Frequently asked questions.

A security layer between employees and AI models. BRANE inspects every prompt, detects sensitive content, applies policies and decides whether a request is allowed, masked, blocked or processed locally.

BRANE is a hardware appliance on NVIDIA hardware — plug it in, configure, get going. Optionally with controlled cloud access in hybrid mode.

Local-only, local-preferred (recommended for enterprise rollout), cloud-preferred with controls, cloud-only with governance, and air-gap capable.

Yes. BRANE sits in front: sensitive prompts are masked or processed locally, non-critical ones go to the best model — per policy.

Classification in under 200 ms per prompt — with no noticeable delay in the workflow.

Yes. Audit logs with content, documented routing decisions and pseudonymization provide the evidence GDPR and EU AI Act Article 50 require.

BRANE is designed for self-service deployment in under 30 minutes. For larger environments we offer optional onboarding (identity, SSO, custom policies).

In local mode, no prompt content leaves your environment. In hybrid mode only masked or tokenized content leaves the appliance — never cleartext.