Elon Musk's X (Twitter) has added an "opt-out" clause that denies users from using user posts for AI learning. In the end, he bowed his head to the European Union's privacy regulations.
Bloomberg reported on the 27th that X added a method of changing the setting so that it could not train "Grook" with user posts.
X is basically set up to use user data for AI learning. Musk also boasted of Grook's strength by citing real-time performance. Currently, there is also a service that summarizes news events using posts.
However, on this day, X said, "All users can use public posts to control whether they can train Groch, an AI search helper," adding, "This setting is available on the web platform and will be released on mobile soon."
This is in accordance with the EU's privacy regulations. However, the opt-out method is likely to become a problem in the future.
TechCrunch said on the same day that the Irish Data Protection Commission (DPC), an EU privacy watchdog, was paying attention to X's user post policy.
"I've been talking to X about this for months, and we talked about it yesterday as well," said Graham Doyle, DPC's deputy director. "I'm surprised at the quick policy shift. I've got X to follow up today, and I'm waiting for a response."
DPC is an organization that complies with the EU's General Data Protection Regulations (GDPR). Recently, Meta's user data issue has become a hot topic of conversation.
On the 14th of last month, DPC put the brakes on learning "Meta AI" with user data on Meta's Facebook and Instagram. The key reason is that it interfered with the exercise of user rights by adopting an opt-out method that places the user denial clause in an invisible place and allows users to choose.
For this reason, Meta postponed the launch of 'Meta AI' in the EU, and protested that the multimodal model to be unveiled later would not be serviced in the EU.
X also adopts the same opt-out method as Meta, so it is highly likely that the DPC said it is a follow-up.
As such, amid the growing interest of regulators in AI learning of social media user data, the existing embedded opt-out method continues to be pointed out as a problem.