Skip to content
Power Prompt Documentation

Authorised AI Models Governance

To comply with regulatory standards (GDPR, EU NIS2, and the EU AI Act), the Platform Administrator defines model access policies across the enterprise.

1. Allowlist Principles

In Power Prompt, users and managers can only select AI models previously vetted and enabled by the administrator:

  • Eliminating Shadow AI: No user can connect an unapproved model without prior validation by the technical or security team.
  • Granular Provider Selection: Administrators can authorise specific providers (e.g. West Europe-hosted Azure OpenAI) while strictly disallowing public consumer API endpoints.
  • Sovereign Local Model Support: Capability to enforce self-hosted open-weights models (vLLM, Ollama, TGI) for business units handling confidential or regulated data.

2. Managing Authorised AI Models

  1. Navigate to Administration > AI Models & Providers.
  2. The detected models list displays each model along with its status:
    • Approved (Whitelisted): Immediately available in the model selector across all workspaces.
    • Disabled: Completely hidden and blocked from execution.
  3. For each approved model, administrators can configure:
    • Maximum Context Window.
    • Token Cap per Request (Max Output Tokens).
    • Restriction on unverified models for production environments.