Every prompt is checkedbefore it leaves your network.
BRANE classifies, masks, routes and audits every prompt. Locally, in the cloud or hybrid. Made in Stuttgart, EU-sovereign by design.
What happens in a personal account
Every prompt is a potential data leak.
An employee pasting a customer list is doing nothing forbidden. They are using their personal account, and the rules there are not the ones your company signed.
- Training
On by default in personal accounts.
With ChatGPT and Gemini, training is the preset. Switching it off is up to the user.
OpenAI Help Center · 2026 - Humans read along
The vendor warns you itself.
Google asks its users not to enter anything they would not want a reviewer to see.
Google Gemini · 2026 - Retention
One thumbs up, five years.
Rate an answer in Claude and the entire conversation is filed away for up to five years.
Anthropic · 2026 - No way back
Three years after you delete it.
What a Gemini reviewer already read stays. Deleting your activity does not reach it.
Google Gemini · 2026
Key Capabilities
Six capabilities. One central control layer.
Every step between employee prompt and cloud AI — visible, controllable, auditable.
Detect what nobody else sees.
Real-time classification of every data class in under 200 ms. Detects PII, IP, contracts, code, trade secrets.
Understand what is between the lines.
Not every confidential text carries a pattern. BRANE weighs what the text is about and treats it as sensitive when it describes the business itself — even when it holds no name, no number and no address.
Mask before it leaks.
Sensitive entities are pseudonymized before cloud egress. The AI sees only the context. Detected sensitive data becomes placeholders. Automatically reversible via on-prem vault.
Route locally what stays local.
Sensitive classes stay in your environment, low-risk classes reach the cloud's best model.
Continuous audits for every prompt.
SHA-256 hash chain as append-only trail. Export as CSV, JSON or PDF — ready for your SIEM. EU AI Act compliant from day one.
Guide people before they send.
In-line guidance inside the AI tool — not a lecture pop-up after the leak. Sandboxes for safe experimentation.
With BRANE in between
The same three providers. Four points different.
BRANE sits between your people and the cloud model. The same question goes out, without the part that has to stay in.
- Training
No training on your data.
The cloud model never receives sensitive data. What it never gets, it cannot train on.
- Humans read along
Reviewers see no sensitive data.
Names and numbers are replaced before sending. Anyone reading along sees only the masked request.
- Retention
Only masked data is stored.
If the provider keeps the request, it holds nothing that leads back to your company.
- No way back
There is nothing to retrieve.
No plain text sits with the provider. Which placeholder stands for whom is known only to your server.
Enforce policy, prove every decision.
Policy enforcement
Turn company AI rules into automatic controls.
Audit trail
Document what happened, who acted, which policy applied and where the request was processed.
Implementation
Self-service deployment in under 30 minutes.
BRANE is a turnkey AI Interaction Firewall: plug it in, connect users, pick a mode, get going — without a long implementation project.
- 01
Deploy BRANE
Connect power and network on a certified appliance, or roll it out to spec on your own hardware or a virtual platform. In every form BRANE ships preconfigured with Safe Prompt, Guardian Routing, Audit Trail and the Admin Console.
- 02
Open the Admin Console
Start the guided setup and select your deployment mode: local-only, local-preferred, cloud-preferred or cloud-only with controls.
- 03
Connect users & access
Connect your identity setup or create the first user groups. Define who gets access and which teams start.
- 04
Activate policies
Choose predefined policy templates or configure your own rules for sensitive data, model access, masking, blocking and local routing.
- 05
Start secure AI usage
Employees use AI through the BRANE interface. Every prompt is checked, routed and logged — per your policies.
Enterprise onboarding available for identity, SSO and custom policy integration.
All of BRANE
Where BRANE sits, how it protects, how it runs, which models it connects, what keeps running beside it, and what comes next.
Further information
BRANE. In your environment. Today.
30-minute live demo with a BRANE engineer. See classification, masking, routing and audit on one control layer: in one session.
CONTACT
Let's get in touch.
Request a demo, book a workflow audit, or just say hello — we reply within one business day.
Or reach us directly at: info@ai-z-group.com