An AI-built app must meet the same GDPR obligations as any other software: a lawful basis for each processing purpose, data minimisation, processor agreements with every service involved, transparent privacy information, security measures and support for data subject rights. The specific risks of AI-built apps are unknown data flows created by generators, personal data sent to AI APIs and hosting outside the EU, all of which you can address with a data map, processor review and anonymization before AI processing.
On this page
The GDPR does not care how your code was written. An app generated in an afternoon with Lovable or built over weeks with Claude Code carries the same obligations once it processes personal data of people in the EU. What changes is the risk profile: generators wire up analytics, authentication providers, email services and AI APIs without anyone writing down where data goes. This guide covers the GDPR topics that matter most for AI-built apps and a practical way to get them under control. It is not legal advice; involve your data protection officer or counsel for final decisions.
Start with a data map, not a privacy policy
Many teams start by generating a privacy policy. That reverses the order. First establish which personal data the app processes, for which purpose, where it is stored and which third parties receive it. Agentic tools help: ask Claude Code or Codex to list every external service called from the codebase, every table and column that holds personal data, every analytics or tracking script and every environment variable that points to a third-party API. Then verify the list manually, because SDKs can send data without explicit calls in your code.
Pay attention to data you did not intend to collect. Error tracking tools capture request bodies and user identifiers, logs store IP addresses and email addresses, and AI chat features keep full conversation histories. Each of these is personal data with its own retention question. Decide how long each category is kept and implement deletion jobs, rather than leaving it to defaults.
The result feeds your record of processing activities under Article 30 and tells you which agreements and disclosures you need.
Processors, AI APIs and international transfers
Every service that processes personal data on your behalf is a processor and needs a data processing agreement (DPA). In a typical AI-built app that includes the hosting provider, the database and auth provider, the email service, error tracking, analytics and any AI model API. Most large providers offer standard DPAs; make sure they are accepted for your account.
AI APIs deserve particular attention. Check whether the provider uses API data for training (many business API terms state that it does not by default, but verify for your plan), how long prompts are retained, and where processing happens. If data leaves the EU, you need a transfer mechanism such as the EU-US Data Privacy Framework for certified US companies or standard contractual clauses. Our article on hosting AI-built apps in Europe covers data residency options.
Minimise what reaches AI models
The most effective GDPR measure for AI features is to not send personal data at all. Many use cases, such as summarising support tickets, classifying documents or drafting replies, work just as well when names, addresses, phone numbers and IDs are replaced with placeholders. Textwash Pro anonymizes text automatically using local machine learning, so personal data is removed before anything is sent to an external model. This supports data minimisation and privacy by design under Article 25 and simplifies the transfer question considerably.
Consent, transparency and user rights
AI builders often add analytics or third-party fonts and embeds that set cookies or transmit IP addresses before consent. Under the ePrivacy rules as implemented in member states, non-essential cookies and similar technologies require prior consent. Audit what loads before the user interacts with your banner. Self-host fonts and load embeds only after consent.
Your privacy notice must describe the actual processing, including AI features and the providers involved. The app must support data subject rights in practice: export of a user's data, deletion of an account including backups within a reasonable period, and correction of data. Generated apps rarely implement account deletion properly; it often deletes the login but leaves profile rows, uploads and logs behind.
A GDPR checklist for AI-built apps
- Create a data map of personal data, purposes, storage locations and recipients.
- Define a lawful basis for each purpose and document it in your record of processing activities.
- Sign DPAs with every processor and document international transfer mechanisms.
- Configure AI APIs for no training use where available and anonymize text before sending it.
- Block non-essential cookies and third-party requests until consent is given.
- Implement working data export, deletion and correction, including files and logs.
- Secure the data: access control, encryption in transit, backups and breach notification procedures.
- Write the privacy notice last, based on the verified data map.
Security is part of GDPR compliance under Article 32. Misconfigured database permissions are the most common cause of leaks in AI-built apps, so read our guide on Supabase Row Level Security and the security audit checklist.
Organisational support
For enterprises and public bodies, compliance extends to governance: which teams may use which AI tools, with what data and under which audit trail. KalaidoCreate provides an AI platform with compliance features and governance for exactly this setting. If you need the implementation done, the OnlineAppCreator custom development team builds and hardens apps with privacy by design, from our base in Freiburg, Germany, under ISO 9001 certified processes.
Frequently asked questions
Does GDPR apply to an app I built with an AI tool?
Yes. If the app processes personal data of people in the EU, GDPR applies regardless of how the code was produced.
Can I send user data to OpenAI or Anthropic APIs under GDPR?
It can be possible with a DPA, a lawful basis, a valid transfer mechanism and appropriate settings, but the safest approach is to anonymize or remove personal data before it reaches the model.
Do I need a cookie banner for my AI-built web app?
You need consent for non-essential cookies and similar tracking. If the app only uses strictly necessary cookies, such as a login session, a banner may not be required; check what your generator added.
Who is responsible if the AI builder platform leaks data?
You remain the controller for your users' data. The platform is typically your processor, so you need a DPA with it and must have chosen it with appropriate care.