X updates its TOS to ban scraping and crawling without “prior written consent”, effective September 29, likely to stop companies training AI models on its data
Elon Musk-owned X, formerly Twitter, has updated its terms of service to prohibit scraping and crawling — likely to fend off any AI models training on its data.
Context & Ripple Effects
Days earlier, X had expanded its own ability to use collected data for AI training, while Musk said the effort would use public rather than private material. The new restriction therefore draws a sharper line between X’s own AI-training access and outside parties’ access to the same public-facing corpus.
Later policy changes reinforce that this was part of an evolving control strategy: X eventually allowed certain third-party collaborators to train on user data under its terms, while subsequently barring developers from using X API or content to train frontier models.
First-order effects
- Companies that crawl X without written permission now face a contractual prohibition, raising the compliance and legal risk of collecting its posts for model training or other data products.
- X gains a clearer basis to control access to its public content and distinguish authorized AI use from unapproved collection.
Second-order effects
- AI developers and data-collection vendors that relied on X content must seek permission, alter collection practices, or substitute other sources; X can use access terms as a negotiating lever.
- The policy makes X’s data position more asymmetric: it can pursue internal training while setting conditions for outsiders, a split later echoed in collaborator access for third-party AI training.
Third-order effects
- If platforms increasingly treat public posts as licensable AI inputs, training-data access shifts from an open-web collection problem toward negotiated platform permissions and enforceable terms.
- The boundary will likely be tested through product and API rules as well as contracts; X’s later ban on training frontier models with API or X content shows how such restrictions can broaden beyond web scraping.
The trend: This is an early instance of social platforms converting public-content access into a controlled, commercial AI-data boundary.