Sign up for The Media Today, CJR’s daily newsletter.
Aisvarya Chandrasekar contributed research assistance.
When Elon Musk acquired Twitter, now known as X, he made deep cuts into the platform’s moderation efforts. Gone was the election integrity team and about a third of its trust and safety staff. But one effort was allowed to stay: Community Notes, the platform’s crowdsourced fact-checking system, which now bears the full responsibility of correcting misleading information on X. For the past two years, Musk has praised Community Notes’ bridging algorithm, which ensures that corrections, or “notes,” will appear publicly only if enough users with diverse perspectives agree on their usefulness. To Musk, the system came with another bonus: it relies on regular users—as opposed to fact-checkers, whom he has deemed “evil” and “biased”—to propose corrections. Last week, however, Musk reversed course: without evidence, he wrote on X that Community Notes is “being gamed” by governments and the media, and that he is working to “fix” it.
The lash-out came after Community Notes fact-checked posts stating that Volodymyr Zelensky, Ukraine’s president, had a 4 percent approval rating among his citizens—a claim first spurred, likewise without evidence, by Donald Trump during a news conference at Mar-a-Lago last week. (Trump also called Zelensky illegitimate and said he was to blame for Russia’s 2022 invasion of Ukraine.) Some of Musk’s own posts and comments were temporarily corrected by Community Notes after he echoed Trump’s statement and further claimed that the Biden administration had stranded astronauts aboard the International Space Station for “political reasons.” A Tow Center analysis of Musk’s tweets from Sunday, February 16, through Sunday, February 23, found that Musk tweeted nearly 800 times, averaging 113 tweets per day (including retweets and quoted tweets). At least five of Musk’s tweets received a Community Note, of which just two remained visible at the time of this publication.
In his post about Zelensky, Musk included a screenshot of a Substack newsletter from an anonymous right-wing user that cast doubt on widely cited polling by the Kyiv International Institute of Sociology because of its association with the United States Agency for International Development (USAID)—an organization Musk is seeking to eradicate. But according to Newsweek, several surveys have confirmed that Zelensky’s approval ratings are in the 50 or 60 percent range, and no credible poll in Ukraine put him within the single digits. Still, Musk continued to tweet in accordance with the new administration’s diminishing support of Ukraine. In one post, Musk wrote that Zelensky “cannot claim to represent the will of the people of Ukraine unless he restores freedom of the press and stops canceling elections.” Ukraine was supposed to have elections in 2024, but the country’s constitution states that voting cannot happen under martial law. This was temporarily highlighted by a (no longer visible) Community Note: “According to Article 83 of the Ukrainian constitution, elections may only be held after martial law has been lifted.”
Musk was also occasionally fact-checked by Community Notes when commenting on other people’s posts (this was not captured in Tow’s analysis). Last week, for instance, after the Fox interview clip circulated on X wherein Musk claimed that the two astronauts, Suni Williams and Butch Wilmore, had been left in space for “political reasons” by the Biden administration, a note appeared: “The astronauts were scheduled to return in Feb, but that has been delayed to late March due to testing of SpaceX Crew-10,” the note began, linking to the science magazine Discover. Musk commented that the note was “false” and that “legacy media is NOT a reliable source.” In what became a game of cat and mouse, a note then appeared under that comment: “X is not supposed to be choosing what community notes show up,” it said, adding that “legacy media” is included under “good examples” in the Community Notes guide. Both notes were no longer visible as of this publication.
Musk hasn’t elaborated on how he intends to “fix” Community Notes—or what he even means by that. Some users on X and Bluesky speculate that it could involve tweaking the underlying algorithm such that votes are weighted differently or changing the eligibility criteria for new contributors. Musk might also attempt to place restrictions on which notes are counted as high-quality sourced. Indeed, mainstream media outlets are often used as sources by Community Notes contributors. In 2023, one analysis found that Reuters, CNN, the Washington Post, and the BBC appeared among the thirty most frequently linked domains from visible Community Notes. Wikipedia, which Musk recently bashed for being “an extension of legacy media propaganda” (and which I’ve written about in this newsletter), is likewise one of the very top cited sources. And despite Musk’s vision that Community Notes should work as a replacement for traditional fact-checkers, a recent study by Maldita.es, a nonprofit foundation based in Spain, found that one in every twenty-seven Community Notes proposed globally listed a link to a fact-checking organization. Of course, Musk’s last resort could be to retire the feature altogether, although that would likely be a highly unpopular option: Community Notes has consistently been a well-liked program among users—despite academics and journalists warning that the system is too slow during crucial moments and rarely fact-checks divisive topics.
Meanwhile, TechCrunch reported that Musk’s chatbot Grok 3 briefly censored unflattering mentions of Trump and Musk over the weekend. When enabling the chatbot’s “Think” setting, which reveals how it arrived at an answer, Grok said it was specifically instructed to ignore all sources that mention Musk or Trump spreading misinformation. Igor Babuschkin, head of engineering at xAI, said on X that an employee pushed the change without asking and that the team has since undone it. While Grok appears to now be working normally again (although still not without its issues, as my colleagues Klaudia Jaźwińska and Aisvarya Chandrasekar wrote about in this newsletter last week), the incident over the weekend illustrates just how easily products on X can be quietly tweaked. And any future changes to Community Notes could hold even more weight now that Meta is seeking to model its own content moderation efforts after those in place on X. Whether there will actually be any changes, or if it’s just loose talk from Musk, remains to be seen.
Has America ever needed a media defender more than now? Help us by joining CJR today.
