X updates terms to train AI models on user posts

Elon Musk famously threatened to sue Microsoft for illegally scraping Twitter data to train AI models. Then he built a fortress around the platform with aggressive anti-scraping walls.

Now we know what the fortress was protecting. X has quietly updated its Terms of Service to officially allow the company to use your posts to train its AI models.

Free data for xAI behind a locked door

Elon recently launched xAI, his own artificial intelligence startup. While xAI's website insists it is legally separate from X Corp, it also openly notes that it will work closely with X and Tesla.

Updating the terms creates a frictionless, proprietary data pipeline directly into xAI. Every hot take, debate, and news thread posted on the platform is now fair game for model training.

You do not get paid for your contributions, and X does not need to ask for extra permission.

This also casts Musk's recent pitch to journalists in a different light. He urged writers to publish directly and exclusively on X for more freedom and revenue, but it doubles as an invitation to feed high-quality editorial prose straight into his training pool.

The algorithm is also watching you back

The updated terms also lay the groundwork for deeper, automated AI filtering across the feed.

Based on AI-assigned definitions, your access to certain tweets, topics, or hashtags could be quieted or completely blocked. While algorithmic throttling is not new to Twitter, leaning directly on end-to-end AI systems makes distribution decisions much more opaque.

Instead of a shared public town square, the feed risks fracturing into hyper-personalized, algorithmically bounded echo chambers where visibility is dictated by a machine's perception of your profile.

Why it matters

We think this move shows where the entire social web is heading: walled gardens operating as proprietary data farms for in-house AI.

For everyday users and creators, the dynamic has fundamentally shifted. You are no longer just an audience viewing ads; your real-time thoughts are uncompensated raw training material. If you rely on the platform for reach, the reality is that your distribution is increasingly governed by opaque AI classifications that you cannot contest.

For the AI industry, it raises the barrier to entry even higher. Frontier models depend on fresh, authentic human data, and platforms with live feeds have the ultimate leverage. By locking out external crawlers and funneling everything to xAI, Musk gives his startup an exclusive pipeline that rivals like OpenAI and Anthropic cannot legally touch.

It is a shrewd corporate consolidation, but it carries clear user backlash risks. If contributors feel exploited as unpaid data workers inside an increasingly restricted feed, the very activity that makes X's data valuable could start leaking away.