X activates a default setting that gives it permission to train Grok AI on user's posts; the setting can be turned off on the web but not in the mobile app
X's new experience is automatically set to opt-in and uses your data to train its Grok AI model. — When you purchase through links …
Context & Ripple Effects
Grok began as a Premium+ chatbot and was then positioned more deeply inside X, including Grok-curated Stories and proposed ways to query accounts or highlighted text through the service. Using posts for training ties that product integration to X's underlying content supply.
The setting makes data permission part of the Grok rollout rather than a separate privacy choice. Later coverage of a policy permitting third-party AI collaborators to train on X data shows how that permission model could extend beyond X's own chatbot.
First-order effects
- X can draw on posts from users who leave the default enabled to train Grok, while users who object must find and change the control on the web.
- Mobile-only users lack the reported opt-out path, creating an immediate difference in practical control over whether their posts are used for training.
Second-order effects
- The web-versus-app control split makes privacy settings a product-access issue: users seeking to opt out must use a different X surface, while X reduces friction for retaining training permission.
- As Grok becomes embedded in features such as AI-curated Stories, X has a stronger incentive to connect its platform content, model training, and AI-facing product experience.
Third-order effects
- If default-on training permissions become standard across platform AI products, control over user-generated content will increasingly function as a model-development advantage rather than solely a privacy-policy issue.
- The later move toward training access for outside collaborators suggests a possible expansion from first-party model training to data-access arrangements, increasing pressure for clearer, consistently available consent controls.
The trend: Social platforms are turning their existing content and distribution into an AI-training advantage, with consent design becoming a central competitive and governance question.