Italy has serious doubts about the legality of ChatGPT and the way OpenAI trained the model. Because everything indicates that the GDPR was not previously complied with, the country wants to block the AI ​​chatbot.

OpenAI has trained ChatGPT on data from the internet. The AI ​​is therefore able to provide a lot of factual information on certain facts, even if they relate to people. This implies that the training data contained data about Europeans and this is not allowed. Finally, the GDPR prohibits the processing of personal data by European citizens without their consent, especially when that processing takes place outside the EU.

Improper processing

Since ChatGPT has a lot to say about Europeans, the Italian data protection authority concludes that OpenAI has done away with this GDPR legislation. In addition, ChatGPT does not shy away from filling knowledge gaps with fabricated data that may sound realistic, but are not based on anything.

For example, ChatGPT knows what the signer is writing about, but the AI ​​isn’t afraid to include the posts of its competitor peers in the short bio. This brings us to a second problem: the GDPR gives Europeans the right to rectify mistakes in data held about them. This is currently not possible with ChatGPT.

All of this suggests that OpenAI isn’t quite right legally in the EU. Italy therefore wants to ban ChatGPT in the country and gives OpenAI twenty days to respond.

GDPR vs. AI

The decision is far-reaching and complex. On the one hand, it sets an interesting precedent. The EU has rules about data and everyone has to follow them. The end does not justify the means: breakthroughs in AI are no excuse for being careless with European data.

On the other hand, the question arises to what extent the GDPR applies or should apply to AI and machine learning training. Too strict an interpretation can completely limit AI research in Europe. At least the fence is off the dam now. These and other legal challenges will drive the development of AI.