X agrees to not use some EU user data to train AI chatbot

DUBLIN (Reuters) - Social media platform X agreed on Thursday not to train its AI systems for now using the personal data collected from European Union users before they had the option to withdraw their consent, an Irish court heard on Thursday.

Ireland's Data Protection Commission, the lead EU regulator for most of the top U.S. internet firms due to the location of their EU operations in the country, this week sought an order to suspend or restrict X from processing the data of users for the purposes of developing, training or refining its AI systems.

Elon Musk-owned X has said it allows all users to decide if their public posts can be used by the platform's artificial intelligence chatbot, Grok. To do so users have to untick a box in their privacy settings to opt out.

However Judge Leonie Reynolds said it was clear that X began processing EU users' data to train its AI systems on May 7 and only offered the option to opt out from July 16. The feature was also not initially rolled out to all users, she said.

A lawyer for the platform formerly known as Twitter said the data collected from EU users between May 7 and Aug 1 would not be used until proceedings on the Irish Data Protection Commission's (DPC) order are decided by the court.

Lawyers for X are due to file opposition papers against the suspension order by Sept 4, the court heard.

On a post on the social media platform on Wednesday, the X Global Government Affairs account said the order sought by the regulator was "unwarranted, overboard and singles out X without any justification."

The regulator's concerns over how X uses the data follows Meta Platforms' decision in June not to launch its Meta AI models in Europe for the time being after the Irish DPC told it to delay its plan.

Alphabet's Google also agreed to delay and make changes to its Gemini AI chatbot earlier this year following consultations with the Irish regulator.

(Reporting by Padraic Halpin; Editing by David Holmes)