Frequently Asked Questions

Because real usage already exists. In most organizations, employees use generative AI (ChatGPT, Claude, etc.) to move faster — often without a security layer. That’s Shadow AI.

SafeDoc doesn’t try to replace internal policies. It turns an invisible risk into controlled usage: anonymize before sending, keep control, then de‑anonymize on the way back.

Shadow AI is the use of external AI tools outside an official framework (no control, no audit, no document protection). It can be small (pasting a paragraph) or massive (whole case files).

SafeDoc is a document security layer that makes this usage safer without killing productivity.

No. SafeDoc is not an analysis AI. It’s a security layer that enables you to use existing AI tools on sensitive documents while significantly reducing exposure risk.

SafeDoc follows a simple workflow:

• Anonymize sensitive information before sending to external AI
• Reduce overly identifying context (depending on level)
• Generate a copy‑ready version + mapping (depending on mode)
• De‑anonymize the AI output back using the mapping

Goal: preserve analytical value while significantly reducing risk.

• Personal data: names, emails, phones, addresses, IDs…
• Confidential information: org names, internal references, numbers, IBAN, amounts…
• Overly precise context: exact dates, fine locations, financial precision (depending on level)

Scope depends on mode (Anonymize / Pseudonymize / Fake data) and level (N1–N3). N3 is on the roadmap for Q3 2026.

Yes. Use the restore flow: import your mapping file (e.g., mapping_xxx.json) and paste the AI answer containing tokens.

SafeDoc re-injects values locally so the analysis becomes usable.

SafeDoc is designed to minimize exposure:

• Controlled processing with automatic purge (depending on configuration)
• Encryption in transit
• No use of your documents to train models

For contractual/GDPR details: see the legal page →

A summary tool processes content — it doesn’t solve the core issue: sensitive documents being sent to external AI without protection.

SafeDoc is a document security layer: anonymize before AI, keep control, then de‑anonymize the output back.

• Single document: one contract, one report, one piece.
• Data Room: a batch (contracts + annexes + supporting docs). Pseudonymization is consistent: the same entity keeps the same token across the batch.

Advanced is currently in development (roadmap). For launch, we keep it muted.

No. SafeDoc aims to significantly reduce exposure and re-identification risk, but no tool can guarantee “zero risk” in every context. Final review and correct settings remain essential.