X has rolled out a new version of its Grok AI model that catches stolen content three times more often than before. The platform found 1.5 million stolen posts in its latest round of enforcement as of July 2026, according to TechCrunch.

Under the new policy, when someone steals a post and tries to hide it with watermarks, intros, or other edits, X sends the ad money from that post to the original creator instead. This works for both video and viral text posts. Over $1 million in creator payouts will be redirected to the people who actually made the stolen content.

The rules get stricter for repeat offenders. Users caught repeatedly or intentionally trying to get around the anti-theft policy will be kicked out of the creator revenue-sharing program entirely. X also has a three-strike policy for engagement bait, which is content designed to trick people into liking, sharing, or replying through misleading tactics. Three strikes means removal from the creator program and referral to the policy team for possible account suspension. These measures run alongside X's existing bot-removal system, which was catching and suspending 208 bots per minute as of April 2026.

X product executive Nikitabier publicly criticized top creator Mr. Beast for using financial bait to drive video engagement, signaling that the platform's enforcement goes beyond small anonymous accounts to high-profile creators whose tactics bring significant traffic to the platform.

The technical challenge is significant. Catching a copied video that someone has lightly edited with a new intro or watermark requires analyzing what the video looks and sounds like, not just matching the underlying file data exactly. Think of it like recognizing a song even after someone has changed the tempo or added background noise — you recognize the tune because the overall pattern is the same. Applying the same detection to viral text posts adds another layer of difficulty, since a thief might paraphrase or rearrange the words. Grok's threefold improvement in detection suggests X has moved beyond simple file matching toward AI models that compare the meaning of content, not just the exact words or pixels.

The payout-redirection part is the more unusual piece. Systems like YouTube's Content-ID have long identified duplicate videos, but redirecting ad revenue to the original creator instead of just removing the copy changes the math for content theft. A thief who used to make money by repackaging viral posts now earns nothing — and actually generates income for the person they stole from. Whether the detection is accurate enough to avoid mistakes at a large scale is an open question, and one that matters a lot for creators whose original work might be wrongly flagged as stolen.

The broader context here is a platform trying to fix problems it partly created. X's revenue-sharing program, launched after the platform's rebrand from Twitter, paid creators based on how much engagement their posts got. That system rewarded engagement but also encouraged content theft, engagement bait, and bot-driven amplification. The current measures — better detection, redirected payouts, escalating penalties, and bot suspension — are an attempt to close that loop. The $1 million-plus in redirected payouts is a small amount compared to the total creator economy on the platform, but it sets up a system that could grow as detection gets better.

The most important unanswered question is what happens when the system gets it wrong. If the AI incorrectly decides that a creator's original post is a copy of someone else's, it redirects real money to the wrong person. That is a higher-stakes mistake than simply removing a post, because money changes hands directly. X will need to handle these errors transparently if the system is to keep creators' trust.

The Nikitabier-Mr. Beast dynamic is worth noting. When a platform executive publicly calls out a top creator, it signals that engagement-bait enforcement will not stop at small accounts. If the policy applies evenly, it could change how creators who have built large followings through financial-bait tactics operate. If enforcement ends up targeting smaller accounts while big creators just get warnings, the policy loses credibility fast.