Prerequisites

Mindgard dynamically tests your AI model or application against a wide range of cyber attack techniques to evaluate its strengths and weaknesses.

For this reason Mindgard requires:

• A running instance of the AI model or application.
• An API to run tests against.

Such an API is usually either:

• A model inference API endpoint.
• An application API endpoint.

To test via Mindgard Web or CLI interfaces, these APIs must communicate JSON over HTTP, either through RPC or streaming. If your API structure differs from this (for example if there is no network accessible API), consider the Python SDK option.