Cisco AI Defense

Cisco AI Defense is an enterprise security product from Cisco aimed at organizations that use third-party AI applications or build their own AI applications. Cisco describes it as end-to-end protection covering asset discovery, risk detection, and real-time threat mitigation, with an Explorer Edition for red teaming AI models and agents.

What It Is

Cisco AI Defense is an AI security solution that addresses two scenarios. For organizations using AI, it surfaces third-party AI applications in use and lets teams define policies for employee access, threat protection, and sensitive data loss prevention. For organizations developing AI, it detects AI assets across the environment, assesses models algorithmically for vulnerabilities, and deploys guardrails to protect AI applications in real time.

Core Components

The product page lists five components:

• AI Model and Application Validation: algorithmic red teaming to identify safety and security vulnerabilities across models.
• AI Runtime Protection: guardrails embedded in the network that block adversarial attacks and harmful responses in production.
• AI Cloud Visibility: automatic inventory of AI models and connected data sources across distributed environments.
• AI Access: monitoring and managing access to third-party AI applications, with policies that limit sensitive data exposure.
• AI Supply Chain Risk Management: governance and security over AI models and files.

Detection and Network Enforcement

Cisco states that it enforces AI security at the network level without agents or libraries, decoupling AI development from security. It says its guardrails go beyond prompt injection and malicious URL detection to cover model denial of service, code detection, and off-topic attacks. Cisco says detections are informed by threat intelligence from its AI research lab and Talos, and that Splunk customers can use Splunk data for additional insights.

Standards Alignment and Explorer Edition

Cisco says AI Defense helps align to NIST, MITRE ATLAS, and the OWASP LLM Top 10 through a single integration. The Explorer Edition lets AI builders red team applications before moving to the enterprise solution, producing a report for security reviews. Access to the full product starts with a demo request.