Hooks can send prompts to an organization’s security server, which decides whether to allow or deny a request before inference. Learn how this control relates to protecting data when using AI.
Inference hooks can send each prompt subject to control to an organization’s security server. That server decides whether to allow or deny the request before it reaches inference.
Engineers can use this mechanism to apply security checks before processing. The description does not specify which checks are performed or how the system is implemented.
If you study or apply this approach with AI, avoid putting personal data or confidential information in prompts. Use fictional examples or remove identifiers, and check your organization’s policies before submitting internal material.
Keep the policy decision separate from model processing: define which requests should be allowed or blocked, then test the expected cases. Do not assume the described mechanism is available in Rota Nacional.