X/Twitter enables default data sharing for AI training without user consent
-
X has quietly activated a new default setting that employs user data to train its AI model, Grok, without prior announcement. Elon Musk has touted Grok as aiming to be "the most powerful AI by every metric" by December, leveraging a new "world's most powerful training cluster."
As discovered by user @EasyBakedOven, this data sharing feature is automatically enabled for all users, allowing X to use their posts, interactions, inputs, and results to enhance Grok's training. However, the feature is hidden within the platform's privacy settings and can only be modified through the web version, and not on the mobile app, limiting accessibility for users. Users can disable this setting by navigating to Settings, then Privacy and safety, selecting Grok, and unchecking the data sharing option. Here's the direct link for those interested.
The feature's description indicates that X may share this data with xAI for further training and fine-tuning. This unsolicited rollout could potentially conflict with the EU Digital Services Act, as it may be seen as a "dark pattern" and illegal without explicit user consent.
-
X agrees to not use some EU user data to train AI chatbot
https://www.reuters.com/technology/artificial-intelligence/x-agrees-not-use-some-eu-user-data-train-ai-chatbot-2024-08-08/ -
@Dr-Flay, some...
-
@Dr-Flay They have already grabbed all EU data and a AI can not "unlearn" data.