COVID-19 origins: the hypotheses, and what they were called

← COVID-19: a sourced timeline ← What the US government said Shoshin: a beginner's mind →

This page does not claim to know where the virus came from, and nothing on it should be read as evidence for any origin. The origin is unresolved. American intelligence agencies remain split, at low confidence, and no primary source here settles it. What this page tracks is narrower and answerable from documents: which hypotheses were on the table, what was published when, what the lab hypothesis was called at each point, and what Anthony Fauci said at each of those points.

Sources on this page are unrestricted — journals, government, newspapers, platform policy — because the subject includes how publications and platforms treated the question, which cannot be told from government documents alone. Every source is still labelled by what it is, and every quotation was read in the linked document.

A rule the repository sets for itself, and how it applies here. These pages exclude "reception": how a record was received, who aligned with whom, what any audience concluded. That rule survives intact. A dated, attributable, quotable publication or a written platform policy is not reception — it is a document, with an author and a date, exactly like an order or a memo. "The media said" is a characterisation of a crowd and stays out. Individual named publications saying specific things on specific days stay in.

Chain of custody. Every row is graded by how far its attribution goes, and one grade is not admitted at all:

on the record the person's own words, in a transcript, a recording, or a paper they signed. You can watch or read them saying it.
document published by the body whose act it is — an agency report, a journal statement, a platform policy, a newspaper's own headline or editor's note.
named reporting journalism whose source is named and on the record, or which publishes the underlying document.
anonymous — not admitted a claim resting on unnamed sources does not appear on this page at all, regardless of the outlet's reputation or how widely the claim has been repeated. See "Sought, and not included".

No laundering through links. A citation here points at the artifact itself — the recording, the transcript, the report — never at a piece describing another piece. "This article reports that another article reported" is how a claim acquires authority it never earned, one hop at a time, and each hop is where the qualifiers fall off. Where we cannot reach the artifact, the row says we cannot reach it. That is why one significant assessment is absent from this page entirely.

An office is not a person, and the difference is dated. An institution's only ultimate voice is whoever holds its top office, so a director speaking in that capacity is the institution speaking — but only while they hold it. Rows carry the office and the years it was held, because the same human being carries different authority before and after. The clearest case on this page is Anthony Fauci: in 2021 he speaks as the sitting director of NIAID, and in 2024 he speaks as a private citizen under subpoena. Same person, same subject, two different things.

The process and the prompts This page's source

The hypotheses, as most often stated

Three claims have carried most of the public argument. They are not equally supported and the page does not treat them as though they were.

The horizontal axis is time. The vertical axis carries no magnitude and no ranking. Rows are stacked in date order so every document has somewhere to sit with its citation.

Each row is tagged with the hypothesis it bears on, and with whether it supports it, weighs against it, or only concerns what it was called.

Our voice appears in two places only: which documents share this axis, and the lines marked "our call".

The argument, stated plainly

This is ours, and it is the only claim on the page that is. The boundary of respectable opinion was drawn before the evidence could support drawing it, and it was redrawn quietly. In February 2020 a hypothesis that United States intelligence agencies would later judge plausible — and that two of them would come to favour — was formally condemned in a medical journal as a conspiracy theory, in a statement organised by a man whose own institution funded coronavirus research at the laboratory in question, a competing interest not disclosed for sixteen months. A month later a paper declared no laboratory scenario plausible. A major newspaper called the idea debunked. A platform made it a removable offence. Each of those was reversed or corrected between May 2021 and 2023, and every correction was quieter than the original.

That is a claim about the labelling, not about the virus. If the lab hypothesis is eventually ruled out entirely, this page's argument is unaffected: the objection is to condemning a hypothesis as illegitimate while the evidence to settle it did not exist, not to the answer that condemnation happened to favour.

The strongest objections to it

These are not strawmen. The first one is the most serious criticism this page faces and we think it lands partially.

Two very different claims are being blurred together, and the blur favours us. What was condemned in early 2020 was largely the engineered bioweapon claim — that the virus was deliberately built or released. That claim was poorly supported then and remains unsupported: the intelligence community judged the virus was not developed as a biological weapon. The accidental research-related incident hypothesis is a separate and far more credible proposition. A page that lets the vindication of the second rehabilitate the first is doing exactly the sleight of hand it accuses others of. The rows below are tagged by hypothesis for this reason, and readers should check whether we kept the line straight.
Scientists updating on evidence is correct behaviour, not hypocrisy. The signatories of the February 2020 statement and the authors of the March 2020 paper were writing at a moment of near-total ignorance, and several publicly changed position as evidence accumulated. Treating a revised view as an exposed lie punishes the exact behaviour we spent the rest of this repository asking for.
Platforms faced real harm, not hypothetical harm. Moderating a pandemic information space included genuinely false and dangerous claims. Drawing the line in the wrong place is a serious error; it is not the same as drawing it in bad faith, and a page that does not distinguish those is not being fair.
The newspaper correction does not vindicate the politician. The Washington Post's editor's note withdraws the words "debunked" and "conspiracy theory" on the ground that no determination had been made. That is a statement about the label, not an endorsement of what Senator Cotton actually said, parts of which invoked the deliberate-release framing.
Absence of a proven natural origin is not evidence of a lab origin. Both remain open. A timeline of failed condemnations can read as a scoreboard for the other side, and that reading is not supported by anything here.
Fourteen documents is a sample, and we picked it. The literature is enormous and a different fourteen could support a different emphasis.

Sought, and not included

The attribution rule cost us three things a reader would reasonably expect to find here. Publishing what a rule excludes is the only way the rule means anything.

The Department of Energy's assessment that a laboratory incident was the likely origin, reported in February 2023. It is a real and significant assessment. It is also known publicly only through reporting that rests on unnamed officials describing a classified annex. No named person has stated it on the record and no document has been released. Under the rule above it is not admissible, however widely it has been repeated. This is why the FBI row exists instead: its director said it himself, on camera, with his name on it.
The teleconference of 1 February 2020 between Anthony Fauci, Francis Collins and several virologists, which is central to almost every account of this story. There is no transcript. What is on the record is correspondence around the call and later testimony about it, and the two do not agree. A row asserting what was said on that call would be a row asserting something nobody can produce.
Any claim resting on an unnamed source, however credible the outlet. We do not care who anyone believes Deep Throat was; an anonymous source is not attributable, and a page that admits one has no principled way to refuse the next. If a document is released or a named person goes on the record, it becomes admissible that day.

How to argue with this page

You don't have to persuade us. Everything here sits in a single public file that anyone can edit; you change what you think is wrong and send us your version, and we accept it publicly or say publicly why not. The most useful contributions to this page in particular: documents that show the engineered claim and the accidental claim being carefully distinguished at the time, which would weaken our argument; and any early, prominent, dated defence of the accidental hypothesis as legitimate, which would strengthen it.

See the file everything lives in Send us a change Take your own copy Or just tell us what's wrong

All documents

Date	Document	Kind	Bears on	Source