Virology is a complex field that requires years of study to enter. As such, we should not expect that we can get a good sense of how concerning engineered pandemics are based on arguments alone. Instead, we should rely on the views of experts. Problematically, though, we don't have good enough data to know what experts actually think.
As far as I know, there have been three studies asking experts about their concern about engineered pandemics, but two of them have significantly biased methodologies, and the other one has a small sample size.
In this post, I go over how I found the surveys. Then I discuss the surveys themselves and why I think they’re biased. Then, I conclude by suggesting that someone should do a large non-biased survey.
Methodology
I found these studies by looking at Google, Google Scholar, and asking Chat-GPT.
Studies
Forecasting Research Institute’s 2023 Forecasting Existential Risks
The first study is the Forecasting Research Institute’s 2023 Forecasting Existential Risks survey. This study sought out participants based on their affiliation with organizations concerned about existential risk. It selected a sample of 80 experts on existential risks, 44% of whom had “spent more than 200 hours working directly on causes related to existential risk in the previous year” and 42% of whom “had attended an EA meetup.” Of these 80 experts, 12 were “graduate students with backgrounds in synthetic biology.”
When these graduate students were asked how likely they thought a catastrophic pandemic from engineered pathogens would be by 2100, they estimated 1.15% and then later updated to .8%. When asked how likely they thought extinction from an engineered pathogen would be by 2100, they estimated .7% then later updated to 1%.
I don’t think we should base our views to any significant degree on this survey since it had a very small sample size, the experts had not completed their education, and they were selected on the basis that they were already concerned about existential risks.
Forecasting Research Institute’s 2025 “Forecasting LLM-enabled Biorisk and the Efficacy of Safeguards”
The second study is Forecasting Research Institute’s 2025 “Forecasting LLM-enabled Biorisk and the Efficacy of Safeguards,” which examined how LLMs might increase the likelihood of a human-cased pandemic. This survey brought together 46 people “with expertise in biosecurity and/or wet lab biology.” 78% of these individuals had doctoral degrees. Their sampling approach was far less biased, selecting for participants based on their experience in biosecurity. They asked 1500 experts to participate in the survey, but only 46 completed it, giving a participation rate of 3%. They mention that the low participation rate may be due to the fact that they told participants to expect the survey “to take between 5 and 15 hours
These experts put a .3% chance of a human-cased biological catastrophe in 2028.
Given the extremely low participation rate and the fact that the survey focuses on what may be a fringe concern, I think we should not update significantly on these results.
Sabra et al.’s (2026) paper “Improving governance in the age of synthetic biology, artificial intelligence, and diverging threats” by
The third study is Sabra et al.’s (2026) paper “Improving governance in the age of synthetic biology, artificial intelligence, and diverging threats.” In this survey, 25 highly experienced biosecurity experts were requested to participate. 13 of them did. Of these, 7 considered themselves to be highly knowledgeable about synthetic biology warfare agents.
Overall, the survey found that experts place significant concern on the threat from engineered pathogens, although their concern for natural pathogens was slightly higher
I think this survey had generally good design, although, obviously the sample size was quite small. That said, it still seems like a quality enough study to conclude that there is likely some concern about engineered pathogens among biosecurity experts.
Conclusion
Overall, I think that only the last survey had sufficiently good design to draw any significant conclusions from it.
Given this lack of quality data, I think that someone should do a survey asking a wide number of biosecurity and virology experts to see how much of a concern engineered pandemics really represent within the field. They should ensure the survey is not biased by making sure that participants are not selected for based on their concern for engineered pandemics.
This is especially important given that Coefficient Giving has given over 290 million dollars in grants for biosecurity with a special concern for engineered pandemics. If it turns out that engineered pandemics are a relatively fringe concern, that could indicate that Coefficient Giving is seriously misallocating resources.
This is obviously a thing we want to know, and someone probably should poll the same virologists repeatedly over time as well.
I wouldn’t expect virologists to do an especially good job of pricing in risk from AI. AI people are better calibrated on virology than virology people are on AI. Folks at the intersection of these 2 fields are extremely sparse, so we tend to go with the AI person who’s read a little virology as our preferred podcast guests. This introduces some systemic errors in the dialogue, which could be corrected by polling virologists. Wet lab biology tends to instill a kind of humility that working with computers or theory doesn’t provide.
I’d want to poll virologists on specific line items like risk from cloud labs, gene synthesis screening, feasibility of first-timers executing key lab techniques, etc. A single “fear of pandemic” metric isn’t the right level of detail.
Agreed, I'd really want more like a RAND policy maker/rationalist to write out some theories of how people could do bio terrorism, and then grab the individual steps and ask viroligists. subject matter experts are usually quite myopic and can't see the bigger picture.
https://dianzhuo-wang.github.io/ fwiw this guy seems like he might have an interesting perspective.
Yeah, I agree that it's important to get more information than merely "fear of pandemic."
If you haven't already, I'd recommend reading Abi Olvera's series on this. She argues that AI people think about virology completely differently than virology people do. From the perspective she offers, engineered pandemics seem a lot less concerning than the AI community makes them out to be.
I just checked it out! I have mixed feelings about that series. I think the sentiment is probably directionally correct, but I found myself disagreeing with a lot of the claims and the framing. It reads to me like a non-expert summarizing expert views, and garbling them in the process. It didn’t seem very “AGI-pilled” either, for lack of a better term.
We don’t have many high quality takes on this stuff written out in public, which is a problem. I think the infohazard argument against this stuff is weak.
I'm glad you looked at it!