ChatGPT · Copilot · Gemini · Claude
Anonymize your documents before you give them to ChatGPT
Asking ChatGPT, Copilot, Gemini or Claude to summarize a contract, review a brief or draft a reply to a client saves a lot of time. But when you paste the text or upload the file, names, ID numbers, addresses and case details leave your organization for a third party with its own terms of use and retention. For a law firm, an accounting practice or an HR team, that can clash with the GDPR or with professional secrecy.
In Spain, plans are arranged with our sales team.
Blacking things out does not solve it: a text full of black boxes leaves the AI without context and the answers get worse. With Anonimatum's pseudonymization, every person is replaced by a consistent fictitious name, dates are shifted while keeping the intervals, and amounts keep their proportions. The assistant understands who did what, but not who anyone is.
Each run produces an equivalence map that stays with you. With it you know that Marta Gil is really your client, and you can put the real names back into the answer. Anonimatum itself does not use external language model APIs: the substitution is done by its own AI on EU servers.
From your document to the AI's answer
- 1
Pseudonymize the document
Upload the Word file or PDF to Anonimatum and apply pseudonymization: names, ID numbers, emails and addresses become fictitious and consistent.
- 2
Check the map
Make sure in the equivalence map that every person has a substitute and that no real data is left in the text.
- 3
Work with the AI
Paste or upload the pseudonymized version into ChatGPT, Copilot, Gemini or Claude and ask for the summary, review or draft.
- 4
Put the real names back
Use the equivalence map to swap the pseudonyms in the answer for the original data before you use it.
Fictitious document
Before and after: an employment query for the AI
Fictitious email from an employee to her adviser. People, company, ID number and phone are replaced with consistent equivalents before asking the AI for a draft reply.
From: Lucía Romero Vidal
To: employment adviser
I have worked at Transportes Arenal, S.L. since 2019, ID 12345678Z.
My manager, Raúl Esteban Mora, told me about a shift change
with no notice and a pay cut of €300 a month.
Raúl Esteban Mora says it is a management decision.
Can I challenge it? My phone number is 600 000 000.
From: Marta Gil Soto
To: employment adviser
I have worked at Logística Norte, S.L. since 2019, ID 87654321X.
My manager, Óscar Prieto Lara, told me about a shift change
with no notice and a pay cut of €300 a month.
Óscar Prieto Lara says it is a management decision.
Can I challenge it? My phone number is 611 111 111.
The same person always gets the same pseudonym, so the AI can follow the case; the facts and the amount stay because they are exactly what needs analysing.
Why pseudonymize instead of just blacking out
AI needs context
With black boxes the assistant cannot tell the employee from the manager or the company. With consistent pseudonyms, the answers stay useful.
Consistent throughout the document
The same original always produces the same substitute, so a person mentioned twenty times is still one person for the AI.
The map stays with you
The equivalence map is exported as JSON or encrypted and never leaves your organization. Without it, the pseudonyms point to no one in particular.
No external LLM in between
Pseudonymizing does not require sending the original to any third-party model: Anonimatum processes it with its own AI in the EU.
Anonymizing before using AI: frequently asked questions
Is pseudonymized data still personal data?
Yes: under the GDPR it remains personal data as long as a map exists that can reverse it. But the risk changes considerably, because the AI provider only receives pseudonyms while the map stays with you. If you will not need the real names, anonymize instead of pseudonymizing.
Is removing names enough?
Rarely. An ID number, a case reference, an address or a very specific detail can also identify someone. Review everything detected in the preview and add patterns or word sets for anything specific to your field.
Does it work the same with ChatGPT, Copilot, Gemini and Claude?
Yes, because Anonimatum works beforehand: you prepare the document, then use it in whichever assistant you prefer. There is no integration with those tools; the result is an ordinary Word file or PDF.
Can I paste text instead of uploading the file?
Yes. Download the pseudonymized document and copy the passage you need. What matters is that the AI only ever receives the pseudonymized version, never the original.
Does Anonimatum use ChatGPT under the hood?
No. Detection and substitution run on Anonimatum's own models on servers in the European Union, with no external language model APIs. Your original document never passes through a generative AI provider.
Use AI without handing over your clients' data
Pseudonymize your next document at app.anonimatum.com and keep the equivalence map to restore the real names afterwards.
Try it in the appNot in Spain?