GDPR disregarded?

AI training with user data: Complaint against Meta

The data protection organization noyb, run by Austrian Max Schrems, has filed complaints in ten EU countries and Norway against the US company Meta, the parent company of Facebook and Instagram, among others, regarding the development of artificial intelligence (AI). The accusation: Meta uses user data to train AI models without asking users for permission beforehand. This is a violation of the EU General Data Protection Regulation (GDPR).

According to a press release issued on Thursday, noyb (none of your business) is calling on the responsible data protection authorities in the eleven countries to initiate urgent proceedings to put an immediate end to this practice. The eleven countries are: Austria, Belgium, France, Germany, Greece, Italy, Ireland, the Netherlands, Norway, Poland and Spain.

Unspecific information emails sent out
Meta is currently sending out emails to users of its social media. In it, the company informs users about a relatively unspecific adjustment to its privacy policy: "We are updating our privacy policy because we are expanding AI at Meta." Users can object to this.

This also affects non-public, personal chat messages. We are talking about "posts, photos and their captions, and messages you send to an AI". "In order to offer you these user experiences, we will in future rely on the legal basis of legitimate interest when we use your information to further develop and improve AI at Meta."

Right to object insufficient for noyb
In noyb's view, however, this right to object is not sufficient. Meta must explicitly ask for consent: "Meta's privacy policy would theoretically allow any purpose of use. All of this is extremely worrying because it involves the personal data of around 4 billion people."

There is fierce competition between tech giants such as Meta, Google and Microsoft when it comes to the development of AI systems. Meta is aiming to develop its own "Meta AI", which will generate texts and images. In order to teach the system its own language skills, the software is trained with many texts from human users. It works in a similar way with photos and images.