Between 2022 and 2025, the three biggest social platforms dismantled the rules that limited health misinformation. Here is what each owner did, what they say about it, and what the record shows.
Our page on Wan v. OpenAI is about one chatbot and one patient. This page is about the larger environment those chatbots learn from and that most people now get their health news in. During the pandemic, X (then Twitter), Facebook, Instagram and YouTube all built rules to label, demote or remove false health claims. Within three years, each had taken those rules down.
The owners give a consistent reason: the rules suppressed legitimate debate, the fact-checkers were biased, and government officials leaned on the companies to remove posts. That argument deserves to be stated fairly, and it is below. But the practical result for anyone selling a miracle cure, an anti-vaccine message or a reason to stop taking your medication is the same on every platform: the brakes are off.
| Platform | Owner | What was dropped | When | What replaced it |
|---|---|---|---|---|
| X (Twitter) | Elon Musk | COVID-19 misleading-information policy; amnesty for suspended accounts | 23 Nov 2022 | Community Notes; periodic spam-bot purges |
| Facebook, Instagram, Threads | Meta (Mark Zuckerberg) | Third-party fact-checking and the penalties attached to it (US) | Announced Jan 2025; ended 7 Apr 2025 | Community Notes, which do not reduce a post’s reach |
| YouTube | Alphabet | Standalone COVID-19 rules; bans on creators terminated under them | COVID rules ended Dec 2024; reinstatement offered Sept 2025 | A general medical-misinformation policy; no fact-checkers |
Weeks after Musk bought Twitter, a one-line note appeared on its rules page: as of 23 November 2022 the company was no longer enforcing its COVID-19 misleading-information policy. There was no announcement. Under that policy, Twitter had suspended more than 11,000 accounts and removed nearly 100,000 pieces of content. The change came as Musk promised a general amnesty for suspended accounts and after he laid off roughly half the staff, including much of content moderation.
Before the purchase, Musk vowed to defeat the spam bots or die trying. X does run purges — in October 2025 Musk said 1.7 million reply-spam bots had been removed, and a larger sweep followed in 2026. Those sweeps target spam behaviour. An account that posts health falsehoods at a human pace, or a bot that does the same thing quietly, is not what they are built to catch, and there is no longer a rule against the content itself.
Musk is not a bystander. A 2024 NewsGuard analysis found that four of the five accounts he engaged with most often were repeat spreaders of false claims. In April 2026 NewsGuard documented a 50-million-view boost he gave a COVID vaccine hoax built on 2,133 death reports to Germany’s Paul Ehrlich Institute. The institute itself had judged only 74 deaths possibly or probably linked to the vaccine, out of about 197 million doses.
Musk’s AI company builds Grok, which is built into X and was partly trained on X posts. An audit published in BMJ Open in April 2026 put five public chatbots through 50 health questions each on cancer, vaccines, stem cells, nutrition and athletic performance. About half of all answers were judged problematic. Grok produced significantly more highly problematic answers than expected — 29 of 50, or 58% — and the authors pointed to its training on X as a likely reason. (The audit tested older model versions; newer ones may perform differently.)
In July 2023, X sued the Center for Countering Digital Hate, a non-profit that had published reports on hate and misinformation on the platform. In March 2024, U.S. District Judge Charles Breyer dismissed the suit under California’s anti-SLAPP law, writing that “this case is about punishing the defendants for their speech.” X appealed. Separately, X won a Ninth Circuit injunction in September 2024 against parts of a California law that would have required platforms to report how they handle misinformation and hate speech.
Readers of this site know the pattern: a critic documents a problem, and the answer is a lawsuit. We have been on the receiving end of that ourselves.
Meta started paying independent fact-checkers in 2016, partly to deal with public-health misinformation. In January 2025 it announced it would end the program in the United States, saying experts had their own biases and that a program meant to inform had too often become a tool to censor. It also stopped demoting fact-checked posts and replaced full-screen warnings with a smaller label. Fact-checking ended on Facebook, Instagram and Threads in the US on 7 April 2025; previously issued strikes were removed.
Its replacement, Community Notes, is modelled on X’s. A note appears only when contributors with differing viewpoints agree it is accurate, and, unlike the old system, a note does not limit how widely the post is shown.
YouTube banned misinformation about approved vaccines in 2021. It retired its election-fraud rule in 2023 and folded its remaining COVID-19 rules into a broader medical-misinformation policy in December 2024. In September 2025, responding to a subpoena from House Judiciary Committee chair Jim Jordan, Alphabet’s counsel wrote that YouTube would let creators back on the platform if they had been terminated under the retired COVID and election rules. The letter said senior Biden administration officials had pressed the company over content that did not break its rules, and that YouTube would not use fact-checkers.
In fairness, YouTube still has a medical-misinformation policy on the books, which X does not. But the reinstatements bring back channels with large audiences that were removed precisely for repeated health falsehoods.
Each point has some truth in it. None of them explains why the answer to imperfect enforcement was to stop enforcing anything — or why it took effect on the claims most likely to kill people.
For decades the quack’s biggest problem was distribution: a newsletter, a radio slot, a booth at the health show. The platforms solved that problem for them, then spent three years taking away the few obstacles left. The people who own these companies are entitled to their views on free speech. They are not entitled to pretend that a feed engineered to reward outrage, with no rule against telling a mother her child’s vaccine causes autism, is a neutral public square.
Corrections are welcome.