X’s new AI training opt-out setting draws regulatory scrutiny – SiliconANGLE

The Irish Data Protection Commission, the regulator that oversees X Corp.’s business practices in the European Union, has sent the company questions over a newly added privacy setting.

Users of the Elon Musk-owned social network noticed the setting in question earlier today. When it’s enabled for an account, that account’s information can be incorporated into the training dataset of Grok, a large language model series developed by another Musk-run company. The setting is turned on by default.

The data collection covered by the new setting can currently only be disabled on the web version of X. A similar opt-out option is expected to arrive on the social network’s mobile apps “soon.” It’s unclear whether X began gathering users’ information for AI training purposes before it rolled out the ability to turn off data collection.

“All X users have the ability to control whether their public posts can be used to train Grok, the AI search assistant,” X said in a statement. “This option is in addition to your existing controls over whether your interactions, inputs, and results related to Grok can be utilised.”

The Irish Data Protection Commission has sent X questions “seeking clarity” about the new setting, the Financial Times reported today. “It took us by surprise that they were rolling this out,” a spokesperson for the regulator told the paper.

The Grok series of LLMs for which X has started collecting user data is being developed by xAI, an artificial intelligence venture that Musk launched last year. The company raised $6 billion in funding at a $24 billion valuation this past May. It’s using a portion of the capital to build an AI training supercomputer that will feature 100,000 graphics processing units when it becomes fully operational.

The most advanced LLM in the Grok series, Grok 1.5V, made its debut in April. The model can process not only text input but also images. Grok 1.5V achieved a score of 53.6% on MMMLU, a popular benchmark for comparing multimodal LLMs, which puts it about five percentage points behind Google LLC’s Gemini Pro 1.5.

X’s AI development push is not the only aspect of its business  that has drawn regulatory scrutiny. According to Politico, the Irish Data Protection Commission is investigating the company in connection with “at least five cases” of potential regulatory compliance violations. Each case could potentially lead to fines equal to up to 4% of X’s worldwide annual revenue.

Separately, the European Commission is investigating the company over its compliance with the EU’s DSA tech regulation. Last month, officials determined that X’s blue checkmarks are deceptive and run afoul of the DSA. The European Commission is also evaluating whether X is meeting its regulatory obligations to remove illegal user-generated content and provide researchers with access to data about its platform.

Photo: NASA/Wikimedia Commons