Incident to report to a regulator?Incident to report? Intake Line

EDPB Draft Guidelines Set GDPR Tests for Web Scraping to Train Generative AI

EDPBEuropean Data Protection Board draft Guidelines 03/2026 say private entities scraping the web to train generative AIArtificial Intelligence should filter data, skip sites that clearly oppose scraping and publish notice details. Feedback closed on 30 October 2026, so the text is not final.

8 July 2026 · Consultation

What happened

The European Data Protection BoardAn independent EU body made up of the heads of the national data protection authorities of the EEA countries and the European Data Protection Supervisor. It ensures the GDPR is applied consistently and coordinates cooperation between authorities. adopted draft Guidelines 03/2026 on 7 July 2026 for public consultation. Feedback ran from 8 July to 30 October 2026 and has closed. The consultation page offers version 1 and the final text is still to come, so these are not yet final guidelines.

What it changes

The draft covers private entities that scrape internet sources to train generative AIArtificial Intelligence, directly or through a contractor. It says roles should be analysed case by case. Legitimate interest under Article 6(1)(f) needs three cumulative conditions to be met.

On data minimisation, the draft expects precise collection criteria, data mapping and filters. It says to exclude certain categories of sites, such as those used mainly by minors, and sites that clearly oppose scraping through measures such as robots.txt, ai.txt or CAPTCHA. A robots.txt file not applying to a site does not amount to GDPRGeneral Data Protection Regulation consentA freely given, specific, informed and unambiguous indication of a person's wishes, by a statement or clear affirmative action, agreeing to the processing of their personal data. The ePrivacy Directive, the DMA and the DGA rely on the same GDPR standard.. Where informing people individually is impossible or would take disproportionate effort, Article 14(5)(b) still requires you to make the information public, for example in a notice on your website.

Who is affected

Private entities that scrape web data, or have it scraped, to train generative AIArtificial Intelligence, and the suppliers that provide those datasets.

What to do now

  • Record who is controllerThe person, public authority, agency or other body that, alone or jointly with others, decides the purposes and means of processing personal data. The controller carries primary responsibility for GDPR compliance and must be able to demonstrate it., joint controller or processorA person, public authority, agency or other body that processes personal data on behalf of a controller. The processing must be governed by a contract or another binding legal act under Union or Member State law meeting GDPR Article 28(3), and the processor must follow the applicable documented-instruction requirements. at each step of your training data pipeline.
  • Write down your legitimate interest assessment, or revisit the one you have.
  • Add collection filters and honour robots.txt, ai.txt and CAPTCHA signals.
  • Publish the Article 14 information, including the sources you use.
  • Ask dataset suppliers how they meet the same points, and recheck when the final text appears.

Get updates for the EU

One email a week when something changes. Add other countries if you need them.

Countries
European Union

Pick one or more

One email a week, only when something changed. Each email has an unsubscribe link. Privacy notice

RSS Feed for EU

Loading the form…