parsing DNS and cloud logs to identify both sanctioned and shadow AI applications across cloud environments. It then uses proprietary machine learning models and large language models (LLMs) to run automated 'algorithmic red teaming' — simulating hundreds of potential attacks against an enterprise's own AI models to identify vulnerabilities before they are exploited. At runtime, it applies guardrails to block adversarial inputs and prevent sensitive data from leaking through AI model outputs. Cisco describes it as 'self-optimizing, leveraging Cisco's proprietary machine learning models' fed with real-time threat intelligence from Cisco Talos. The biggest expansion of the product since its launch was announced in February 2026, extending protections to agentic AI systems.