Privacy policy

1. Who we are

Pseudonymizer (pseudonymizer.eu) is a service of Litic B.V., a private limited liability company (besloten vennootschap) under Belgian law, trading as Litic AI, Uitbreidingstraat 84, 2600 Antwerp, Belgium, enterprise number 0817.919.143, VAT number BE0817919143. Privacy questions? Email privacy@litic.ai or call +31 40 369 00 28.

Our role depends on the type of data. For the content of the documents you have processed, the legends, the results, the packages and the dispatches, you (or your organization) decide why and how they are processed; we act as processor and process those data only on your instructions. Our standard data processing agreement automatically applies as part of the agreement, for business customers and for consumers (article 9 of the terms of service). The text is published at pseudonymizer.eu/data-processing-agreement (Dutch: pseudonymizer.eu/nl/verwerkersovereenkomst); a signed copy is available on request via privacy@litic.ai at no extra cost.

If you are a business customer, you are the controller of those data. If you are a consumer and use the service only for yourself or your household, for example for your own mail, your own employment or divorce case or the medical records of a family member, this often falls under the household exemption of Article 2(2)(c) GDPR. The GDPR then does not apply to you, but it does apply to us; we then process the data in the same way and with the same safeguards. If you publish the result, make it available to an indefinite number of people or use it for a purpose outside your private life, such as an association or volunteer work, the GDPR does apply to you and you are the controller.

For account, billing, usage and connection data we are the controller ourselves. This policy covers both kinds of processing.

2. What data we process

• Account data: your name and email address, the sign-in method you choose, your customer type (consumer or business) and, if you are a business customer, your company name and VAT number.
• Usage and billing data: credit balance, transactions, your billing address and (via Stripe) payment data. We do not store card details ourselves.
• Acceptance log of the terms: which version of the terms you accepted and when, your customer type and language, and the IP address and browser (user agent) you used. This is our evidence that the agreement was concluded (article 6:234 of the Dutch Civil Code).
• Withdrawal requests (only if you withdraw): your name, email address, the order date, your comment and the IP address from which you made the request.
• Confirmation emails: a log of the order confirmation and withdrawal emails we send you, including the text, as evidence that you received the statutory information.
• Documents you upload for pseudonymization: these are processed temporarily and not stored by us. If you use our API or MCP server, we keep the result (the pseudonymized text and the legend) for at most 24 hours so your integration can retrieve it. If you send more than 100,000 characters with “Send to LawBot Pro”, we keep the pseudonymized text of those documents (without the legend) for at most 24 hours as a package, so your agent can retrieve it.
• Connection data for LawBot Pro (only if you use that feature): your ChatGPT token (stored encrypted), the names and IDs of your linked agents, your prompt templates and the names of your dossiers.
• Connected apps (only if you connect the Pseudonymizer app in ChatGPT or a connector in Claude): which app you granted access, when, and the associated access tokens (stored only as an irreversible hash).
• Dispatch data: for each dispatch to LawBot Pro or Claude, the time, the agent, the dossier, the number of documents and characters, the credits, the status and the link to the chat in ChatGPT. We do not keep instructions or document content, with the sole exception of the temporary package for large dispatches (see above).

3. Documents and pseudonymization

When you upload a document, we use it at the moment of processing to detect personal data. We use artificial intelligence for this: a large language model proposes a legend with the data found and their pseudonyms. We reach that language model through OpenRouter, with processing in Europe and no data retention by the AI provider (see section 6). We first convert scanned documents into text using text recognition (OCR) by Mistral AI. The language model does not rewrite your text and does not make decisions about people.

Replacing data with pseudonyms follows fixed rules. In the web app this happens locally in your browser, and you can check the legend first. If you use our API or MCP server, the replacement works the same way in our server function in the EU. We do not use the content of your documents, legends and results to train or improve AI models, and we do not build profiles from them. Section 7 sets out how long we keep data.

4. Send to LawBot Pro (ChatGPT) and Send to Claude

With “Send to LawBot Pro” you send pseudonymized documents with an instruction to an agent in your own ChatGPT Business workspace, such as LawBot Pro. This only happens on your own instruction. You see in advance exactly which text will be sent, and you confirm it.

What goes to ChatGPT: the pseudonymized text of the documents you choose, your instruction and the file names, which are pseudonymized as well. Your legend, which links pseudonyms to real data, is never sent. Before each dispatch we automatically check whether any data from your legend is still in the text; if so, we show you which data, and you send only after an explicit confirmation. You decide what is sent.

Please note: for you, pseudonymized data remains personal data, because you can re-identify it with the legend. Whether it is also personal data for a recipient who does not have the legend depends, according to the Court of Justice of the EU (judgment of 4 September 2025, case C-413/23 P), on the means that recipient can reasonably use to re-identify it. That is why the legend is never sent and why we check before each dispatch whether any data from your legend is still in the text. We nevertheless advise you to treat the text you send as personal data and to align your arrangements accordingly.

OpenAI receives this data as a service provider of your ChatGPT workspace, under the agreement and settings that you or your organization have with OpenAI. For these dispatches, OpenAI is not a sub-processor of ours. Make sure your use of ChatGPT fits your own arrangements, such as a data processing agreement with OpenAI and your agreements with clients. Where and for how long OpenAI processes the data depends on the settings and terms of your workspace.

The agent’s answer appears in ChatGPT, in a chat of the account the linked token belongs to. We do not receive or keep that answer.

We only use your ChatGPT token to start the dispatches and test messages you initiate, and to check their status. The token is stored encrypted and never shown again after you enter it. You can delete it at any time on your account page, and revoke it in ChatGPT.

Send to Claude: here we send nothing ourselves. We keep the pseudonymized documents as a package for at most 24 hours and give you an instruction text that you paste into Claude yourself; Claude then retrieves the package through the Pseudonymizer connector you connected in Claude. Anthropic is then the recipient you choose yourself, under your agreement with Anthropic. The same credits apply as for a dispatch to LawBot Pro.

Integrations via our API or MCP server can use these features too, likewise only after a confirmed preview.

5. Legal bases

For the data for which we are the controller (account, billing, usage and connection data), the legal bases are performance of the contract (delivering the service), our legal obligations (such as retaining billing records and the information duties for distance contracts), our legitimate interest (keeping the service safe and usable, and keeping evidence of the conclusion, confirmation and withdrawal of the agreement) and, where applicable, your consent.

For the content of your documents, the legends, the results, the packages and the dispatches, the legal basis is yours as the controller: you decide why you have a document processed and you warrant that you may do so. If your use as a consumer falls under the household exemption (section 1), you do not need a legal basis. In both cases we process those data only on your instructions, within the contract, and not for our own purposes.

We carry out dispatches to LawBot Pro or Claude only on your explicit instruction for each dispatch.

6. Processors and sub-processors

We use carefully selected providers, with processing within the EU or with appropriate safeguards. For document processing these are our sub-processors; the list matches Annex 3 of the data processing agreement (as at 28 October 2026):

• Supabase (Supabase Pte. Ltd, Singapore): authentication, database and server functions. Storage and primary processing in the EU region of our project; EU standard contractual clauses in Supabase’s data processing addendum.
• Vercel: hosting of the web application in the EU region Frankfurt, with a global network for static files. EU-US Data Privacy Framework (certified) and EU standard contractual clauses.
• OpenRouter and the AI provider selected via OpenRouter: detection of personal data. Prompts and completions are processed in Europe via eu.openrouter.ai (EU/EEA, United Kingdom or Switzerland), with zero data retention at the AI provider; our account and operational data at OpenRouter (authentication, security logs) may be processed in the US. Adequacy decisions (UK, Switzerland) and EU standard contractual clauses.
• Mistral AI: OCR for scanned documents (EU, France).
• Cloudflare (Turnstile): protection against automated abuse at sign-up and sign-in. This processes technical browser and device data, never document content (EU/global; EU standard contractual clauses and EU-US Data Privacy Framework).
• For the data for which we are the controller ourselves, we also use: Stripe for payments and, via Stripe Tax, the VAT calculation based on your address and VAT number (Stripe processes only account and payment data, no documents); and Brevo (France, processing within the EU) to send all emails of the service: sign-in and sign-up emails, order confirmations, reminders and acknowledgements of withdrawals. For this, Brevo processes your email address and the content of the email, never documents.
• OpenAI (ChatGPT) is not a sub-processor of ours. If you use “Send to LawBot Pro”, OpenAI is the recipient you choose yourself (see section 4). The same applies to Anthropic (Claude) with “Send to Claude”.

7. Retention periods

We keep account data for as long as you have an account. We do not retain uploaded documents. Results of API and MCP requests (pseudonymized text and legend) are deleted automatically after at most 24 hours. We keep your ChatGPT token, agents, prompt templates and dossiers until you delete them or close your account. Dispatch data for “Send to LawBot Pro” is deleted automatically after 30 days; packages for large dispatches after at most 24 hours. We keep a connected app’s access until you revoke it on your account page or until the connection has not been used for 30 days. We keep transaction and billing data for as long as legally required. For documents, results, packages and dispatch data, the same periods apply as in article 5.3 of the data processing agreement.

We keep the acceptance log of the terms, withdrawal requests and the log of confirmation emails for up to five years after the end of the agreement, in line with the limitation period. If you delete your account, those logs remain without a link to your account, but with your email address, because otherwise they would no longer serve as evidence.

8. Your rights

You have the right to access, rectification, erasure, restriction, objection and portability of your data. Send your request to privacy@litic.ai.

If you appear in a document that one of our customers has had processed, that customer is the controller for those data. Please address your request to that customer. If we receive such a request, we refer you to that customer and help the customer with the handling where needed. If a consumer used the service only for themselves or their household, the GDPR does not apply to that consumer; you can then contact us and we will tell you whether we still hold any data about you. Because we do not keep document content, or keep it for at most 24 hours (section 7), this is usually not the case.

If you disagree with how we handle your data, you can file a complaint with the Belgian Data Protection Authority (gegevensbeschermingsautoriteit.be), our lead supervisory authority because we are established in Belgium. You can also contact the supervisory authority of the country where you live or work, such as the Autoriteit Persoonsgegevens in the Netherlands.

9. Security

Processing takes place within the EU, connections are encrypted and sign-in is passwordless (magic link, passkeys or an external provider). ChatGPT tokens are stored encrypted (AES-256) and can only be decrypted by our servers. No service is entirely risk-free; report suspected abuse to privacy@litic.ai.

Where our infrastructure is located, which jurisdiction it falls under and which measures we take against access by governments outside the EU is explained at pseudonymizer.eu/infrastructure-and-government-access (article 28 of the Data Act).

10. Cookies and local storage

We only use functional storage needed to keep you signed in and remember your language preference. We do not use advertising or tracking cookies.

11. Changes

We may update this policy. The most recent version is always on this page. We will inform you where possible of significant changes.