A €158,000 GDPR fine: compliance is not optional

Character.AI learns that offering a service in Italian means playing by EU rules

On 3 July 2026, Italy’s data protection authority (Garante) fined Character Technologies Inc. €158,000, following an ex officio investigation opened in November 2024. The investigation concerned the availability of Character.AI to users in Italy, including through an Italian-language interface.

The company operates Character.AI, a generative AI platform where users create and interact with virtual characters through chat.


Why was the company fined?

Italy’s data protection authority identified a series of GDPR violations. The fine was set at €158,000 after Garante took into account mitigating circumstances, including the company’s cooperation, the absence of previous infringements, the limited duration of the infringements and the lack of intent.

  1. Inadequate privacy notices (Articles 12–14 GDPR)

Garante reviewed two versions of Character.AI’s privacy notice, dated October 2023 and August 2025, and found both fell short of the transparency requirements set out in Articles 12 to 14 of the GDPR.

Neither notice clearly explained the purposes for which user data was processed, the legal basis relied on for that processing, how long the data would be retained, or the safeguards applied when data was transferred outside the EU.

For a platform whose core function is ongoing, data-intensive conversation with an AI system, that lack of clarity matters more than it might for a simpler service, users had little way of understanding what happened to what they typed, or for how long.

  1. Late DPIA and EU representative (Articles 35(1), 27(1) and 27(4) GDPR)

Under the GDPR, any processing likely to result in high risk to individuals requires a Data Protection Impact Assessment (DPIA) before the processing begins, not after, and processing children’s data through an AI chat interface plausibly qualifies.

Character.AI conducted its DPIA late, only after the platform had already been operating and growing its user base across the EU. Garante found a breach of Article 35(1) GDPR, which requires a DPIA to be carried out before processing where it is likely to result in a high risk to individuals. The company also appointed an EU representative only on 31 May 2025. Character.AI relied on the exception in Article 27(2) GDPR, arguing that its processing of EEA users’ data had been only occasional and that the EEA was not a priority market. Garante disagreed: the service had been offered in Italian since 8 April 2024 and the privacy policy expressly addressed EEA residents, indicating that the company was targeting people in the Union within the meaning of Article 3(2)(a) GDPR. Garante also found that the representative’s contact link did not initially work, leading to a finding under Article 27(4).

  1. Insufficient age verification (Articles 24(1) and 25(2) GDPR)

Arguably the most consequential finding in Garante’s decision was the inadequacy of Character.AI’s age-verification safeguards. Between 8 April 2024 and 8 April 2025, users under 16 could register, including after the introduction of an age gate that did not function correctly, and profiles were public by default. The 16-year threshold had been chosen by Character.AI itself; in Italy, the relevant age under Article 8 GDPR is 14, as reflected in Article 2-quinquies of the Italian Privacy Code. Garante therefore found that the measures did not adequately implement data protection by design and by default under Articles 24(1) and 25(2) GDPR. The platform’s immersive conversational format and its capacity for long, personal exchanges with AI-generated characters are relevant to the seriousness of the risk, but this is an assessment rather than a statement that the format was expressly designed by Garante to keep users engaged.


The territorial scope question

Character.AI argued that, until mid-2025, its processing of EEA users’ data was only occasional and the EEA was not a priority market, so it considered itself exempt from appointing an EU representative under Article 27(2) GDPR. Garante disagreed. Since 8 April 2024 the service had been offered in Italian, and its privacy policy expressly addressed EEA residents. These were relevant indications of targeting under Article 3(2)(a) GDPR. Mere accessibility of a service in the EU would not, by itself, have been enough. Because Character.AI had no EU establishment, the one-stop-shop mechanism did not apply, leaving Garante competent to act on its own.

The territorial-scope test therefore turns on whether the service targets people in the Union, rather than simply whether EU users can access it. Here, Garante relied on concrete indicators of targeting, including the Italian-language interface and the privacy policy addressing EEA residents.


What does compliance look like in practice? Garante ordered the measures below, with a 120-day deadline for implementation and reporting:

  • An age gate that effectively blocks under-16 users in the EEA, rather than a simple self-declared age at sign-up;
  • Cooling-off periods preventing blocked minors from immediately re-registering with a different age;
  • Private profiles set by default for accounts identified as belonging to minors;
  • Timely appointment of an EU representative, before entering the European market;
  • Updated privacy notices addressing storage periods and international transfers;
  • Verification and review of the data used for model training.
 

What should companies take away from this decision?

  • An EU audience, not an EU address, is what triggers GDPR obligations where a service targets EU users; the company’s headquarters is not decisive.
  • Age verification for AI products used by minors is now being assessed with the same rigour as data security or consent management, not treated as a secondary feature. The decision also sits in a broader Italian enforcement context: Garante previously sanctioned Luka Inc. (Replika) €5 million over, among other issues, inadequate age verification, while Article 4 of Italy’s Law No. 132/2025 on artificial intelligence provides for parental consent for children under 14 accessing AI technologies.
  • Transparency, timely risk assessments, and local representation are being judged as a single compliance system, not as separate, independent boxes to tick.

For a sector still moving faster than the rules meant to govern it, this fine is a reminder that some obligations were never optional to begin with – and regulators are increasingly willing to prove it.


Sources

age verification failures”

Failures”

Join us

Sign up for our newsletter and take the first step in joining our community! Stay updated with the latest news, events, and educational resources from the hub.

By clicking “Send,” you consent to the processing of your email address for the purpose of sending you news and updates about Legal Accelerators events. We use MailerLite as our email marketing service provider. MailerLite is an EU-based company that operates under the General Data Protection Regulation (GDPR). For further details, please see MailerLite’s Privacy Policy.You have the right to withdraw your consent at any time and we will act immediately, unless there is a legal reason or legitimate interest in not doing so. For withdrawal of consent, please contact us at contact@legalaccelerators.com.