Inside the Window

COVID-19 — the live case

WHO · Imperial College · SAGE · national governments · platform moderation

Status Deliberately not scored. This is the only case in the study still inside its own window — the period in which the official position has not settled and the disconfirming evidence has not finished arriving. Every other record here is scored with hindsight this one does not yet have. Two components have now been worked up in full because they can be evidenced today: the vaccine injury and compensation record, and the origins question and how it was policed, both immediately below. The rest stays open.

What the pillar's method requires of it Separate the claims and score each on its own: Imperial Report 9's 510,000 unmitigated figure — conditional, and reported as a prediction, from the same group whose 1997 vCJD upper bound was ten million against an eventual 178, and whose 20-million-lives-saved estimate the MHRA and ONS now undercut by orders of magnitude; fomite transmission and surface cleaning; the masks reversal; school closures against measured learning loss; the lab-leak hypothesis, moderated as misinformation in 2020–21 and assessed as plausible by several agencies from 2023; transmission-blocking claims; and the delayed-diagnosis and backlog harms. Some will score as forecast-far-off, some as averted, some as dismissed-and-real. Anyone who expects them all to score the same way has decided the answer.

Why it is the hardest test of the whole study Because it is the one case where my own bias is at maximum and yours is too — see prediction P4 in the audit. Build the other three pillars first, fix the scoring rules on cases where the answer is already known, and then apply those rules here without changing them. Doing it in that order is the only thing that will make the conclusion worth anything.

Who gained Deliberately not answered here, and the omission is the point. Every other record in this study names its beneficiaries with hindsight — knowing how it ended, which parties were vindicated, and which interests turned out to have been served. None of that is available yet. What can be said now is structural and applies to all sides: the asymmetry that governs every actor in an emergency is that no one is ever destroyed for having done too much. Where that asymmetry pointed on any particular decision, and who it served, is exactly what this pillar declines to score until the window closes.

Vaccine injury, the AstraZeneca claim, and a scheme designed in 1979

AstraZeneca · UK Government · High Court · the Covid-19 Inquiry

What is claimed Group litigation in the High Court under the Consumer Protection Act 1987, coordinated by Leigh Day, alleging that the Oxford–AstraZeneca vaccine was less safe than persons generally are entitled to expect because of vaccine-induced immune thrombotic thrombocytopenia. Lead claimant Jamie Scott suffered a severe brain injury in April 2021.

What AstraZeneca has and has not conceded In High Court documents in early 2024 the company accepted that the vaccine can, in very rare cases, cause TTS. That is an admission of generic capability only. It disputes the claimants' framing of causation, denies the product was defective, and has conceded nothing about any individual claimant. Generic causation, specific causation and defectiveness are three separate hurdles. The admission was very widely reported as though it settled all three.

Where the case actually stands (verified 8 Sep) No judgment has been handed down. The National Archives' Find Case Law database returns no substantive judgment in the claim. The first case management conference was moved from December 2024 to February 2025; disclosure is in progress. Leigh Day represents 50 individuals and families and is closed to new claims. Its own case page still says trial is "likely in 2025/2026" but carries no date, so that line should be read as aspiration, not status.

And it is shrinking, for a legal reason rather than an evidential one Families have withdrawn after advice that they were unlikely to succeed, because the patient information leaflet from 7 April 2021 disclosed the rare clotting risk — which goes directly to what people were "entitled to expect". Whether anyone was harmed and whether the product was legally defective are different questions, and the second is the one being litigated.

The Inquiry did what the court has not The Covid-19 Inquiry's Module 4 report held both findings at once: the vaccination programme was an overall public health success that saved many lives, and people were seriously injured or died from side effects and government has a duty to support them. Baroness Hallett recommended reforming the Vaccine Damage Payment Scheme — raising the maximum from £120,000 to at least £200,000, indexing it, introducing tiered payments by degree of harm, and covering injuries below the 60% disablement threshold — "as soon as possible".

Three findings for the study 1. The obstacle for most affected people is not corporate wrongdoing but the design of a 1979 compensation scheme: a flat sum with a cliff-edge threshold, which pays the same to someone 60% disabled and someone 100% disabled, and nothing at all below the line. 2. It was an independent inquiry with disclosure powers, not litigation, that moved this — the same countermeasure that recurs throughout these registers. 3. Hallett's report states both halves of a contested case without picking one, which is precisely the discipline this study demands of itself and rarely finds in an official document.

Who gained AstraZeneca is not paying for its own defence. Under the pandemic indemnity agreement the UK government is effectively funding it — a fact stated on the claimant firm's own page. So a successful claim would be met substantially by the taxpayer, and the manufacturer's exposure is far smaller than the case's framing implies. This is the indemnity structure recorded in the Pandemrix entry in Pillar I, operating in plain sight a decade later.

The origins question, and how it was policed

US intelligence agencies · platform moderation · the White House · Congress

Where the assessments actually stand (verified 8 Sep) The intelligence community is split, and mostly at low confidence. The FBI and the Department of Energy assess a laboratory-associated incident as more likely — the FBI with moderate confidence, the DOE with low. The CIA moved to favouring a lab origin in 2025 but at "low confidence", the lowest of three tiers. The National Intelligence Council and four other agencies assess natural zoonotic exposure as more likely. Peer-reviewed work continues to point toward a zoonotic origin linked to the Wuhan wildlife trade. Both hypotheses remain live and the question may never be settled.

The White House page, scored by this register's own rules In April 2025 the administration replaced the federal COVID information resource at covid.gov with a page titled Lab Leak: The True Origins of COVID-19, asserting a lab origin as established. Under the tier rule in Pillar III this is Tier B — an official body's advocacy — and it is not evidence for its own conclusion. That judgement is not about which conclusion it reaches: a government website asserting a contested scientific claim is the same category of object whichever way it points, and this register would be worthless if it applied that test to the WHO and not to the White House.

Why it belongs here anyway Because the operation is one this study already documents: a government replacing a public-information resource with a position on a contested question. Structurally that is what the UK government did in distributing An Inconvenient Truth to every secondary school, which a High Court judge required to be accompanied by corrective guidance. Same mechanism, opposite politics, and both belong in the register on the same terms.

The failure that does not depend on resolving the science The lab-leak hypothesis was labelled misinformation and suppressed on major platforms during 2020–21 — at a time when the intelligence assessments were already divided and no agency held high confidence either way. Whatever the origin turns out to be, treating an unresolved question as settled and moderating the other side of it is a documented institutional failure, and it is the part of this record that can be scored now.

And the transparency record, which is checkable The COVID-19 Origin Act of 2023 passed both chambers unanimously and was signed in March 2023, requiring declassification within 90 days. A further provision in the National Defense Authorization Act ordered another declassification review. Substantial portions of the released material remain redacted under both mandates — which is a concrete, dated, verifiable finding about institutional conduct, and does not require anyone to decide where the virus came from.

Who gained On one side, an administration that had staked itself on the hypothesis and gained a vindication narrative. On the other, from 2020: platforms that avoided a moderation controversy by deferring to an official line, scientists and institutions with funding or reputational exposure to gain-of-function research, and a public-health communications strategy that treated uncertainty as a threat to compliance. Both columns are populated, which is why the record stays classified as unresolved.

What the Covid-19 Inquiry did not publish

UK Covid-19 Inquiry · HART and three other groups · APPG on Pandemic Response and Recovery · The Telegraph

What is established (verified 8 Sep) Four groups holding views critical of the official position — among them the Health Advisory and Recovery Team — applied for core participant status in Module 4, on vaccines and therapeutics. The applications were refused. Each group was, however, invited to submit a written witness statement under Rule 9, and did so. The Module 4 report was published in April 2026. Those statements have still not appeared on the Inquiry's website.

The Inquiry's answer Reported by The Telegraph. An Inquiry spokesperson said it publishes witness statements that are entered into evidence during or after the inquiry's hearings and publishes evidence that the chair considers sufficiently relevant to its hearings and reports. That is a normal policy for a public inquiry receiving very large volumes of material, and it is not a denial.

Why it belongs in this register, whatever you think of the groups HART is a pressure group whose substantive positions on lockdowns and vaccines are strongly disputed by mainstream science, and nothing here endorses them. The finding is structural: the body being examined for how it handled dissenting evidence is also the body deciding which dissenting evidence enters the public record — and its test is relevance as the chair judges it. That is the same defect as the Climategate record in Pillar III, where the resistance to disclosure did more damage to trust than disclosure could have.

Correction — the statements ARE public, and I said otherwise My first draft of this record said the position was unfalsifiable from outside because nobody could judge whether the statements were weak or merely unwelcome. That is wrong. HART published its own statement in June 2024 at the People's Vaccine Inquiry site, and has since reposted all four groups' statements on its own. HART's is a 204-page document by Dr Clare Craig, a diagnostic pathologist and HART co-chair, titled Unsafe and Defective. The material is available to anyone who wants to assess it, and the correct criticism is narrower than the one I made.

A fact that cuts both ways HART says it was initially asked not to publish, and published anyway once the Module 4 hearings slipped from July 2024 to January 2025. Read one way, a party that broke an expectation of confidentiality gives an inquiry a reason not to publish. Read the other, a body that solicits evidence and then neither publishes it nor visibly engages with it has few grounds to complain that the author published it themselves. Both readings should be in the record.

What the narrower criticism actually is Not that the evidence was suppressed — it is public. It is that the Inquiry solicited written evidence under Rule 9 from parties it had refused core participant status, and then neither published those statements alongside the rest of the evidence base nor, on the face of the report, engaged with them. Whether the report engages with their substance is a separate, checkable question that this register has not yet answered.

Having now read it (8 Sep) Unsafe and Defective runs to about 102 pages of text plus figures. It is structured in two parts — safety (adverse-event systems, disability data, myocarditis, deaths, immune and reproductive effects) and effectiveness (eleven numbered claims it sets out to rebut) — and it is referenced predominantly to official UK sources: ONS, PHE, SAGE, MHRA and NHS England, plus German and Norwegian reporting systems and the KCL Zoe study.

Its strongest material Three things are genuinely serious. First, the argument that passive surveillance detects rare single-organ harms well (VITT, myocarditis) and systemic harms against a common background poorly — that is methodologically sound as stated, and she supports it with the Pandemrix precedent, where the narcolepsy estimate was raised a decade later, which is a verified record in Pillar I of this study. Second, the case-counting window argument: that excluding the first two weeks after vaccination from the "vaccinated" group biases effectiveness upward if risk is elevated in that window. She quotes the ONS's own finding that in unadjusted analyses the risk of infection increased following first vaccination, peaking at around 16 days, PHE testing data, and a SAGE observation of an abundance of patients admitted to hospital within 7 days of vaccination. This is a live methodological debate in the peer-reviewed literature, not a fringe claim. Third, the observation that the MHRA, ONS and NHS give UK lives-saved figures far below what the Imperial 20-million global model implies.

Where it overreaches, on its own evidence The executive summary states that the lack of benefit confirms the fraud from the clinical trial data — fraud is asserted, not demonstrated in the text. Several central exhibits are ecological correlations used causally: NHS bed occupancy and ambulance calls rising alongside the rollout, with confounders the document does not dispose of — deferred care, the elective backlog, post-lockdown presentation, capacity constraints predating 2021. And the claim that the vaccinated attended emergency departments five times more often than the unvaccinated is exposed to precisely the denominator problem the document correctly identifies elsewhere: the old and the sick were vaccinated first.

The honest verdict, which is not a comfortable one for either side It is a serious, referenced document by a credentialed pathologist, containing at least one methodological critique that is live in the literature — and it draws causal conclusions its own evidence does not establish. Both halves are true, and neither can be reached by characterising the author. Having read it, I think the non-publication is harder to defend rather than easier: whatever its faults, it is not obviously irrelevant to Module 4. It addresses Module 4's subject directly and argues from the Inquiry's own evidence sources.

What is alleged but not documented The All-Party Parliamentary Group on Pandemic Response and Recovery published an account on 2 January 2026 of a presentation by Dr Clare Craig, alleging that the Inquiry excluded evidence on lockdown effectiveness and harms, care-home policy, infection fatality rates and vaccine safety; that testimony from vaccine-injured people was cut short; and that Carl Heneghan's evidence was set aside. Score this as allegation, not finding: the document cites no published exclusion list and no independent submissions. Separately, 21 MPs and peers warned as early as November 2022 that the Inquiry risked being dominated by one perspective. And on Module 3, HART's own core-participant application lapsed because it was not renewed by the deadline — which is not the Inquiry's doing.

The correction this forces on my own record In the entry immediately above I praised the Module 4 report for stating both halves of a contested case without picking one, and called it precisely the discipline this study demands. That praise needs a qualifier I did not apply: the report states both halves of the evidence the Inquiry admitted, and the admission decision was its own and is not visible. I assessed an institution's output without asking what its process had excluded — in a study whose entire method is asking exactly that. Logged as note 20 in The Deference Audit.

Who gained The Inquiry, in a cleaner and more coherent report than a fully adversarial record would have produced, and in avoiding hearings that would have been used as a platform. Government, in recommendations aimed at scheme design rather than at the decisions themselves. And — the recurring cost — the excluded groups, who gained the strongest argument they have ever had: not that they are right, but that they were not published. An institution that withholds weak evidence makes it indistinguishable from suppressed evidence, and pays for that itself.

The data that only came out under court order

US FDA · CDC · PHMPT · ICAN · Judge Mark Pittman · Judge Matthew Kacsmaryk

The Pfizer licensing documents Public Health and Medical Professionals for Transparency filed a Freedom of Information request in September 2021 for the data the FDA relied on to license the Pfizer vaccine — some 329,000 pages. The FDA proposed releasing about 500 pages a month, a schedule running to roughly 2096 — 75 years (verified 8 Sep). In January 2022 Judge Mark Pittman of the Northern District of Texas held the request to be of paramount public importance and ordered production at 55,000 pages a month.

V-safe The CDC built an active surveillance system — a smartphone check-in tool used by around 10 million people — and did not publish its results. ICAN requested the data in June 2021 and obtained it only after two lawsuits and a court order. Of roughly 10 million users, 782,913 — about 7.7% — recorded a health event requiring medical attention, emergency care or hospitalisation. The first release covered only the tick-box fields; the free-text entries were withheld, and a federal judge ordered the CDC to produce every one of them by 15 January 2025.

What that 7.7% does and does not mean It is not a vaccine-injury rate. V-safe is self-reported, has no control group and no comparator period, and "sought medical care" captures any medical contact in the window for any reason. Presenting it as 782,913 injuries is the same error as reading raw VAERS totals as confirmed harms. What the figure genuinely establishes is that the system recorded a large volume of post-vaccination medical contact — which is exactly what an active surveillance system is for, and exactly what was not published.

A correction to that paragraph, and the asymmetry in it The study's owner put the obvious question to me: if "no control group" disqualifies V-safe, what about the trial? The point is right and it runs against the institutional side, not for it. V-safe never had a comparator by design — it is surveillance, not a study, and criticising it for lacking one is criticising it for being what it is. The Pfizer trial had a randomised control group and gave it up. Unblinding began on 14 December 2020 with the emergency authorisation; it was carried out non-randomly, prioritising older participants, and it converted the randomised controlled trial into an open-label study with a small and selectively-composed residual placebo arm (verified 8 Sep). So the standard I applied to the surveillance system is one the trial itself abandoned, at roughly two months' median follow-up, in precisely the period where a comparator mattered most: long-term safety, and all-cause mortality, where the counts were frozen at a size too small to support any conclusion in either direction. The stated justification was ethical — that an authorised effective vaccine should not be withheld from placebo recipients — and that is a real argument, contested at the time, with alternatives available such as blinded crossover or retention of a subset. It was a choice, made by the sponsor and permitted by the regulator, and it destroyed the evidence that would now settle several of the disputes in this register. Logged as note 22 in The Deference Audit: I applied an evidentiary standard to material used by critics of the official position and did not simultaneously apply it to the official evidence base. That is prediction P1, in its textbook form, and I did it while writing a record about applying standards symmetrically.

Why this belongs in the register, whatever the data show The finding does not depend on interpreting the contents. Two public health bodies collected safety data on a product given to hundreds of millions of people and released it only when compelled by courts, in stages, over three to four years. That is the identical mechanism as Roche withholding the Tamiflu clinical study reports for four years, as the Post Office resisting disclosure, and as the CRU resisting Freedom of Information requests — all already in these registers. It is the study's most repeatable finding: the countermeasure that works is an outsider obtaining the underlying data, and it usually takes a court.

Who gained The manufacturers and the regulators, from delay — a 75-year schedule is not a redaction policy, it is a decision that the public would never see the file. And, once again, the requesters gained the strongest argument available to them, which is not that the data are damning but that they had to sue twice to see it.

Twenty-eight days, counted one way

Public Health England · the daily dashboard · the same bodies assessing vaccine harm

What the measure was England originally counted as a covid death anyone who died at any point after a positive test, regardless of cause and with no time limit. Carl Heneghan and Yoon Loke of the Oxford Centre for Evidence-Based Medicine put it plainly in July 2020: a patient successfully treated and discharged would still be counted even if they had a heart attack or were run over by a bus three months later. The Health Secretary ordered a review, and in August 2020 a 28-day cut-off was adopted, aligning England with the devolved nations and removing about 5,000 deaths from the cumulative total (verified 8 Sep).

The asymmetry, which is the point "Died within 28 days of a positive test" and "died within 28 days of a vaccination" are the same kind of claim. One was adopted as the headline national statistic, reported every evening, and used to justify the most extensive peacetime restrictions in modern British history. The other was met, correctly but selectively, with correlation is not causation. The same institutions, in the same period, applied a permissive temporal-proximity rule in one direction and a strict causal standard in the other.

Why this is stronger than most entries in the register Because it forecloses the usual defence. You cannot say different fields carry different evidentiary norms — it is one set of bodies, one period, one form of inference, two standards. And neither standard is wrong in itself: proximity counting is a defensible real-time proxy, and demanding causal evidence before attributing harm is good practice. The failure is that the choice between them tracked which error the institution preferred to make. Where undercounting was feared, the permissive measure led. Where overcounting was feared, the strict one did. Those are policy judgements about tolerable error, presented to the public as methodological standards.

What the charge should NOT be The 28-day measure itself was not a fabrication and calling it one weakens the case. It was transparently defined, published in parallel with the slower death-registration series on the same dashboard, and it was a tightening of something considerably worse — a correction made in about a month, prompted by outside critics, which by this register's standards is exceptionally fast. The defensible charge is not manipulation of a statistic. It is asymmetric application of evidentiary standards, and the selection of which measure became the headline. That is harder to answer and more damaging, because it cannot be fixed by amending a definition.

The counterweight this record owes The covid count had an external check that vaccine-proximity claims largely lacked: all-cause excess mortality rose in step with the waves, which is a corroborating signal independent of attribution. That does not dissolve the asymmetry, but the two are not perfectly parallel and the register should say so. Note also that excess mortality in the period after 2021 is itself contested, and is one of the arguments in the HART statement above — so the counterweight is weaker than it first appears.

Disclosure I committed this exact failure earlier in this same session: I applied "no control group, therefore no causal inference" to V-safe and did not apply it to a trial that had a control group and gave it up. That is evidence for the mechanism rather than against it. This does not require bad faith or coordination. It requires only that one error feels more costly than the other — which is the finding this whole study keeps arriving at, and the reason it has no villain to prosecute.

Who gained From the permissive count: every policy that needed public compliance, and every institution whose position would have been weakened by a smaller number. From the strict standard on vaccine harm: manufacturers, the rollout, and public health communication that treated uncertainty as a threat to uptake. Both columns are populated by substantially the same people.

Adjudicating inside the window

EU Commission · UK NSOIT · WHO · UN · WEF · the platforms · Stanford Internet Observatory

What exists (verified 8 Sep) The EU Digital Services Act is the consequential instrument; its Code of Conduct on Disinformation was endorsed in February 2025 and folded into the DSA's co-regulatory framework on 1 July 2025, with Google, Meta, Microsoft and TikTok among the signatories alongside fact-checkers and civil-society bodies. Britain has the Online Safety Act 2023 and the National Security Online Information Team, formerly the Counter Disinformation Unit. The WHO launched "infodemic management" in a September 2020 joint statement with the UN, UNICEF, UNDP, UNESCO, UNAIDS, ITU and the IFRC. The UN published Global Principles for Information Integrity in 2024. The WEF has ranked mis- and disinformation the top two-year global risk in 2024 and 2025, and it remains top in 2026.

Why this register bears on it Every one of the other 132 records in this study shares a single feature: the disconfirming evidence existed while the official position was still official. That interval is what this pillar is named after, and it is exactly where information governance operates — necessarily adjudicating before questions settle. And the asymmetry that runs through the whole register predicts the failure mode without needing any bad actor: a moderator who removes something true faces no penalty and no name attached to the error; one who leaves up something later judged harmful faces a hearing. The career-safe direction is over-removal. It is the peanut committee and the Fukushima evacuation in a new medium.

Note also who the register's dissenters were Semmelweis, Alice Stewart, Brooksley Born, Harry Markopolos, Raghuram Rajan. Every one held a minority position against an institutional consensus, and every one was right. Stewart's funding was withdrawn rather than her data answered — which is the pre-digital form of the same mechanism, and cost a generation.

The documented instance, and it is not hypothetical Britain's Counter Disinformation Unit monitored Carl Heneghan, professor of evidence-based medicine at Oxford; Alexandre de Figueiredo; and Molly Kingsley, who campaigned to keep schools open. Heneghan appears elsewhere in this study for a specific reason: with Yoon Loke he identified that England was counting as a covid death anyone who died at any point after a positive test, with no time limit — a criticism the government accepted, acting within weeks to adopt the 28-day measure. A government disinformation unit monitored a critic whose correction that same government adopted and implemented. No inference is required; both facts are on the record.

The other half, which this register insists on These same pages carry Wakefield — Britain stripped of measles elimination status, and 83 dead in Samoa — plus the laetrile clinics and AIDS denialism in South Africa at roughly 330,000 avoidable deaths. False information kills, and that is the genuine case for governance. The finding here is not that this apparatus is illegitimate. It is that nothing exempts it from the asymmetry that ran every other institution in this study, and it presently has no mechanism that would catch its own error: no registry, no independent count, no penalty for removing something true.

What the historical record actually shows Across these registers, information was shut down repeatedly and the outcomes are traceable. Tobacco's doubt campaign delayed regulation by decades against internal knowledge dating from the 1950s. Kehoe's lead standard stood for forty years while Patterson was refused funding and removed from committees. Alice Stewart was corroborated in 1962 and ignored into the 1980s. Fujitsu's knowledge of remote access to Horizon was withheld from the courts through nine hundred prosecutions and at least four suicides. The NEJM's Expression of Concern on Vioxx came five years after publication, and its executive editor learned of Merck's internal memo only on being deposed. Study 329's negative data was reframed by a ghostwriter. Chisso concealed Minamata for over a decade. In none of these did the suppression prevent the harm. In all of them it extended it.

But the second-order effect is the larger finding Suppression does not merely delay a correction. It converts a bounded error into an unbounded one, by spending the credibility needed to manage the next event. Chernobyl is the clean case: the Soviet authorities suppressed the accident's scale in the first days, and the consequence was not calm but a European population that would not afterwards be reassured by anyone — producing 100,000–200,000 terminations of wanted pregnancies that the radiological evidence never warranted. The initial shutdown caused the later over-reaction. The same shape recurs: the tobacco playbook, once recognised, poisoned the well for legitimate scientific communication when the same firms sold it to the oil industry; the Wakefield fraud made the vaccine-safety question so unaskable that the genuine harms in this register — Pandemrix narcolepsy, the thimerosal removal, the 1976 swine flu programme — became difficult to state plainly without being misheard; Climategate's substantive science survived every inquiry and its credibility did not; and the Covid-19 Inquiry's decision not to publish four Rule 9 statements handed the excluded groups the strongest argument they have ever had. The cover-up reliably costs more than the thing covered up, and it is paid in a currency that cannot be replenished on demand and is needed most in the next emergency.

And the timing is exactly wrong Restriction of challenge is heaviest during emergencies — which is precisely when institutions are most likely to be wrong, because they are deciding fast, on thin evidence, at high stakes. This register bears that out: the swine flu programme of 1976, BSE, and the covid decisions were among the fastest and among the worst. The error-correction mechanism is switched off at the moment the error rate peaks.

The limit of this finding, stated plainly This register contains no case in which suppressing information improved the outcome — and that must not be read as evidence that none exists. A successful suppression is by construction invisible: it produces no scandal, no inquiry, no record, and therefore no entry here. The study is selected for failures, so it systematically underestimates how often suppression occurs and overestimates how harmful it is on average. This is the same denominator problem that the front page says the whole study cannot escape, and it applies to this record with particular force. What the evidence supports is narrow: where suppression has come to light, it has almost always cost more than the thing it concealed. That is a claim about the cases we can see.

What would close it Publication of what was flagged, what was removed, at whose request, and with what outcome. An independent count of false positives — true material suppressed — which no body in this record currently produces or has ever published. A ruling on the merits: Murthy v Missouri was decided 6–3 in 2024 on standing alone, so the American courts have never reached the question of whether federal pressure on platforms was constitutional. And note the divergence rather than assuming a monolith: the Stanford Internet Observatory shut in June 2024, the Election Integrity Partnership did not operate in the 2024 cycle, and enforcement is now intensifying in Europe while receding in the United States.

Who gained Governments, who acquired a channel to the platforms that requires no legislation and — in the UK's case — no notice to the person being monitored. The platforms, for whom being told what to remove is safer than deciding, and for whom compliance is cheaper than a DSA fine. The bodies themselves, in mandate, funding and standing — the identical gain recorded against the WHO's pandemic mandate in Pillar II. An entire funded research field. And, honestly, the public, in protection from Wakefield-class harm, which is real. But the structural beneficiary is the one this study keeps meeting: nobody has ever been penalised for removing something true, and until someone is, the incentive points one way only.