{"entities":[{"id":"rand-bio-redteam-report-2024","name":"The Operational Risks of AI in Large-Scale Biological Attacks: Results of a Red-Team Study","subtype":"Report","description":"Research report RR-A2977-2, published 25 January 2024. Date from RAND's publication page for RR-A2977-2; the report PDF itself carries only '© 2024 RAND Corporation'.","sourceIds":["source-ab-rand-redteam","source-ab-rand-landing"]},{"id":"ab-prop-current-system-uplift","name":"Currently available AI systems increase an actor's biological-misuse capability"},{"id":"rand-corporation","name":"RAND Corporation"}],"claims":[{"id":"ab-ev-rand-redteam-no-difference","statement":"RAND’s red-team study states that it found no statistically significant difference in the viability of the plans produced by its red cells with or without large language model assistance, for the two models and the twelve cells of its main comparison.","derivation":"explicit","resolution":"supported","editorialState":"seed","evidenceIds":["evidence-ab-ev-rand-redteam-no-difference-1","evidence-ab-ev-rand-redteam-no-difference-2","evidence-ab-ev-rand-redteam-no-difference-3"],"attributes":{"evaluatedSystem":"Two unnamed large language models, referred to as LLM A and LLM B, as available in summer 2023","evaluation":{"what":"Whether access to a large language model meaningfully increased the feasibility of planning a biological attack, relative to internet access alone","how":"A red-team study: teams worked under an internet-only control condition or with model access, and their outputs were scored on a nine-point scale by subject matter experts through an adjudication process","result":"No statistically significant difference in the viability of the plans produced with or without model assistance, in the combined analysis and in the per-model comparisons","caveats":"The report states that participants might have lacked the expertise to emulate actual attackers, that the study design could have impeded realistic planning, that it did not measure the distance between the capability frontier and the knowledge needed, and that it would widen the test in future work."}}},{"id":"ab-ev-rand-redteam-published","statement":"RAND published 'The Operational Risks of AI in Large-Scale Biological Attacks: Results of a Red-Team Study' (RR-A2977-2) on 25 January 2024.","derivation":"explicit","resolution":"supported","editorialState":"seed","evidenceIds":["evidence-ab-ev-rand-redteam-published-1","evidence-ab-ev-rand-redteam-published-2"]},{"id":"ab-arg-d1-mouton-contests","statement":"RAND's report on the operational risks of AI in large-scale biological attacks (RR-A2977-2, 25 January 2024) states that it found no statistically significant difference in the overall viability of biological weapon attack plans generated with or without the aid of the large language models it tested, and that such planning is beyond the capability frontier of the models available in summer 2023.","derivation":"explicit","resolution":"supported","editorialState":"seed","evidenceIds":["evidence-ab-arg-d1-mouton-contests-1","evidence-ab-arg-d1-mouton-contests-2","evidence-ab-arg-d1-mouton-contests-3"],"attributes":{"thought":{"roles":["contesting","earliest-located"],"register":"nonfiction","qualifier":"The report's conclusions also state what it does not quantify.","engagement":{"kind":"engages-premise","rationale":"The report gives a measured comparison against an internet-only control as its reason, rather than only asserting the contrary."}}}}],"sources":[{"id":"source-ab-rand-redteam","title":"The Operational Risks of AI in Large-Scale Biological Attacks: Results of a Red-Team Study (RR-A2977-2)","authors":["RAND Corporation"],"publisher":"RAND Corporation","publicationDate":null,"version":"Retrieved for the candidate pack on 2026-10-28 (SHA-256 e09a4103027b4f3c9bd3f8d7e1b7f427fea66f4d636f67c5a254a5f88af9438e). Published PDF of RR-A2977-2, 24 pages, © 2024 Retrieved again at admission on 2026-10-30 (SHA-256 e09a4103027b4f3c9bd3f8d7e1b7f427fea66f4d636f67c5a254a5f88af9438e, the same bytes). Both passes of the candidate pack retrieved this file; the bytes are identical, and one source row carries it.","evidenceUrl":"https://www.rand.org/content/dam/rand/pubs/research_reports/RRA2900/RRA2977-2/RAND_RRA2977-2.pdf","language":"en"},{"id":"source-ab-rand-landing","title":"RAND publication page for RR-A2977-2 (publication date)","authors":["RAND Corporation"],"publisher":"RAND Corporation","publicationDate":null,"version":"Retrieved for the candidate pack on 2026-10-28 (SHA-256 88627299fdd2be1a0fb8bc57ca2d13f19dc6b905216e17105a4fc0eaf81440e0). Publication page as served on 2026-10-28, used only for the stated publication date. Retrieved again at admission on 2026-10-30 (SHA-256 8db1fd8fbf0014fc4ede6a59e10c64ef7b8f749e4f0fbb3ef6520a8cea342cae). Its wording can change; earlier wording is not authenticated.","evidenceUrl":"https://www.rand.org/pubs/research_reports/RRA2977-2.html","language":"en"}],"evidence":[{"id":"evidence-ab-ev-rand-redteam-no-difference-1","sourceRole":"primary","support":"supports","verification":"machine_checked","retrievalRoute":"direct"},{"id":"evidence-ab-ev-rand-redteam-no-difference-2","support":"supports","verification":"machine_checked","retrievalRoute":"direct"},{"id":"evidence-ab-ev-rand-redteam-no-difference-3","support":"context","verification":"machine_checked","retrievalRoute":"direct"},{"id":"evidence-ab-ev-rand-redteam-published-1","sourceRole":"primary","support":"supports","verification":"machine_checked","retrievalRoute":"direct"},{"id":"evidence-ab-ev-rand-redteam-published-2","support":"supports","verification":"machine_checked","retrievalRoute":"direct"},{"id":"evidence-ab-arg-d1-mouton-contests-1","sourceRole":"primary","support":"supports","verification":"machine_checked","retrievalRoute":"direct"},{"id":"evidence-ab-arg-d1-mouton-contests-2","support":"supports","verification":"machine_checked","retrievalRoute":"direct"},{"id":"evidence-ab-arg-d1-mouton-contests-3","support":"supports","verification":"machine_checked","retrievalRoute":"direct"}],"at":{"entities":[1247,1254,1262],"claims":[3364,3365,3426],"sources":[1072,1090],"evidence":[4615,4616,4617,4618,4619,4725,4726,4727]},"version":"1lg47t7"}