X agrees to halt AI training on EU users’ data after DPC investigation ends

The European Data Protection Commission has agreed to end its investigation into X after the social media platform agreed to stop training its Grok AI chatbot on users’ data. In a statement released on September 4, the DPC said the Court struck out the case against X based on this agreement.

The regulator filed an application before the Irish High Court on August 8, 2024, alleging that X violated the Data Protection Act by training its chatbot on users’ personal data without their consent. X temporarily suspended the practice following the application but has now agreed to abide by the decision permanently.

X to delete EU users’ data from its AI training model

Per the terms of the agreement, X will permanently erase data of EU users collected between May 7 and August 1 from its AI model and will not collect additional data for training the Grok chatbot. The DPC chairperson, Des Hogans, described this decision as a welcome development, noting that it will protect the rights of EU/EAA citizens.

He said:

“The DPC welcomes today’s outcome which protects the rights of EU/EEA citizens. This action further demonstrates the DPC’s commitment to taking appropriate action where necessary, in conjunction with its European peer regulators. We are grateful for the Court’s consideration of the matter.”

Interestingly, X’s parent company, Twitter International, initially denounced the allegations, claiming that its practices meet all EU General Data Protection Regulations provisions. The company called the orders “draconian,” claiming they would affect how the chatbot functions in the region.

Meanwhile, the DPC has further requested that the European Data Protection Board (EDPB) give its opinion on Article 64 (2)GDPR, particularly on how this provision affects the use of personal data for training AI models.

XAI supercomputer begins operations; Grok set for big upgrade

Although X might have bowed to regulatory pressures surrounding the training of Grok in Europe, the chatbot could be on its way to becoming more powerful following the launch of a new supercomputer to train the AI model. Elon Musk’s AI startup, xAI, has now completed the first phase of its supercomputer facility, which will train AI models.

The first phase was completed within four months, and the Colossus supercomputer is now online and active. Musk posted about it on X, noting that the computer comprises 100,000 Nvidia H100 GPUs and was built in just 122 days. The company plans to add 50,000 Nvidia H200 chips to the cluster, doubling the supercomputer’s capacity.

With hundreds of thousands of the most powerful GPUs now available for training Grok, the chatbot is expected to see accelerated progress, which would be massive given it is already competing with the best OpenAI and Anthropic models.

Some users on X note that Grok could reach Artificial General Intelligence (AGI) by August. In his response, Musk did not confirm whether that would happen but noted that Grok 3 will be “something special” when it arrives at the end of the year.