Activate AI Defense

Claim your subscription in SCC

Access to AI Defense capabilties depends on the subscription, or license, that you activate. The available subscription levels are explained in Licenses.

For all license types, you must activate your AI Defense subscription in Cisco Security Cloud Control (SCC). SCC allows you to launch and manage all Cisco security products. Follow the procedure below.

Onboard for AI Advantage

AI Advantage provides you with essential tools for securing both models in active use and models that you’re developing or using in your applications.

• For models in active use, AI Advantage offers runtime protection for real-time monitoring and control of user interaction with your AI models and third-party models. AI runtime enforces your policies that specify your rules for security and acceptable prompts and responses. Your protected AI models and model endpoints are saved as applications and connections in AI Defense. Event logs allow you to monitor AI model use and view the potential security, privacy, and safety violations detected by AI runtime.

• For AI models that you develop or offer, you can run validation to ensure that each AI model meets required safety and security standards. You can run validation on model endpoints that you specify, and you can take advantage of Multicloud Defense to discover all AI assets (models and applications) in your environment.

• To discover AI assets, you need to set up the Multicloud Defense-AI Defense integration and connect to the Amazon Web Services (AWS) account that hosts your models.

To set up your AI Advantage capabilities, click AI Defense: Dashboard and expand the Get started with AI Defense section at the top of the page. Follow these onboarding flows:

• Discover Assets: Follow this flow to set up your connection to Multicloud Defense to enable automatic discovery Assets.

• Scan your Assets: Follow this flow to set up AI validation to enable testing of automatically discovered Assets and models you specify.

• Secure prompts and responses: Follow this flow to set up runtime to enable real-time monitoring of your production AI applications with guardrails that block adversarial attacks and harmful responses in real time.

Onboard for AI Runtime Essentials

The AI Runtime Essentials license provides runtime protection for real-time monitoring and control of user interaction with your AI models and third-party models that are in active use. Runtime enforces your policies that specify your rules for security and acceptable prompts and responses. Your protected AI models and LLM endpoints are saved as applications and connections in AI Defense. Event logs allow you to monitor AI model usage, and and they let you see the potential safety and security violations detected by runtime.

To get started with runtime, you will choose how to connect AI Defense to your model or application. AI Defense can connect in the following ways:

• API: Add your AI Applications and create policies

• Gateway: Add your AI Applications and create policies

• Multicloud Defense: Configure Multicloud Defense and allow it to find assets

Onboard with AI Validation Essentials

For teams that develop or offer AI models, AI Validation Essentials provides model validation to ensure that each AI model meets required safety and security standards. You can run validation on model endpoints that you specify, and you can take advantage of Multicloud Defense to discover all Assets (models and applications) in your environment. To discover assets, you need to set up the Multicloud Defense-AI Defense integration and connect to the Amazon Web Services (AWS) account that hosts your models.

To get started with AI Validation Essentials, you will choose a connection type:

• Model: Connect your AWS Bedrock account and scan AI models hosted there

• Application: Manually add AI Applications to be scanned

• External: Scan a specific AI model endpoint for flaws

For an overview of subscription levels, see Licenses.