Government’s Half-Million-Pound Review of Online Misinformation Built on AI Summaries of Unread Papers, Ignoring Contradictory Evidence, Free Speech Union Says
Britain is preparing an apparatus of internet censorship without precedent in the modern era, according to a new publication from the Free Speech Union (FSU), which argues that the entire case for regulation rests on a flawed evidence base. The proposed measures include statutory “quality signals” in content ranking, a state-linked database of “false and misleading narratives”, official “prebunking” campaigns across education and public communications, and a Green Paper proposal that social media algorithms be compelled to give prominence to news sources the state deems “trustworthy”. Ofcom, the media regulator, holds a crisis protocol in readiness for the next emergency. The Government’s case for these sweeping powers rests on two central claims: that misinformation saturates the internet and inflicts grave harm, and that platform algorithms are the engine driving it. Yet when tested against primary sources and the peer-reviewed literature, the evidential base for both claims fails, according to the FSU’s analysis, which draws on the Government’s own commissioned review, academic studies published in Science and Nature, and research conducted by the social media platforms themselves.
The proposals, the FSU contends, originate from a small, interconnected ecosystem of campaign groups, consultancies, self-appointed fact-checkers, former broadcasters and parliamentarians whose recommendations consistently point toward more censorship, and whose members enjoy close ties to government. The Centre for Countering Digital Hate (CCDH), a prominent advocate of stricter regulation, was co-founded by Morgan McSweeney, who later became Sir Keir Starmer’s Chief of Staff. After the Southport tragedy, the CCDH chaired an “emergency” roundtable with officials from the Department for Science, Innovation and Technology, the Home Office, Ofcom and the Metropolitan Police. The Global Disinformation Index (GDI), another influential voice in the debate, has been funded by the Foreign Office. When the Government commissioned a review to prove that the problem was real, members of the same lobby served as its “expert” reviewers, creating what the FSU describes as a closed loop in which the demand for censorship produces the evidence that justifies it, with the taxpayers footing the bill for a process that merely ratified a predetermined conclusion.
The Government’s evidence review itself, the FSU argues, is a case study in misinformation. The GO-Science review, commissioned to supply the “full evidence base” that ministers admitted they lacked, cost taxpayers almost half a million pounds. Yet its authors made no attempt to establish what is true or false, relying instead on AI-generated summaries of 228 papers they had not read and on monthly questionnaires returned to the European Digital Media Observatory by fact-checkers. The review’s claim that platform algorithms are “exclusively profit- and engagement-driven” rests on a single citation: an AI-assisted synthesis from Universitas Muhammadiyah Buton, an Indonesian institution ranked 10,370th in the world. The FSU highlights the irony that a review intended to provide the intellectual foundation for regulating misinformation was itself produced using unverified AI summaries and a source of negligible academic standing. The publication also draws attention to commentary from Andrew Doyle, who described the proposals as “deeply Orwellian” in a broadcast for That’s Debatable in August 2026, reflecting growing concern among free speech advocates that the Government is building a machinery of thought control on the shakiest of foundations.
The empirical record, the FSU points out, contradicts the Government’s core assumptions. A 2024 study published in Science found that misinformation flagged by fact-checkers attracted just 0.3% of 2.7 billion vaccine-related views on Facebook during the vaccine rollout. Unflagged “vaccine-skeptical” content containing no falsehoods had 46 times the aggregate impact on vaccination intentions, and the most influential single item was an accurate article from the Chicago Tribune, seen by more than 50 million people. This finding undermines the claim that suppressing flagged misinformation would have any meaningful effect on public health outcomes. A 2024 synthesis published in Nature similarly found that exposure to misinformation is far smaller than is commonly supposed, with “fake news” accounting for roughly 0.15% of the average American’s daily media diet. The FSU argues that these findings are not marginal quibbles but fundamental challenges to the entire rationale for government intervention: if misinformation is a tiny fraction of what people consume, and if accurate reporting is more influential in shaping opinions that regulators deem harmful, then the case for algorithmic re-engineering and state-designated “trustworthy” sources collapses.
The claim that algorithms favour extreme content is likewise unsupported by the evidence. Twitter’s own researchers, studying millions of accounts, found no evidence that its algorithm favoured the far left or the far right over moderate voices. A landmark Facebook experiment found that the algorithm actually served less untrustworthy content than a simple chronological feed, contradicting the narrative that platforms are designed to amplify falsehood. And a study of 46 conspiracist beliefs, tracked in national surveys since 1966, found no evidence that belief in conspiracy theories is rising; significant declines outnumbered increases. Meanwhile, the Government’s chosen exemplar of trustworthiness, the BBC, is struggling for public confidence. Ofcom’s Media Nations 2026 report found that only 33% of adults associate the BBC with impartiality, and only 45% with trustworthiness. Recent events, including the doctoring of a Donald Trump speech by BBC Panorama and Ofcom’s critical ruling on a Gaza documentary, suggest the public may have a point. The FSU contends that building a regulatory system on the notion that state-appointed arbiters can identify “trustworthy” news is dangerously naïve when the state’s own flagship broadcaster is so widely distrusted.
Ultimately, the FSU warns, “misinformation” is being redefined as dissent. For the CCDH, the GDI and the GO-Science review, it is “misinformation” to treat the pace and cost of decarbonisation as matters of economic and political judgement rather than physics. The GDI pronounced “clearly untrue” a vaccine statistic drawn from Public Health England’s own data, then set about persuading advertisers to defund the publisher; its chief executive has explained that content can be “factually accurate but still extremely harmful”. Whoever controls the label controls the debate, and Britain has seen nothing like these proposals since the Star Chamber decreed in 1637 that nothing be published unless licensed. The FSU, a membership organisation with more than 45,000 members, is urging the public to join its campaign against the proposals, arguing that once the state gains the power to designate which statements are acceptable, all other rights are conditional. The publication concludes with a stark warning: the apparatus being assembled in the name of protecting democracy may end up dismantling it, and the evidence used to justify it does not survive contact with the facts.
