Quick Q) Information hazards in open source lit reviews

By SofiiaF @ 2025-01-27T21:32 (+4)

Hypothetical precaution to avoid any issues in biosecurity risk, but if I was utilising completely open source, internet based, (not even behind a journal license) from very public sources (WHO, CDC, Against Malaria Foundation, WMP) for topics to do with pandemics, bioweaponry, biosecurity, global and public health (so slightly more information double edged than say anatomy or physiology or cell genetics), is there any safety measures that people would recommend for this level of secondary literature review of open source research for an unqualified student?

Reasons for and against taking measures:

Pro:

• good academic integrity to avoid any potential malicious or accidental misuse of information (purely in the sense of 'dangerous information that otherwise wouldn't be available to a threat actor, rather than accuracy of information')
• usually good to take precautions against even low probability but high impact risks rather than end up in a tricky spot
• good career/research training for specific niches of interest (biosec, pandemics) that both intrinsically condition me to highlight more risks, but also extrinsic signalling (unsure if that's important but it's a though process list)

Cons:

• open source public data for lit reviews have been vetted by individuals much more equipped and have redacted sections, and the fact all interviews or primary expert opinions I get are run through media experts or science advisors avoids any new information that hasn't already been approved ---> makes further effort to safeguard information hazards seem like too much energy/time for this low stakes style of research
• all information is public already so threat actors wouldn't have any difficulty getting the same information, and there was no substantial monetary or time (or at least not exceedingly lengthy) time barrier ---> plus raising information hazard concerns has the low but potential risk of aknowledging and 'signposting' the risk of the data
• many assumptions made in how threat actors would gain anything (honestly I think 2-3 people will read my work and that is probably just my family) in particular from a secondary lit review, especially since primary data on methods to utilise techniques isn't included (I do broad scope, non method driven, explanations) and they would still need great technical apparatus (e.g. to do gene editing against broad vaccine candidates) which means knowledge, time and money that would raise authorities attention more than reading random papers
• double edged so also could prevent good intentioned or simply curious individuals from learning about technologies
• simply not admitting dangerous theories or ignoring them can mean threat actors acquire it anyway but the defence community is on the backfoot with no previous awareness or no work to neutralise the appeal or exclusivity of a potential weapon/hazard
• the practicalities of discussing vs redacting information hazards would require using up more experts time to ask questions and gain opinion, and would result in some less practical paper reviews without case studies or anecdotes (still removing any instructional data and instead focusing on past experiences from a holistic view of secondary effects)

Sources I found helpful in my consideration:

Podcasts- had some amazing recommendations including bio(un)ethical, and the 80k podcast on pandemic prevention, and also some pure ethics and philosophy of science podcasts

Books- reading Pathogenesis, Behave, etc really helped but suprisingly the books on information hazards in AI actually helped me gain an understanding more on the abstract concepts of dual use information

Papers- Natural Selection Favors AI over humans (may seem unrelated but the whole chapter on informational channels in human vs AI systems painted the assumption of 'all malicious information hazards should be hidden from all' as less credible), and most of all Lewis et al Information Hazards in Biosecurity 

People- lots of chats with very kind individuals, from cold emailling to EA communities to uni groups to organisers and advisors who talk through their experiences with research or with highly complex topics

Questions I have:

• Any case studies or patterns of early-career researchers in these fields using open source public data that still managed to fall into hazardous territory (basically is it even a problem)
• Thoughts on information hazard considerations and whether it is more important in specialised research or if it should be trained into your method throughout
• Practical considerations in information hazards in biological reviews in terms of pandemics, security, weaponry etc especially with potential secondary political, cultural or social effects of discussing public and global health
• Good resources to further understand the discussions in the field on information hazards or open access to research etc

Would love to hear anyone's thoughts!