The Split

I. Four answers to one question

For three years, the most powerful intelligence apparatus on earth was handed a single question — where did COVID-19 come from? — and could not agree on the answer.

That, by itself, is not remarkable. Intelligence is rarely unanimous. What is remarkable is that the U.S. Intelligence Community said so, out loud, in documents it released to the public, and that when it finally named the disagreement it turned out to run right through the middle of the government: two agencies on one side, five and a council on the other, and the Central Intelligence Agency unable to choose.

This is not a story about which side is right. It is a story about a record that never resolved — and about how much louder the public certainty was, in both directions, than anything the agencies actually wrote down.

II. What a “split” actually is

Start with the vocabulary, because it carries the whole story.

When an intelligence agency reaches a judgment, it attaches a confidence level — high, moderate, or low — that describes how good the underlying evidence is, not how strongly the analysts feel. “Low confidence” does not mean an analyst is guessing; it means the sourcing is fragmentary, or contested, or open to more than one reading. Two agencies can hold opposite conclusions and both, honestly, mark them low confidence, because both are reasoning from the same thin record.

That is the situation here. As the IC put it in its August 2021 unclassified summary, “the IC remains divided on the most likely origin of COVID-19,” and all agencies agreed that “two hypotheses are plausible: natural exposure to an infected animal and a laboratory-associated incident.”¹ The disagreement, the summary explained, was not about the facts so much as about how to weigh them: “Variations in analytic views largely stem from differences in how agencies weigh intelligence reporting and scientific publications, and intelligence and scientific gaps.”¹

A divided assessment is not a cover-up, and it is not a verdict. It is what honest analysis looks like when the evidence will not carry a firm conclusion.

III. The two poles

In August 2021, the IC drew the split in numbers without names. “Four IC elements and the National Intelligence Council assess with low confidence that the initial SARS-CoV-2 infection was most likely caused by natural exposure to an animal infected with it or a close progenitor virus.”¹ On the other side: “One IC element assesses with moderate confidence that the first human infection with SARS-CoV-2 most likely was the result of a laboratory-associated incident, probably involving experimentation, animal handling, or sampling by the Wuhan Institute of Virology.”¹ And in the middle, undecided: “Analysts at three IC elements remain unable to coalesce around either explanation.”¹

Two years later, the June 2023 report mandated by the COVID-19 Origin Act drew the same map — and this time it used names.

The Department of Energy and the FBI on the lab side — “although for different reasons,” the report is careful to note. The National Intelligence Council and four other agencies on the natural side. And the agency most associated in the public mind with the question, the CIA, declined to land at all: “The Central Intelligence Agency and another agency remain unable to determine the precise origin of the COVID-19 pandemic, as both hypotheses rely on significant assumptions or face challenges with conflicting reporting.”²

There is a quiet movement buried in those two snapshots. In 2021, exactly one element held the lab view. By 2023, two agencies are named on the lab side. The assessment did not stand still; it shifted, as assessments do when analysts keep working a hard problem — and the documents record the shift without dramatizing it.

IV. What they all agreed on

The split is the headline. But underneath it is a floor of agreement, and the floor is where the loudest public claims go to die.

On the question that drove the most fevered speculation — was the virus built? — the agencies were close to unanimous. The 2023 report: “Almost all IC agencies assess that SARS-CoV-2 was not genetically engineered.”² The 2021 summary said the same, in confidence terms: “Most agencies also assess with low confidence that SARS-CoV-2 probably was not genetically engineered; however, two agencies believe there was not sufficient evidence to make an assessment either way.”¹

And on the most extreme claim of all — that COVID-19 was a weapon — there was no division at all. “All IC agencies assess that SARS-CoV-2 was not developed as a biological weapon.”² The 2021 summary had put it in the IC’s plainest register: “We judge the virus was not developed as a biological weapon.”¹

This is the part of the record that fairness requires stating loudly, because it cuts against a story many people were sure of: the agencies that could not agree on natural-versus-lab did agree, broadly, that the virus was not engineered and unanimously that it was not a bioweapon. A laboratory-associated incident — an accident around ordinary research — is a different claim than a designed pathogen, and the IC kept them apart even when the public conversation did not.

V. Why everyone’s confidence stayed low

The deeper question is why serious analysts, with access most citizens will never have, kept marking their judgments low confidence on both sides of the line.

The answer in the documents is unglamorous: the evidence was never there to settle it. The IC said it would remain stuck “unless new information allows them to determine the specific pathway for initial natural contact with an animal or to determine that a laboratory in Wuhan was handling SARS-CoV-2 or a close progenitor virus before COVID-19 emerged.”¹ Neither of those things happened. And the one actor who could have changed that did the opposite: “Beijing,” the summary noted, “continues to hinder the global investigation, resist sharing information and blame other countries.”¹

That is the honest center of this story. The lab-leaning agencies gave weight to “the inherently risky nature of work on coronaviruses” at the Wuhan Institute of Virology.¹ The natural-leaning agencies gave weight to China’s “lack of foreknowledge” and “the numerous vectors for natural exposure.”¹ Both were reading a record with a hole in the middle where the decisive evidence should be — and both, to their credit, marked their answers accordingly.

VI. What the record establishes — and what it does not

This investigation does not take a side on where COVID-19 came from, because the agencies whose job that is did not, and could not.

The record establishes that the U.S. Intelligence Community was formally and openly divided on the origin of the pandemic; that, by 2023, the Department of Energy and the FBI assessed a laboratory-associated incident as most likely while the National Intelligence Council and four other elements assessed natural exposure and the CIA could not decide; that the split rested on differences in weighing a thin, contested record rather than on the facts being in dispute; and that the agencies were near-unanimous that the virus was not genetically engineered and unanimous that it was not a bioweapon.

The record does not establish that any one of these agencies is correct — “low confidence” and “moderate confidence” are the IC’s own warnings against treating its conclusions as the final word. It does not establish that the split was the product of political pressure or bad faith; nothing in these assessments says so, and the documents attribute the disagreement to analytic weighting, not to interference. It does not establish a lab origin, and it does not establish a natural one. And it does not, on its own, resolve the louder controversies this series has examined elsewhere — Anthony Fauci’s engagement with the agencies as they worked the question, or the scientific confidence gap that ran alongside the intelligence one.

VII. Why it matters

It is tempting, watching a government disagree with itself for three years, to read the disagreement as failure. It is closer to the opposite.

The people with the most access to the classified record looked at it and, in large part, refused to claim more than it could bear. They split. They marked their confidence low. They told the public they had split and why. Against that, the certainty that filled the airwaves — it was definitely a lab leak; it was definitely natural; it was obviously a bioweapon — was not a more rigorous reading of the evidence. It was a less rigorous one.

The split is not the scandal. The pretense, on all sides, that the question had been settled is the thing the record does not support.