ChatGPT is available again in Italy after OpenAI met regulator Garante's demands, including a form for EU users to object to having their data used for training
What changes did ChatGPT make? — 👶 Verify users' ages upon signup … Rupert Breheny : This is an odd one because it sounds like their approach to complying with GDPR is to put an onerous roadblock in front of users that want to opt out of having their data enter the training models. … Tweets: Sam Altman / @sama : we're excited chatgpt is available in 🇮🇹 again! Asela Waidyalankara / @aselawaid : Access to the ChatGPT chatbot has been restored in Italy. It was banned by the Italian data-protection authority at the start of April over privacy concerns. https://www.bbc.com/... Donald Callahan / @dcallahan2 : “Among the changes, OpenAI linked The Verge to a new form that EU users can submit to remove personal data under Europe's #GDPR” I find this absurd. How can @OpenAI remove data that was used in the model training? https://www.theverge.com/... https://twitter.com/... Neil Turkewitz / @neilturkewitz : OpenAI said “it would provide a new form for EU users to exercise their right to object to use of personal data to train its models.” How is that going to work? Preemptive opt-out for any personal data? Post-training removal? Of what? This is delicious. https://www.bbc.com/... Peter Welinder / @npew : Welcome back, Italy! 🎉 https://twitter.com/... @martinsfp : This is very interesting re: GDPR compliance for generative A.I. https://twitter.com/...
Context & Ripple Effects
Italy's regulator had first suspended the service over the asserted lack of a legal basis for collecting and storing personal data, then set explicit conditions around data-processing disclosures and age gating in its compliance roadmap for restoring access.
OpenAI had already introduced controls to turn off chat history and exclude those conversations from model training; the new EU objection mechanism extends that shift from a product setting into a regulator-driven privacy process.
First-order effects
- ChatGPT can resume serving users in Italy after OpenAI implemented age verification and information measures required by Garante.
- EU users gain a formal channel to object to use of their data for model training, though the coverage leaves unresolved how practical that process is and how it applies to data already incorporated into models.
Second-order effects
- OpenAI must operate a more differentiated privacy and access workflow for European users, rather than treating ChatGPT's data practices as uniform across markets.
- The restoration shows that a service suspension can be converted into negotiated operational conditions, giving other AI providers a concrete example of the controls regulators may seek before allowing access.
Third-order effects
- If regulators continue to attach access to data-use controls, frontier-model distribution will increasingly depend on jurisdiction-specific governance features, not only model capability.
- The unresolved treatment of historical training data points to a broader tension: user objection rights are easier to implement for future collection than for information already used in model development.
The trend: Generative-AI access is moving toward conditional market entry, with privacy, age assurance, and data-governance controls becoming part of the product itself.