Using ChatGPT, Claude or Copilot without exposing your company’s data
How to use ChatGPT, Claude or Copilot at work without exposing sensitive data: business plans, settings, anonymisation, usage rules and the right tools.
5 min read
Protection software · Published by Step By Step
AIGuard: anonymise your data before it reaches AI
AIGuard anonymises your data before it reaches AI. This filter, installed on the workstation, intercepts requests sent to artificial intelligence assistants, replaces personal data and secrets with reversible placeholders, then restores the real values in the answer you read.
- ChatGPT
- Claude
- Gemini
- Copilot
- Mistral
- Perplexity
How it works
How does AIGuard anonymise your data before it reaches AI?
Your sensitive data is replaced on your workstation before it reaches the AI, then restored on screen in its answer. The AI only works on placeholders; you read the real answer.
-
Local interception
The AIGuard agent runs on your workstation and captures the request at source, before anything leaves over the network.
-
Substitution before sending
Each piece of sensitive data is replaced with a consistent placeholder: a reversible token or a fictitious name. The AI never sees your real data.
-
Restoration on reading
In the answer, AIGuard puts your real values back locally. You read an accurate, natural answer; the AI only ever saw the placeholders.
What you type
Write a letter to John Smith, case no. 2024-118, payment to IBAN FR76 3000 …
What the AI receives
Write a letter to Tulip Cherry, case no. 2024-118, payment to IBAN FR76 9920 …
Illustrative example: the case number, which is not sensitive, goes through unchanged; the name and IBAN are replaced, then restored on display.
Features
What AIGuard does to protect your exchanges with AI
AIGuard combines several layers of protection, from the workstation to the administration console, so your teams can use AI without exposing company data.
-
Reversible substitution
Email addresses, bank details and IBANs, numbers, secrets and credentials are detected and replaced before sending. You choose to mask, use an anonymous token or a fictitious name.
-
System-level capture
A driver intercepts traffic to the main AI services, both in the browser and in installed applications.
-
Blocking sensitive files
A document containing personal data or secrets is blocked the moment you try to attach it to an AI, before anything is sent.
-
Temporary unblocking
A legitimate file or a false positive? Allow it in one click: it is unblocked for three minutes, then protection switches back on.
-
GDPR audit and compliance
Every exchange is logged (who, when, which service, which types of data were protected) without ever storing the sensitive content.
-
Centralised rules
Ready-made packs (GDPR, cloud secrets) and your own rules, deployed and signed across every workstation from the console.
First call on us
Make AI use in your business secure
AIGuard can be deployed on its own or with AIKeep. Let’s talk about how you use AI and which data needs protecting: we will recommend the right setup.
Going further
AIKeep and our artificial intelligence services
Blog
Further reading: our articles on the topic
-
-
AI, the GDPR and the EU AI Act: what a small business needs to know in 2026
Using AI in a small business in 2026: GDPR duties, the EU AI Act timeline, transparency, AI literacy for staff and choosing the right tools. An overview.
6 min read
FAQ
Frequently asked questions about AIGuard
What is AIGuard for?
AIGuard lets you use ChatGPT, Claude, Gemini, Copilot, Mistral’s Le Chat or Perplexity without handing them your sensitive data. Installed on the workstation, it replaces that data with placeholders before sending, then restores the real values in the answer on screen.
Does the AI still work properly with replaced data?
Yes. Placeholders are consistent (a fictitious name is still a name, an address is still an address), so the model can reason normally. Non-sensitive information, such as a case number, goes through unchanged.
Where are the mappings between placeholders and real data stored?
The mapping between each token and the real value stays encrypted on your workstation. The administration console, hosted in France, only receives anonymous metadata, never your personal data.
Which systems does AIGuard run on?
AIGuard is available on Windows 11, and a macOS version has been announced. Installation takes a single command line and the agent updates itself automatically.
How do AIGuard and AIKeep relate?
Both are published by Step By Step. Within AIKeep, AIGuard sits at the software’s exit point: everything heading for the AI model goes through it, and sensitive data is removed before it is sent.