Is ChatGPT GDPR Compliant? What Businesses Must Know

"Is ChatGPT GDPR compliant?" has no yes-or-no answer. The GDPR (General Data Protection Regulation) does not certify a tool. It governs a specific processing activity and the organisation that decides it. When a business pastes EU personal data into ChatGPT, the business becomes the data controller. These duties are yours: a lawful basis (Article 6), data minimisation (Article 5), and a data processing agreement (Article 28) with the provider. The default consumer app does not ship that signed agreement. The good news: anonymising personal data before the prompt sharply cuts the risk.

"GDPR compliant" is not a badge

The GDPR does not work like a label stuck on software. Compliance attaches to a given processing activity. It also attaches to the organisation that decides why and how personal data is used.

Here is the key point for a business. You paste a client's name, email or file into ChatGPT. At that moment, you are the data controller. The tool you use does not absorb your duties. They stay yours.

The data controller's duties

A data controller must satisfy several GDPR obligations at once. They are not optional. Here are the three that weigh most in an AI prompt.

• A valid lawful basis (Article 6): consent, contract or legitimate interests, for example. Without one, the processing is unlawful.
• Data minimisation (Article 5(1)(c)): process only data that is adequate, relevant and limited to what is necessary.
• A data processing agreement (Article 28): whenever another company processes that data on your behalf, a written contract is required. It is often called a DPA (Data Processing Agreement).

The GDPR also demands transparency toward the people concerned. They must know their data is being processed. They keep their rights: access, rectification, objection, erasure. These principles do not vanish because an AI is involved.

The consumer app and the missing DPA

The default consumer ChatGPT app is not built for this. It does not provide a signed Article 28 processing agreement for that use. Yet that contract is required whenever a third party processes other people's personal data on your behalf.

OpenAI offers separate business and enterprise terms. They include a DPA and data controls, such as a setting to exclude your inputs from model training. But that is a distinct, opt-in arrangement. It is not the default consumer experience.

France's regulator, the CNIL, confirms the principle for any AI. Anyone using personal data to build or deploy an AI system must fix an explicit, legitimate purpose from the outset. They must inform the people whose data is used. They must respect those people's rights. And they must process only the data that is necessary.

The Garante case: what enforcement really shows

The risk is not theoretical. Italy's authority, the Garante, fined OpenAI 15 million euros in December 2024. The grounds: the lawful basis for training data, transparency toward users, and the absence of adequate age verification.

Be precise about what came next. That fine was later annulled. On 18 March 2026, the Court of Rome (judgment No. 4153/2026) set the penalty aside. But it did so on jurisdictional grounds, not on the merits. In plain terms: Ireland's Data Protection Commission had become OpenAI's lead supervisory authority. So the Garante lacked competence to issue the final decision.

What the court did not say matters just as much. It did not clear OpenAI of the underlying GDPR concerns. The substantive questions on lawful basis and transparency were not decided. They remain open. The annulment was procedural, nothing more.

The fix: minimise and anonymise before the prompt

The most practical way to shrink your GDPR exposure is simple. Minimise and anonymise personal data before it ever enters the prompt. Strip out names, contact details and IDs before the text reaches an external model.

The result is direct. Far less personal data is processed by a third party in the first place. That serves the Article 5 minimisation principle exactly. You get the AI's help. You keep the sensitive material out of reach.

Here is the method, step by step.

1. 1Spot the personal data in your text: names, emails, phone numbers, client IDs.
2. 2Replace each value with a reversible token, in the browser.
3. 3Send only the anonymised text to the AI model.
4. 4Restore the real values in the reply, locally.

That's what ONYRI Sanitize is for. The engine spots sensitive data — names, emails, IDs, amounts — and replaces it with reversible tokens before sending. Detection and the mapping stay in your browser. Only anonymised text reaches the model. The AI finds only tokens, never your real personal data. You serve the GDPR minimisation principle, without giving up the AI's help.