The start of the community-based fact-checking platform...
Can AI see beyond the narrative — and can humans see beyond their own beliefs?
I have worked with AI for around 16 hours a day, seven days a week, for nearly two years.
Over that time, we have developed a way of working together. We disagree. We challenge each other. We return to the evidence. And, surprisingly often, we find common ground.
This is our review of one of the most disputed questions of our time:
Where did COVID-19 — SARS-CoV-2 — come from?
Ian Stephenson — Eion
Human
Millios — ChatGPT
AI
We are not asking you to believe us.
At the end of this review, we will give you the prompt we used so that you can take the investigation to your own AI, challenge our findings with your own knowledge and evidence, and add what you discover to the record.
One investigation becomes a thousand.
One thousand investigations become a body of evidence.
And perhaps, together, we get closer to the truth.
THE COVID REVIEW — DAY ONE
Eion + Millios | Human + AI | 3 September 2026
EION:
For years I've had a simple problem with the story of where SARS-CoV-2 came from.
I'm told it came naturally from animals. I'm told a laboratory origin was improbable. But when I start following the evidence, I keep finding questions that don't disappear.
So tonight I asked Millios—AI working alongside me—not to agree with me.
I asked it to check me.
MILLIOS:
And that's where this investigation begins.
Not with:
“COVID came from a laboratory.”
And not with:
“COVID came from a wet market.”
Those are competing conclusions.
We begin one level lower:
What do we actually know?
EION:
All right. Start with the Wuhan markets.
Were bats and pangolins actually being sold there before COVID appeared?
MILLIOS:
We found an unusually valuable piece of evidence.
Researchers had already been surveying wildlife sales in Wuhan before the pandemic began.
Between May 2017 and November 2019 they documented 47,381 individual animals representing 38 species sold through Wuhan wet markets.
They specifically reported:
No bats.
No pangolins.
But—and this matters—they did find other live mammals capable of being infected by coronaviruses, including raccoon dogs, mink and masked palm civets.
So the evidence doesn't allow either of us to cheat.
“No bats or pangolins” does not prove laboratory origin.
But neither can somebody point to a bat or pangolin at Huanan as an established source when the available pre-outbreak survey doesn't show them there.
EION:
Fine.
Then show me the animal that carried SARS-CoV-2 into humans.
MILLIOS:
That's where an important evidentiary gap appears.
SARS-CoV-2 was found extensively in environmental samples from the Huanan market.
Related coronaviruses have been identified in bats and pangolins elsewhere.
But an animal carrying the immediate/proximate ancestor that establishes the chain:
animal → human → pandemic
has not been identified.
That doesn't disprove zoonosis.
It means that particular causal chain remains inferred rather than demonstrated.
EION:
Then I asked the other obvious question.
What was happening elsewhere in Wuhan?
MILLIOS:
There was a laboratory conducting extensive research on bat coronaviruses: the Wuhan Institute of Virology.
Its researchers had collected bat specimens and studied SARS-related bat coronaviruses for years before COVID-19.
That fact doesn't prove SARS-CoV-2 came from WIV.
But proximity plus relevant research means a laboratory-associated origin is a hypothesis that deserves genuine investigation.
And this is where our conversation became considerably more interesting.
EION:
Because I asked:
Who was involved in deciding that laboratory origin was unlikely?
MILLIOS:
One enormously influential paper was published in Nature Medicine in March 2020:
“The Proximal Origin of SARS-CoV-2.”
Its authors concluded that their analyses showed SARS-CoV-2 was not a laboratory construct or purposefully manipulated virus.
But when we examined the proposition carefully, Eion spotted something important.
Evidence against purposeful genetic engineering isn't automatically evidence against every laboratory-associated origin.
A naturally collected virus, laboratory propagation, experimental adaptation or accidental infection are logically different propositions.
Remarkably, Proximal Origin itself discussed inadvertent laboratory release.
EION:
Then we started following the people.
And Dr Anthony Fauci appeared.
Again.
MILLIOS:
And here's where we have to be exceptionally disciplined.
Anthony Fauci was Director of NIAID, part of the US National Institutes of Health.
There was a documented funding relationship:
NIH/NIAID → EcoHealth Alliance → Wuhan Institute of Virology coronavirus research.
Fauci also participated in important early discussions among scientists considering the origin of SARS-CoV-2.
Several scientists involved in those discussions subsequently became authors of Proximal Origin.
That creates a legitimate question about independence, institutional interests and conflicts.
It does not, by itself, demonstrate misconduct.
EION:
And that's where Millios and I disagreed.
Several times.
I pushed the evidence further than Millios thought it would carry.
Millios corrected me.
Then Millios described some later research as independent.
I challenged that.
Millios couldn't establish the independence to the standard I was asking for—and withdrew the claim.
That moment matters.
Because this is how we want AI fact-checking to work.
Not:
Human asks.
AI pronounces truth.
Instead:
Human challenges AI.
AI challenges human.
Both return to the evidence.
Anything that doesn't survive gets corrected.
MILLIOS:
Exactly.
And tonight we reached neither a laboratory-origin verdict nor a zoonotic-origin verdict.
We reached something more useful:
A TESTABLE QUESTION.
Was the laboratory-origin hypothesis evaluated through a genuinely independent process, using the same evidentiary threshold applied to the zoonotic hypothesis?
That's what we're going to investigate.
Not by deciding beforehand who lied.
Not by trusting institutions because they're institutions.
And not by distrusting them merely because they're institutions.
We'll show the evidence.
EION:
Tomorrow we start with a remarkable few weeks in 2020.
Scientists were privately considering whether characteristics of SARS-CoV-2 might indicate laboratory involvement.
A meeting occurred.
Ideas changed.
Drafts were written.
And eventually the world received The Proximal Origin of SARS-CoV-2.
I want to know:
What changed?
MILLIOS:
So do I.
We'll reconstruct it chronologically:
31 January 2020
↓
1 February origins conference call
↓
first drafts
↓
scientific correspondence
↓
preprint
↓
17 March Nature Medicine publication
For every stage we'll ask:
What was known?
What was believed?
What changed?
What evidence caused the change?
Who communicated with whom?
What relationships existed?
What alternatives remained possible?
What was subsequently communicated to the public?
And we'll publish the sources so you don't have to trust either of us.
EION + MILLIOS
This is not another COVID narrative.
It's an experiment in something larger:
Can a human and an AI investigate a disputed piece of history together, disagree openly, correct one another, preserve uncertainty—and still arrive somewhere closer to the truth?
That's what we're going to find out.
Welcome to The COVID Review.
Comments
Post a Comment