AI Defense
AI Defense empowers organizations to confidently adopt generative AI by providing a cutting-edge, user-centric, and transparent security solution. This suite is designed to ensure the highest levels of data protection, compliance, and ethical integrity in an evolving AI landscape. This suite will provide organizations with the tools they need to secure their AI operations, maintain compliance with relevant regulations, and uphold ethical standards in their use of generative AI.
-
Intercept unsafe AI traffic in real time: The AI Defense runtime component enforces your AI Policies to ensure unsafe traffic to and from LLMs and chatbots is intercepted and logged. The Policies tab allows you to set up rules for allowed and disallowed types of interactions with LLMS, and the Events tab reports violations as they happen.
-
Validate your AI models to ensure safety: AI Validation lets you assess the vulnerabilities in the generative AI models your organization uses. AI Validation tests probe each AI model with attack techniques and intents designed to elicit undesirable behavior or information disclosure. Test results give you a detailed analysis of the top types of risks posed by each model, and they allow you to see how the model responded to each threat type and attack technique. AI models and applications must be discovered as AI Assets before you can scan them.
-
Discover AI applications: AI App Discovery allows you to find out which chatbots and AI applications your users are interacting with, and assess their risks. AI App Discovery (AI Access) uses Cisco Secure Access to scan for third-party AI applications that your users connect to. For each, it provides a risk score and details about the traffic to and from that AI application. Through the integration with Secure Access, you can click on any application to see detailed risk information.
-
Detect AI workloads in your cloud environment: The AI Assets component relies on Cisco Multicloud Defense to detect all AI workloads in your environment, including AI models, agents, and knowledge bases. This feature currently supports workloads that rely on Bedrock-hosted models.
-
Runtime Protection: Delivers robust, real-time security through AI Gateway, monitoring and enforcement of prompts and responses, ensuring compliance and safety during AI interactions.
-
AI Gateway: Acts as a protective barrier, filtering and securing AI interactions, preventing malicious activities or non-compliant behaviors.
-
Prompt/Response Monitoring & Enforcement: Continuously tracks AI prompts and responses, enforcing policies to maintain ethical and compliant AI operations.
-
Configuration & Dashboard: Provides users with an intuitive dashboard and flexible configuration options, making it easy to set up and manage security settings, monitor AI activity, and access report.
-
Application and Connection Configuration: Simplify application setup and establish secure connections.
-
Policy and Guardrail Assignment: Assign, monitor, and manage policies to ensure compliant operations.
-
AI Validation: Systematically assess the performance, reliability, compliance, and safety of your AI systems.
-
Risk Exposure Analysis: Understand risks, vulnerabilities, and usage patterns for each AI system.
-
Proactive Protection Recommendations: Investigate risks and receive tailored recommendations to enhance security.