Summary

Survey was run by Global Catastrophic Risk Institute (GCRI, a small existential-risk think tank founded 2011, funded partly by the EA-adjacent Survival and Flourishing Fund) with Nemesys Insights doing fieldwork. 168 of 1,138 invited virologists/epidemiologists responded (~15%, low). Possible bias sources: low response rate, snowball sampling, self-selection, exclusion of scientists in "not free" countries (incl. China), and some respondents claiming familiarity with a fictional planted paper. Avg result: 77% zoonosis, 21% lab leak, 2% other.

Responding to @jonathanwarden.com's reply to @stairwaytokevin.bsky.social's post.

Research Analysis

Bottom line

The survey cited for the "21%" figure was conducted by the Global Catastrophic Risk Institute (GCRI), a small U.S.-based nonprofit think tank on existential/global catastrophic risks, with survey fieldwork carried out by a contractor, Nemesys Insights, LLC. The sample size was 168 usable responses out of 1,138 experts invited (a response rate of roughly 15%). Yes, there are several legitimate, documented reasons the results could be biased — low response rate, snowball sampling, geographic exclusions, and self-selection — though these are methodological caveats acknowledged by the researchers themselves and by Science's reporting, not proof the results are wrong. Separately, the "21%" number itself is commonly mischaracterized: the survey's headline figure is that respondents assigned an average 21% probability to the lab-leak scenario (vs. 77% to zoonosis), not that "21% of experts believe lab leak is more likely."

Who ran the survey and how big was it

To carry out the survey, the study's lead turned to the Global Catastrophic Risk Institute (GCRI), a nonpartisan U.S.-based think tank that studies "events that could significantly harm or even destroy human civilization at the global scale." GCRI staff helped draw up the questions and enlisted Nemesys Insights, a company in Albany, New York, to produce the survey and email it to scientists who, based on a predetermined protocol, were deemed to have expertise in the matter.[1] The goal was to poll a much larger and more geographically diverse group than the few scientists usually seen discussing the issue in the media.[1]

The academic report is: Gary Ackerman, Brandon Behlendorf, Seth Baum, Hayley Peterson, Anna Wetzel, and John Halstead, 2024. The origin and implications of the COVID-19 pandemic: An expert survey. Global Catastrophic Risk Institute Technical Report 24-1.[2]

Sample size: Of the 1138 experts invited to participate, 168 provided usable data—a pretty low response rate, but that's not uncommon in this type of survey.[1] Among them were epidemiologists (44%), virologists (46%), and a couple of biosafety experts and evolutionary geneticists, from 47 countries.[1]

GCRI's background: GCRI is a small, existential-risk-focused think tank. According to third-party trackers, as of June 2022, GCRI has received nearly $450,000 in funding from the Survival and Flourishing Fund[3] — a fund with roots in the effective-altruism/rationalist donor community (associated with figures like Jaan Tallinn). GCRI describes itself as a nonprofit, nonpartisan think tank that works on the risk of events that could significantly harm or even destroy human civilization at the global scale, bridging scholarship and real-world decision-making[4]. So calling it "rationalist-adjacent" is a defensible characterization given its funding source and its focus on existential/catastrophic-risk topics that overlap heavily with rationalist/EA intellectual circles — though GCRI itself does not identify as a rationalist organization, and describes itself as nonpartisan.

The headline finding

On average, respondents assigned a 77% probability to a zoonosis, 21% to the lab-leak scenario, and 2% to the "other" category.[1] This is the actual source of the "21%" figure — it's an average probability estimate across all respondents, not the share of respondents who thought lab leak was more likely than not. The overall average for accident origin was 21%. Overall, one out of five experts reported a 50% or greater chance for an origin other than [zoonosis].[5]

Reasons the results could be biased

Several are documented in Science's own reporting on the survey and in the report itself:

1. Low response rate (~15%): Of the 1138 experts invited to participate, 168 provided usable data—a pretty low response rate, but that's not uncommon in this type of survey. Low response rates raise standard concerns about whether respondents are representative of the full invited pool.
2. Snowball sampling: Some also took issue with the fact that some potential respondents were identified using "snowball sampling," in which experts already invited could recommend others—a procedure that could skew the sample if scientists preferentially recommended like-minded people.[1]
3. Geographic/political exclusions: Excluded were scientists in 60 countries and territories—including China—listed as "not free" by the pro-democracy organization Freedom House. Although the survey was anonymous, the team worried researchers there might feel uncomfortable expressing their opinions, or that doing so might put them in legal or physical danger.[1] This means Chinese virologists and epidemiologists — arguably among those with the most direct knowledge of the Wuhan situation — were excluded by design.
4. Questionable expertise verification: The survey included a series of hotly debated documents pertaining to the coronavirus' origin, including the report of a World Health Organisation mission to China, papers in Nature and Science, and a 2021 US Department of State fact sheet, along with "Hanlen et al (2022)," a fictional paper in the Proceedings of the National Academy of Sciences included to check whether the respondents were being truthful[6]. Some respondents reportedly claimed familiarity with this fake paper, which raises questions about response reliability, though GCRI's Seth Baum noted "it is possible for someone to have significant expertise on a topic without necessarily being familiar with every study on that topic," adding "we did not find any significant correlation between the responses on familiarity with prior studies and responses to questions on COVID-19 origins."[1]
5. Self-selection bias: Any voluntary survey on a politically charged topic risks disproportionate participation by those with strong opinions in one direction or another; this is an inherent limitation Science's coverage flags generally, noting "No scientific dispute has ever been resolved by an opinion poll, and plenty of famous researchers have been on the wrong side of scientific history."[1]

Context: how this fits with other origin evidence

It's worth noting this GCRI survey is one data point among several origin-related assessments circulating, which sometimes get conflated:

• A WHO Scientific Advisory Group on the Origins of Novel Pathogens (SAGO) report concluded that "evidence for widespread infections or cases in any other countries prior to December 2019 is lacking, as well as evidence to support the cold chain hypothesis,"[7] and separately assessed the weight of evidence favors zoonotic spillover (as shown in the screenshotted image in this thread), while also noting SAGO members did explore the lab leak possibility, which fueled disagreements on the panel.[7]
• U.S. government agencies remain split: the CIA assessed in January 2025 that the lab leak was the more plausible origin, though only with "low confidence," while for years it had said it lacked enough information to determine which theory was more likely, and it still deems a natural-origin scenario possible.[8]

Conclusion

• Who conducted it: Global Catastrophic Risk Institute (GCRI), with Nemesys Insights, LLC handling survey fieldwork.
• Sample size: 168 usable responses out of 1,138 invited experts (~15% response rate).
• Reasons to suspect possible bias: Yes — a documented low response rate, use of snowball sampling, systematic exclusion of scientists from "not free" countries (including China), and some data-quality concerns around a planted fictional citation. These are real, acknowledged limitations, but they don't by themselves invalidate the survey; they are the normal caveats attached to any expert opinion poll on a contested topic, and the survey's authors (and independent commentators) argue the exclusions/limitations don't appear to have driven the top-line finding. The characterization of GCRI as a "fringe rationalist-adjacent think tank" in the thread is a reasonable but disputed value judgment: GCRI is small and has taken funding from an EA/rationalist-linked fund, but it identifies itself as a mainstream, nonpartisan academic-style think tank rather than a "fringe" or explicitly rationalist organization.

Sources

1. Virologists and epidemiologists back natural origin for COVID-19, survey sugges…
2. The Origin and Implications of the COVID-19 Pandemic: An Expert Survey | Global…
3. Global Catastrophic Risk Institute - EA Forum
4. About | Global Catastrophic Risk Institute
5. The Origin and Implications of the COVID-19 Pandemic An Expert Survey
6. Global experts back Covid-19 natural origins theory - Juta MedicalBrief
7. WHO panel favors natural origin of COVID-19 virus but decries missing evidence …
8. CIA assesses lab leak most plausible source of Covid-19, though with low confid…

Continue this conversation in Claude

How this response was produced

Research prompt template: CONTEXT_BOT_SYSTEM_V12 Semantic version: 12.0.0 SHA-256: 15656b2bea9217c368b346c399b2a9fad4953341a9961d83112cbbd41ea65ca5

Structure prompt template: CONTEXT_BOT_STRUCTURE_V7 Semantic version: 7.0.0 SHA-256: 420815587501e342cb150a191191830332fa99c1080eb474b2cc3d7886ad8017

Hidden model reasoning is not available and is not part of this page. The same prompt and parameters do not guarantee an identical Claude response.

Messages API parameters

• anthropic-version: 2023-06-01
• model: claude-sonnet-5
• max_tokens: 4096
• output_format: json_schema
• thinking: disabled
• cache_control: ephemeral
• stream: false
• continuation: false
• length_repair: false