{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,22]],"date-time":"2026-06-22T11:27:54Z","timestamp":1782127674148,"version":"3.54.5"},"reference-count":66,"publisher":"Public Library of Science (PLoS)","issue":"5","license":[{"start":{"date-parts":[[2025,5,8]],"date-time":"2025-05-08T00:00:00Z","timestamp":1746662400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["www.ploscompbiol.org"],"crossmark-restriction":false},"short-container-title":["PLoS Comput Biol"],"abstract":"<jats:p>As a result of rapidly accelerating artificial intelligence (AI) capabilities, multiple national governments and multinational bodies have launched efforts to address safety, security and ethics issues related to AI models. One high priority among these efforts is the mitigation of misuse of AI models, such as for the development of chemical, biological, nuclear or radiological (CBRN) threats. Many biologists have for decades sought to reduce the risks of scientific research that could lead, through accident or misuse, to high-consequence disease outbreaks. Scientists have carefully considered what types of life sciences research have the potential for both benefit and risk (dual use), especially as scientific advances have accelerated our ability to engineer organisms. Here we describe how previous experience and study by scientists and policy professionals of dual-use research in the life sciences can inform dual-use capabilities of AI models trained using biological data. Of these dual-use capabilities, we argue that AI model evaluations should prioritize addressing those which enable high-consequence risks (i.e., large-scale harm to the public, such as transmissible disease outbreaks that could develop into pandemics), and that these risks should be evaluated prior to model deployment so as to allow potential biosafety and\/or biosecurity measures. While biological research is on balance immensely beneficial, it is well recognized that some biological information or technologies could be intentionally or inadvertently misused to cause consequential harm to the public. AI-enabled life sciences research is no different. Scientists\u2019 historical experience with identifying and mitigating dual-use biological risks can thus help inform new approaches to evaluating biological AI models. Identifying which AI capabilities pose the greatest biosecurity and biosafety concerns is necessary in order to establish targeted AI safety evaluation methods, secure these tools against accident and misuse, and avoid impeding immense potential benefits.<\/jats:p>","DOI":"10.1371\/journal.pcbi.1012975","type":"journal-article","created":{"date-parts":[[2025,5,8]],"date-time":"2025-05-08T17:38:52Z","timestamp":1746725932000},"page":"e1012975","update-policy":"https:\/\/doi.org\/10.1371\/journal.pcbi.corrections_policy","source":"Crossref","is-referenced-by-count":26,"title":["Dual-use capabilities of concern of biological AI models"],"prefix":"10.1371","volume":"21","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-1030-8776","authenticated-orcid":true,"given":"Jaspreet","family":"Pannu","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Doni","family":"Bloomfield","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Robert","family":"MacKnight","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Moritz S.","family":"Hanke","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0001-2968-3365","authenticated-orcid":true,"given":"Alex","family":"Zhu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Gabe","family":"Gomes","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Anita","family":"Cicero","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Thomas V.","family":"Inglesby","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"340","published-online":{"date-parts":[[2025,5,8]]},"reference":[{"issue":"6711","key":"pcbi.1012975.ref001","doi-asserted-by":"crossref","first-page":"831","DOI":"10.1126\/science.adq1977","article-title":"AI and biosecurity: the need for governance","volume":"385","author":"D Bloomfield","year":"2024","journal-title":"Science"},{"key":"pcbi.1012975.ref002","article-title":"EpochBiological sequence models in the context of the AI directives","author":"N Maug","year":"2024"},{"issue":"6681","key":"pcbi.1012975.ref003","doi-asserted-by":"crossref","first-page":"349","DOI":"10.1126\/science.ado1671","article-title":"Protein design meets biosecurity","volume":"383","author":"D Baker","year":"2024","journal-title":"Science"},{"issue":"7992","key":"pcbi.1012975.ref004","doi-asserted-by":"crossref","first-page":"570","DOI":"10.1038\/s41586-023-06792-0","article-title":"Autonomous chemical research with large language models","volume":"624","author":"DA Boiko","year":"2023","journal-title":"Nature"},{"key":"pcbi.1012975.ref005","article-title":"On the societal impact of open foundation models","author":"S Kapoor","year":"2024","journal-title":"arXiv"},{"key":"pcbi.1012975.ref006","unstructured":"Patwardhan T, Liu K, Markov T, Chowdhury N, Leet D, Cone N, et al. Building an early warning system for LLM-aided biological threat creation. 2024. [cited 2024 Mar 27]. Available from: https:\/\/openai.com\/research\/building-an-early-warning-system-for-llm-aided-biological-threat-creation"},{"key":"pcbi.1012975.ref007","author":"CA Mouton","year":"2023"},{"key":"pcbi.1012975.ref008","author":"J Kaplan","year":"2020"},{"key":"pcbi.1012975.ref009","author":"JB Sandbrink","year":"2023"},{"key":"pcbi.1012975.ref010"},{"key":"pcbi.1012975.ref011"},{"issue":"6691","key":"pcbi.1012975.ref012","doi-asserted-by":"crossref","first-page":"36","DOI":"10.1126\/science.adl0625","article-title":"Regulating advanced artificial agents","volume":"384","author":"M Cohen","year":"2024","journal-title":"Science"},{"key":"pcbi.1012975.ref013","article-title":"Responsible AI x Biodesign."},{"issue":"1","key":"pcbi.1012975.ref014","doi-asserted-by":"crossref","first-page":"97","DOI":"10.1038\/s44286-023-00002-4","article-title":"Self-driving laboratories to autonomously navigate the protein fitness landscape","volume":"1","author":"JT Rapp","year":"2024","journal-title":"Nat Chem Eng"},{"issue":"8","key":"pcbi.1012975.ref015","doi-asserted-by":"crossref","first-page":"633","DOI":"10.1016\/j.cels.2023.04.006","article-title":"In vitro continuous protein evolution empowered by machine learning and automation","volume":"14","author":"T Yu","year":"2023","journal-title":"Cell Syst"},{"key":"pcbi.1012975.ref016","article-title":"Laboratories in the cloud","author":"M Field","year":"2019","journal-title":"Bull At Sci"},{"key":"pcbi.1012975.ref017","author":"Microsoft Research AI4Science","year":"2023"},{"key":"pcbi.1012975.ref018","author":"S Wu"},{"issue":"8040","key":"pcbi.1012975.ref019","doi-asserted-by":"crossref","first-page":"808","DOI":"10.1038\/d41586-024-03815-2","article-title":"AI could pose pandemic-scale biosecurity risks. Here\u2019s how to make it safer","volume":"635","author":"J Pannu","year":"2024","journal-title":"Nature"},{"key":"pcbi.1012975.ref020","volume-title":"AI Safety Institute approach to evaluations","author":"GOV.UK"},{"key":"pcbi.1012975.ref021","volume-title":"Frontier AI taskforce: first progress report","author":"GOV.UK"},{"key":"pcbi.1012975.ref022","volume-title":"Introducing the AI Safety Institute","author":"GOV.UK"},{"key":"pcbi.1012975.ref023","volume-title":"Inspect","author":"Inspect"},{"key":"pcbi.1012975.ref024","volume-title":"The Bletchley declaration by countries attending the AI safety summit, 1-2 November 2023","author":"GOV.UK"},{"key":"pcbi.1012975.ref025","author":"The White House","year":"2023"},{"key":"pcbi.1012975.ref026","volume-title":"Memorandum on advancing the United States\u2019 leadership in artificial intelligence; harnessing artificial intelligence to fulfill national security objectives; and fostering the safety, security, and trustworthiness of artificial intelligence","author":"The White House","year":"2024"},{"key":"pcbi.1012975.ref027","article-title":"Simulating 500 million years of evolution with a language model","author":"T Hayes","year":"2025","journal-title":"Science"},{"key":"pcbi.1012975.ref028"},{"key":"pcbi.1012975.ref029","year":"2023"},{"key":"pcbi.1012975.ref030","volume-title":"United States Government Policy for oversight of dual use research of concern and pathogens with enhanced pandemic potential | OSTP","author":"The White House","year":"2024"},{"key":"pcbi.1012975.ref031","volume-title":"Removing barriers to American leadership in artificial intelligence","author":"The White House","year":"2025"},{"key":"pcbi.1012975.ref032"},{"key":"pcbi.1012975.ref033"},{"key":"pcbi.1012975.ref034","article-title":"Framework for guiding funding decisions about proposed research involving enhanced potential pandemic pathogens (P3CO)","year":"2022"},{"key":"pcbi.1012975.ref035","article-title":"Biotechnology research in an age of terrorism","volume-title":"Biotechnology research in an age of terrorism [Internet","year":"2004"},{"key":"pcbi.1012975.ref036","volume-title":"Biodefense in the Age of Synthetic Biology [Internet","year":"2018"},{"key":"pcbi.1012975.ref037","volume-title":"National Academies of Sciences E. Governance of dual use research in the life sciences: advancing global consensus on research oversight: proceedings of a workshop [Internet","year":"2018"},{"key":"pcbi.1012975.ref038"},{"issue":"6069","key":"pcbi.1012975.ref039","doi-asserted-by":"crossref","first-page":"660","DOI":"10.1126\/science.1217994","article-title":"Public health and biosecurity. Adaptations of avian flu virus are a cause for concern","volume":"335","author":"K Berns","year":"2012","journal-title":"Science"},{"issue":"1","key":"pcbi.1012975.ref040","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1038\/s41596-021-00655-6","article-title":"Protocols and risks: when less is more","volume":"17","author":"J Pannu","year":"2022","journal-title":"Nat Protoc"},{"issue":"5","key":"pcbi.1012975.ref041","doi-asserted-by":"crossref","first-page":"975","DOI":"10.1111\/risa.13235","article-title":"Information hazards in biotechnology","volume":"39","author":"G Lewis","year":"2019","journal-title":"Risk Analysis"},{"issue":"4","key":"pcbi.1012975.ref042","doi-asserted-by":"crossref","first-page":"296","DOI":"10.1089\/genbio.2023.0025","article-title":"Generative artificial intelligence-assisted protein design must consider repurposing potential","volume":"2","author":"S Ekins","year":"2023","journal-title":"GEN Biotechnol"},{"issue":"4","key":"pcbi.1012975.ref043","doi-asserted-by":"crossref","first-page":"389","DOI":"10.1038\/s41592-021-01100-y","article-title":"Low-N protein engineering with data-efficient deep learning","volume":"18","author":"S Biswas","year":"2021","journal-title":"Nat Methods"},{"issue":"6615","key":"pcbi.1012975.ref044","doi-asserted-by":"crossref","first-page":"49","DOI":"10.1126\/science.add2187","article-title":"Robust deep learning-based protein sequence design using ProteinMPNN","volume":"378","author":"J Dauparas","year":"2022","journal-title":"Science"},{"issue":"7989","key":"pcbi.1012975.ref045","doi-asserted-by":"crossref","first-page":"1070","DOI":"10.1038\/s41586-023-06728-8","article-title":"Illuminating protein space with a programmable generative model","volume":"623","author":"JB Ingraham","year":"2023","journal-title":"Nature"},{"issue":"6","key":"pcbi.1012975.ref046","doi-asserted-by":"crossref","first-page":"691","DOI":"10.1038\/s41587-020-00793-4","article-title":"Deep diversification of an AAV capsid protein by machine learning","volume":"39","author":"D Bryant","year":"2021","journal-title":"Nat Biotechnol"},{"key":"pcbi.1012975.ref047","year":"2023"},{"key":"pcbi.1012975.ref048"},{"key":"pcbi.1012975.ref049"},{"issue":"6","key":"pcbi.1012975.ref050","doi-asserted-by":"crossref","first-page":"521","DOI":"10.1089\/hs.2022.0109","article-title":"Beyond biosecurity by taxonomic lists: lessons, challenges, and opportunities","volume":"21","author":"P Millett","year":"2023","journal-title":"Health Secur"},{"issue":"6625","key":"pcbi.1012975.ref051","doi-asserted-by":"crossref","first-page":"1170","DOI":"10.1126\/science.adf6020","article-title":"Strengthen oversight of risky research on pathogens","volume":"378","author":"J Pannu","year":"2022","journal-title":"Science"},{"key":"pcbi.1012975.ref052","volume-title":"Globalization, biosecurity, and the future of the life sciences [Internet]","year":"2006"},{"key":"pcbi.1012975.ref053","unstructured":"Letts K. Proposed biosecurity oversight framework for the future of science. 2023."},{"issue":"11","key":"pcbi.1012975.ref054","doi-asserted-by":"crossref","first-page":"901","DOI":"10.1136\/medethics-2014-102619","article-title":"The ethics of biosafety considerations in gain-of-function research resulting in the creation of potential pandemic pathogens","volume":"41","author":"NG Evans","year":"2015","journal-title":"J Med Ethics"},{"issue":"7984","key":"pcbi.1012975.ref055","doi-asserted-by":"crossref","first-page":"818","DOI":"10.1038\/s41586-023-06617-0","article-title":"Learning from prepandemic data to forecast viral escape","volume":"622","author":"NN Thadani","year":"2023","journal-title":"Nature"},{"issue":"3","key":"pcbi.1012975.ref056","doi-asserted-by":"crossref","first-page":"189","DOI":"10.1038\/s42256-022-00465-9","article-title":"Dual use of artificial intelligence-powered drug discovery","volume":"4","author":"F Urbina","year":"2022","journal-title":"Nat Mach Intell"},{"key":"pcbi.1012975.ref057","author":"M Guo","year":"2024"},{"key":"pcbi.1012975.ref058"},{"key":"pcbi.1012975.ref059","author":"N Li","year":"2024"},{"key":"pcbi.1012975.ref060"},{"key":"pcbi.1012975.ref061"},{"key":"pcbi.1012975.ref062"},{"key":"pcbi.1012975.ref063"},{"key":"pcbi.1012975.ref064"},{"key":"pcbi.1012975.ref065"},{"key":"pcbi.1012975.ref066","year":"2023"}],"container-title":["PLOS Computational Biology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1012975","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,5,8]],"date-time":"2025-05-08T17:39:01Z","timestamp":1746725941000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1012975"}},"subtitle":[],"editor":[{"given":"Jason A.","family":"Papin","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2025,5,8]]},"references-count":66,"journal-issue":{"issue":"5","published-online":{"date-parts":[[2025,5,8]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pcbi.1012975","relation":{},"ISSN":["1553-7358"],"issn-type":[{"value":"1553-7358","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,5,8]]}}}