X is facing a surge of highly realistic AI-generated war visuals that many users cannot distinguish from genuine footage. The platform announced a policy to suspend creators from revenue sharing for 90 days for undisclosed AI war videos, with permanent penalties for repeat offenders. Experts praise the intent but warn enforcement is difficult given monetization incentives, gaps in automated detection, and limits of Community Notes. The volume of deepfakes continues to outpace professional fact-checkers.
AI Deepfakes Flood X Amid Middle East War — New Policy Struggles to Curb Viral Fabrications
The Middle East war has unleashed a deluge of AI-generated visuals on X (Kirill KUDRYAVTSEV)(Kirill KUDRYAVTSEV/AFP/AFP)
AI-generated videos circulating on Elon Musk's X depict scenes such as American soldiers allegedly captured by Iran, an Israeli city in ruins, and US embassies on fire. These lifelike deepfakes have surged since the Middle East fighting escalated, often leaving users—and some automated systems—unable to tell fabrication from reality.
What X Announced
Last week X said it will suspend creators from its revenue-sharing program for 90 days if they post AI-generated war videos without a clear disclosure that they are artificial. Nikita Bier, X’s head of product, warned that repeated violations could lead to permanent removal from the program. The change marks a notable reversal for a platform widely criticized for becoming a haven of disinformation since Elon Musk’s acquisition in October 2022.
Mixed Reactions From Officials
Senior State Department official Sarah Rogers praised the move as “a great complement” to X’s Community Notes—its crowd-sourced fact-checking tool—because demonetizing misleading content reduces its reach. But many researchers and fact-checkers remain skeptical about enforcement and the policy’s likely impact.
Researchers Say The Problem Persists
“The feeds I monitor are still flooded with AI-generated content about the war,”
Joe Bodnar of the Institute for Strategic Dialogue told AFP, noting major posts from monetized “blue check” accounts continue to draw millions of views. Bodnar highlighted an AI clip shared by a prominent premium account that portrayed an Iranian ‘nuclear-capable’ strike on Israel, which outperformed Bier’s own announcement in views.
Platform Incentives And Technical Gaps
Researchers have warned that X’s model—where premium accounts and paid blue checkmarks can earn payouts tied to engagement—creates a strong financial incentive to publish sensational or fabricated material. AFP fact-checkers have documented numerous AI fakes, including videos showing tearful American soldiers in a bombed embassy, captured US troops next to Iranian flags, and a destroyed US naval fleet. One premium account posted an AI video of Dubai’s Burj Khalifa in flames and, despite a request from Bier to label it as AI, the clip remained online and accumulated more than two million views.
Compounding the problem, X’s own AI assistant, Grok, reportedly gave incorrect confirmations that some AI visuals were genuine when users asked for fact-checks, creating additional confusion.
Limits Of Community-Based Fact-Checking
Even with strict demonetization, many accounts spreading AI content are not part of X’s revenue-sharing program and therefore fall outside that enforcement channel. Those accounts can still be annotated through Community Notes, but studies raise doubts about its effectiveness. A Digital Democracy Institute of the Americas study found that more than 90 percent of Community Notes are never published, highlighting significant limitations in the system’s reach and timeliness.
Transparency Concerns And Sanctions Risks
The Tech Transparency Project reported that X appeared to profit from more than two dozen premium accounts connected to Iranian officials and state-controlled outlets pushing propaganda—potentially creating sanctions risks. X later removed blue checkmarks from some such accounts, the report said.
Expert Assessment
Alexios Mantzarlis, director of the Security, Trust, and Safety Initiative at Cornell Tech, called the policy “a reasonable countermeasure to viral disinformation about the war,” but warned that metadata can be stripped from AI content and that Community Notes are relatively rare. He said X faces a difficult trade-off if it is to achieve both high precision (not penalizing legitimate posts) and high recall (catching most bad actors).
As deepfake tools become more accessible and convincing, the platform’s ability to enforce disclosure, maintain high-quality fact-checking, and limit monetization incentives will determine whether the new policy meaningfully slows the spread of wartime disinformation.
Help us improve.
