Update, 12/08, 4:39 PM: Privacy organization noyb has filed complaints about Grok’s AI training. A total of nine complaints were sent to various mostly European countries, including the Netherlands, Belgium, Austria, France, Greece, Italy, Spain and Poland.
According to the organization, extra attention should be paid to the matter because the Irish Data Protection Commission (DPC) only appears to be concerned about mitigation measures and the fact that X allowed the data collection to continue despite a warning from the DPC. Because the trial took place behind closed doors, noyb can only deduce this and not confirm it. Max Schrems, chairman of noyb, explains what the organization wants to achieve: “Companies that interact directly with users simply need to show them a yes/no prompt before using their data.”
Update, 9/08, 10:35 am: X will temporarily stop feeding data from European users to Grok. The chatbot has been training on the data of all X users since its launch in 2023. The social media platform was recently taken to court because this allegedly violated the GDPR. In court, X reportedly agreed not to use the data collected between May 7 and August 1, writes Reuters.
Original, 8/08, 12:11 pm: X has been taken to task for training Grok based on posts shared on the social media platform. This is because users did not give explicit permission to do so. That is against GDPR rules, according to the Irish regulator.
The Irish Data Protection Commission (DPC) says X does not respect the data privacy of users of the social media platform. This reports Irish broadcaster RTE. The platform is in violation of GDPR due to Grok’s training method. This AI model trains itself with content from X. Users do not give permission for it, which is, however, necessary in the EU.
Adjustments to opt-in needed
Here, the DPC is cracking down on Grok’s training method, for which X does ask permission. X users have been able to opt out of using their own data in training since July. The DPC finds the move insufficient because X users approve the training by default. The regulator reported this problem to the social media platform but received no response.
A GDPR violation can result in a fine of up to four percent of annual revenue. The DPC will transfer the case to the European Data Protection Board (EDPB).
Second version
Grok has been around since November 2023 and is a product of xAI. At launch, Elon Musk, the founder of xAI, stated that the AI chatbot would have real-time access to all of X’s information, which Musk said would be “a huge advantage. Under European laws, the net could become particularly disadvantageous.
A second version of Grok is currently undergoing training and requires data. Grok-1.5 is however the first successor to the chatbot. That brought improvements in performance on coding and math tasks.
Also read: AI chatbot ‘Grok’ from co-founder OpenAI launched: ‘sarcastic and based’