In December 2024 the Italian Garante fined OpenAI €15 million over ChatGPT — the headline generative-AI penalty in Europe. In March 2026 a court erased it.
What the court did
On 20 March 2026 the Tribunale di Roma — the Rome civil court that hears challenges to Garante sanctions — annulled the €15 million fine in full (ANSA). The court had already suspended the penalty in March 2025 pending a ruling on the merits.
The Garante’s original decision — announced on 20 December 2024 — found that OpenAI had trained ChatGPT on personal data without an adequate legal basis, breached its transparency duties, and ran no age-verification capable of keeping out under-13s.
The full case history — the exact GDPR articles, the six-month information campaign OpenAI was ordered to run, and judgment no. 4153/2026 — is in our registry: Garante AI fines.
Why it matters
This is the first major European generative-AI fine to be overturned at first instance. It underlines a pattern this hub tracks across GDPR enforcement: a headline number is not a final, collected penalty — the €746 million Amazon fine in Luxembourg fell the same way. Most early AI enforcement still runs on GDPR rails, and those rails are now being tested in court.
Related: GDPR pillar · Garante profile · Garante AI fines registry