{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,11,16]],"date-time":"2026-11-16T20:31:17Z","timestamp":1794861077254,"version":"4.9.1"},"reference-count":93,"publisher":"Oxford University Press (OUP)","issue":"1","license":[{"start":{"date-parts":[[2024,5,29]],"date-time":"2024-05-29T00:00:00Z","timestamp":1716940800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,1,4]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>This article examines the challenges of regulating artificial intelligence (AI) systems and proposes an adapted model of regulation suitable for AI's novel features. Unlike past technologies, AI systems built using techniques like deep learning cannot be directly analyzed, specified, or audited against regulations. Their behavior emerges unpredictably from training rather than intentional design. However, the traditional model of delegating oversight to an expert agency, which has succeeded in high-risk sectors like aviation and nuclear power, should not be wholly discarded. Instead, policymakers must contain risks from today's opaque models while supporting research into provably safe AI architectures. Drawing lessons from AI safety literature and past regulatory successes, effective AI governance will likely require consolidated authority, licensing regimes, mandated training data and modeling disclosures, formal verification of system behavior, and the capacity for rapid intervention.<\/jats:p>","DOI":"10.1093\/polsoc\/puae020","type":"journal-article","created":{"date-parts":[[2024,5,16]],"date-time":"2024-05-16T00:52:06Z","timestamp":1715820726000},"page":"85-97","source":"Crossref","is-referenced-by-count":64,"title":["When code isn\u2019t law: rethinking regulation for artificial intelligence"],"prefix":"10.1093","volume":"44","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-6970-698X","authenticated-orcid":false,"given":"Brian","family":"Judge","sequence":"first","affiliation":[{"name":"Center for Human-Compatible AI, University of California\u2014Berkeley , Berkeley, CA,","place":["United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mark","family":"Nitzberg","sequence":"additional","affiliation":[{"name":"Center for Human-Compatible AI, University of California\u2014Berkeley , Berkeley, CA,","place":["United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Stuart","family":"Russell","sequence":"additional","affiliation":[{"name":"Center for Human-Compatible AI, University of California\u2014Berkeley , Berkeley, CA,","place":["United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"286","published-online":{"date-parts":[[2024,5,29]]},"reference":[{"key":"2025102108212253700_R1","volume-title":"The Oxford handbook of AI governance","author":"Acemo\u011flu","year":"2022","edition":"online"},{"key":"2025102108212253700_R2","volume-title":"Power and progress: Our thousand-year struggle over technology and prosperity","author":"Acemo\u011flu","year":"2023"},{"key":"2025102108212253700_R3","article-title":"The Bletchley Declaration","author":"AI Safety Summit","year":"2023"},{"key":"2025102108212253700_R4","article-title":"Concrete problems in AI safety","author":"Amodei","year":"2016"},{"key":"2025102108212253700_R5","article-title":"Frontier AI regulation: Managing emerging risks to public safety","author":"Anderljung","year":"2023"},{"key":"2025102108212253700_R6","doi-asserted-by":"publisher","DOI":"10.1016\/j.chb.2020.106572","article-title":"The importance of assurance that \u2018humans are still in the decision loop\u2019 for public trust in artificial intelligence: Evidence from an online experiment","volume":"114","author":"Aoki","year":"2021","journal-title":"Computers in Human Behavior"},{"issue":"2","key":"2025102108212253700_R7","doi-asserted-by":"publisher","DOI":"10.1111\/phc3.12964","article-title":"Artificial intelligence: Arguments for catastrophic risk","volume":"19","author":"Bales","year":"2024","journal-title":"Philosophy Compass"},{"key":"2025102108212253700_R8","article-title":"Abu Dhabi makes its Falcon 40B AI model open source","volume-title":"Reuters","author":"Barrington","year":"2023"},{"issue":"3","key":"2025102108212253700_R9","doi-asserted-by":"publisher","first-page":"393","DOI":"10.1111\/gove.12179","article-title":"Instrument constituencies in the policy process","volume":"29","author":"B\u00e9land","year":"2016","journal-title":"Governance"},{"issue":"4","key":"2025102108212253700_R10","doi-asserted-by":"publisher","first-page":"111","DOI":"10.1353\/jod.2023.a907692","article-title":"AI and catastrophic risk","volume":"34","author":"Bengio","year":"2023","journal-title":"Journal of Democracy"},{"key":"2025102108212253700_R11","article-title":"Existential risks: Analyzing human extinction scenarios and related hazards","volume":"9","author":"Bostrom","year":"2002","journal-title":"Journal of Evolution and Technology"},{"key":"2025102108212253700_R12","volume-title":"Superintelligence: Paths, dangers, strategies","author":"Bostrom","year":"2014"},{"key":"2025102108212253700_R13","volume-title":"Artificial intelligence safety and security","author":"Bostrom","year":"2018"},{"issue":"1","key":"2025102108212253700_R14","doi-asserted-by":"publisher","first-page":"83","DOI":"10.1016\/0304-4076(94)01598-T","article-title":"General purpose technologies: \u2018Engines of growth\u2019?","volume":"65","author":"Bresnahan","year":"1995","journal-title":"Journal of Econometrics"},{"issue":"1","key":"2025102108212253700_R15","doi-asserted-by":"publisher","first-page":"213","DOI":"10.1146\/annurev-soc-090820-020800","article-title":"The society of algorithms","volume":"47","author":"Burrell","year":"2021","journal-title":"Annual Review of Sociology"},{"issue":"11","key":"2025102108212253700_R16","doi-asserted-by":"publisher","first-page":"1721","DOI":"10.1080\/13501763.2022.2126515","article-title":"Governing AI\u2014Attempting to herd cats?","volume":"29","author":"B\u00fcthe","year":"2022","journal-title":"Journal of European Public Policy"},{"issue":"2","key":"2025102108212253700_R17","doi-asserted-by":"publisher","first-page":"141","DOI":"10.1177\/0952076719827068","article-title":"Causal logics and mechanisms in policy design: How and why adopting a mechanistic perspective can improve policy design","volume":"36","author":"Capano","year":"2021","journal-title":"Public Policy and Administration"},{"issue":"4","key":"2025102108212253700_R18","doi-asserted-by":"publisher","first-page":"311","DOI":"10.1080\/13876988.2015.1031977","article-title":"Bringing governments back in: Governance and governing in comparative policy analysis","volume":"17","author":"Capano","year":"2015","journal-title":"Journal of Comparative Policy Analysis"},{"key":"2025102108212253700_R19","article-title":"Open problems and fundamental limitations of reinforcement learning from human feedback","author":"Casper","year":"2023"},{"key":"2025102108212253700_R20","doi-asserted-by":"crossref","DOI":"10.1145\/3630106.3659037","article-title":"Black-box access is insufficient for rigorous AI audits","author":"Casper","year":"2024"},{"issue":"2133","key":"2025102108212253700_R21","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1098\/rsta.2018.0080","article-title":"Governing artificial intelligence: Ethical, legal and technical opportunities and challenges","volume":"376","author":"Cath","year":"2018","journal-title":"Philosophical Transactions"},{"issue":"1","key":"2025102108212253700_R22","doi-asserted-by":"publisher","first-page":"5","DOI":"10.1038\/s42256-018-0003-2","article-title":"Bridging near-and long-term concerns about AI","volume":"1","author":"Cave","year":"2019","journal-title":"Nature Machine Intelligence"},{"key":"2025102108212253700_R23","volume-title":"The alignment problem: Machine learning and human values","author":"Christian","year":"2020"},{"issue":"6691","key":"2025102108212253700_R24","doi-asserted-by":"publisher","first-page":"36","DOI":"10.1126\/science.adl0625","article-title":"Regulating advanced artificial agents","volume":"384","author":"Cohen","year":"2024","journal-title":"Science"},{"issue":"4","key":"2025102108212253700_R25","doi-asserted-by":"publisher","first-page":"307","DOI":"10.1016\/j.polsoc.2014.10.001","article-title":"Making sense of governance","volume":"33","author":"Colebatch","year":"2014","journal-title":"Policy and Society"},{"key":"2025102108212253700_R26","article-title":"Towards automated circuit discovery for mechanistic interpretability","author":"Conmy","year":"2023"},{"key":"2025102108212253700_R27","article-title":"AI research considerations for human existential safety","author":"Critch","year":"2020"},{"issue":"6","key":"2025102108212253700_R28","doi-asserted-by":"publisher","first-page":"833","DOI":"10.1080\/02723638.2020.1746096","article-title":"The transition to autonomous cars, the redesign of cities and the future of urban sustainability","volume":"42","author":"Cugurullo","year":"2021","journal-title":"Urban Geography"},{"key":"2025102108212253700_R29","volume-title":"The Oxford handbook of AI governance","author":"Dafoe","year":"2022","edition":"online"},{"key":"2025102108212253700_R30","article-title":"Towards Guaranteed Safe AI: A Framework for Enuring Robust and Reliable AI Systems","author":"Dalrymple","year":"2024"},{"issue":"2","key":"2025102108212253700_R31","doi-asserted-by":"publisher","DOI":"10.1016\/j.giq.2021.101666","article-title":"The perils and pitfalls of explainable AI: Strategies for explaining algorithmic decision-making","volume":"39","author":"de Bruijn","year":"2022","journal-title":"Government Information Quarterly"},{"issue":"3","key":"2025102108212253700_R32","doi-asserted-by":"publisher","first-page":"377","DOI":"10.1017\/eis.2023.1","article-title":"Engines of power: Electricity, AI, and general-purpose military transformations","volume":"8","author":"Ding","year":"2023","journal-title":"European Journal of International Security"},{"key":"2025102108212253700_R33","article-title":"GPTs are GPTs: An early look at the labor market impact potential of large language models","author":"Eloundou","year":"2023"},{"issue":"2","key":"2025102108212253700_R34","doi-asserted-by":"publisher","first-page":"421","DOI":"10.1177\/00323217221126665","article-title":"Artificial intelligence and the political legitimacy of global governance","volume":"72","author":"Erman","year":"2022","journal-title":"Political Studies"},{"key":"2025102108212253700_R35","article-title":"Pause giant AI experiments: An open letter","author":"Future of Life Institute","year":"2023"},{"issue":"3","key":"2025102108212253700_R36","doi-asserted-by":"publisher","first-page":"411","DOI":"10.1007\/s11023-020-09539-2","article-title":"Artificial intelligence, values, and alignment","volume":"30","author":"Gabriel","year":"2020","journal-title":"Minds and Machines."},{"issue":"2","key":"2025102108212253700_R37","doi-asserted-by":"publisher","first-page":"218","DOI":"10.1162\/daed_a_01911","article-title":"Toward a theory of justice for artificial intelligence","volume":"151","author":"Gabriel","year":"2022","journal-title":"Daedalus"},{"key":"2025102108212253700_R38","doi-asserted-by":"crossref","DOI":"10.1145\/3531146.3533229","article-title":"Predictability and surprise in large generative models","author":"Ganguli","year":"2022"},{"key":"2025102108212253700_R39","article-title":"It\u2019s time to create a national registry for large AI models","author":"Hadfield","year":"2023"},{"key":"2025102108212253700_R40","article-title":"Cooperative inverse reinforcement learning","author":"Hadfield-Menell","year":"2016"},{"key":"2025102108212253700_R41","article-title":"Open-source AI is uniquely dangerous","volume-title":"IEEE Spectrum","author":"Harris","year":"2024"},{"key":"2025102108212253700_R42","article-title":"Unresolved problems in ML safety","author":"Hendrycks","year":"2022"},{"key":"2025102108212253700_R43","volume-title":"Designing public policies: Principles and instruments","author":"Howlett","year":"2019"},{"issue":"6","key":"2025102108212253700_R44","doi-asserted-by":"publisher","first-page":"487","DOI":"10.1080\/13876988.2020.1774367","article-title":"Policy-makers, policy-takers and policy tools: Dealing with behavioural issues in policy design","volume":"22","author":"Howlett","year":"2020","journal-title":"Journal of Comparative Policy Analysis"},{"key":"2025102108212253700_R45","volume-title":"NIST\u2019s AI risk management framework plants a flag in the AI debate","author":"Kerry","year":"2023"},{"issue":"1","key":"2025102108212253700_R46","doi-asserted-by":"publisher","first-page":"52","DOI":"10.1093\/polsoc\/puae012","article-title":"Why and how is the power of Big Tech increasing in the policy process? The case of generative AI","volume":"44","author":"Khanal","year":"2024","journal-title":"Policy and Society"},{"key":"2025102108212253700_R47","volume-title":"Machine learning and knowledge extraction. CD-MAKE 2023. Lecture notes in computer science","author":"Kieseberg","year":"2023"},{"key":"2025102108212253700_R48","article-title":"Specification gaming: The flip side of AI ingenuity","volume-title":"DeepMind","author":"Krakovna","year":"2020"},{"issue":"1","key":"2025102108212253700_R49","doi-asserted-by":"publisher","first-page":"3","DOI":"10.1111\/rego.12512","article-title":"Trustworthy artificial intelligence and the European Union AI act: On the conflation of trustworthiness and acceptability of risk","volume":"18","author":"Laux","year":"2024","journal-title":"Regulation & Governance"},{"key":"2025102108212253700_R50","doi-asserted-by":"crossref","DOI":"10.1145\/3600211.3604701","article-title":"The bureaucratic challenge to AI governance: An empirical assessment of implementation at U.S. federal agencies","author":"Lawrence","year":"2023"},{"key":"2025102108212253700_R51","volume-title":"Code and other laws of cyberspace","author":"Lessig","year":"1999"},{"key":"2025102108212253700_R52","article-title":"Code is law: On liberty in cyberspace","volume-title":"Harvard Magazine","author":"Lessig","year":"2000"},{"key":"2025102108212253700_R53","volume-title":"Controlling the atom: The beginnings of nuclear regulation, 1946\u20131962","author":"Mazuzan","year":"1984"},{"key":"2025102108212253700_R54","article-title":"UK to invest \u00a3900m in supercomputer in bid to build own \u2018BritGPT\u2019","volume-title":"The Guardian","author":"Milmo","year":"2023"},{"key":"2025102108212253700_R55","article-title":"12 Tentative ideas for AI policy","author":"Muehlhauser","year":"2023"},{"key":"2025102108212253700_R56","article-title":"Repurposing the wheel: Lessons for AI standards","author":"Narayanan","year":"2023"},{"issue":"2","key":"2025102108212253700_R57","doi-asserted-by":"publisher","first-page":"367","DOI":"10.1007\/s00146-019-00887-x","article-title":"The race for an artificial general intelligence: Implications for public policy","volume":"35","author":"Naud\u00e9","year":"2020","journal-title":"AI and Society"},{"key":"2025102108212253700_R58","article-title":"The alignment problem from a deep learning perspective","author":"Ngo","year":"2023"},{"key":"2025102108212253700_R59","article-title":"GPT-4 technical report","author":"OpenAI","year":"2023"},{"key":"2025102108212253700_R60","article-title":"Language models can explain neurons in language models","author":"OpenAI","year":"2023"},{"key":"2025102108212253700_R61","volume-title":"Reforms at risk: What happens after major policy changes are enacted","author":"Patashnik","year":"2014"},{"key":"2025102108212253700_R62","doi-asserted-by":"publisher","DOI":"10.1111\/rego.12563","article-title":"European artificial intelligence \u201ctrusted throughout the world\u201d: Risk-based regulation and the fashioning of a competitive common AI market","author":"Paul","year":"2023","journal-title":"Regulation & Governance"},{"key":"2025102108212253700_R63","article-title":"Exploiting novel GPT-4 APIs","volume-title":"Far","author":"Pelrine","year":"2023"},{"key":"2025102108212253700_R64","article-title":"The U.S. regulates cars, radio and TV. When will it regulate AI?","volume-title":"New York Times","author":"Philbrick","year":"2023"},{"issue":"4","key":"2025102108212253700_R65","doi-asserted-by":"publisher","first-page":"595","DOI":"10.2307\/2950710","article-title":"When effect becomes cause: Policy feedback and political change","volume":"45","author":"Pierson","year":"1993","journal-title":"World Politics"},{"issue":"2","key":"2025102108212253700_R66","doi-asserted-by":"publisher","first-page":"251","DOI":"10.2307\/2586011","article-title":"Increasing returns, path dependence, and the study of politics","volume":"94","author":"Pierson","year":"2000","journal-title":"American Political Science Review"},{"issue":"2","key":"2025102108212253700_R67","doi-asserted-by":"publisher","first-page":"178","DOI":"10.1080\/14494035.2021.1929728","article-title":"Steering the governance of artificial intelligence: National strategies in perspective","volume":"40","author":"Radu","year":"2021","journal-title":"Policy and Society"},{"key":"2025102108212253700_R68","article-title":"Bing\u2019s AI chat: \u201cI want to be alive\u201d","volume-title":"New York Times","author":"Roose","year":"2023"},{"key":"2025102108212253700_R69","doi-asserted-by":"publisher","DOI":"10.51593\/20190041","article-title":"Key concepts in AI safety: Robustness and adversarial examples","author":"Rudner","year":"2021","journal-title":"CSET Issue Brief"},{"key":"2025102108212253700_R70","volume-title":"Human compatible: Artificial intelligence and the problem of control","author":"Russell","year":"2019"},{"issue":"3","key":"2025102108212253700_R71","first-page":"60","article-title":"Banning lethal autonomous weapons: An education","volume":"38","author":"Russell","year":"2022","journal-title":"Issues in Science and Technology"},{"key":"2025102108212253700_R72","article-title":"Written testimony. Hearing before the Committee on the Judiciary, United States Senate","author":"Russell","year":"2023"},{"key":"2025102108212253700_R73","article-title":"Biden\u2019s elusive AI whisperer finally goes on the record","volume-title":"Politico","author":"Scola","year":"2023"},{"issue":"4","key":"2025102108212253700_R74","doi-asserted-by":"publisher","first-page":"478","DOI":"10.1093\/polsoc\/puad027","article-title":"The role of policy design in policy continuation and ratcheting-up of policy ambition","volume":"42","author":"Sewerin","year":"2023","journal-title":"Policy and Society"},{"issue":"1","key":"2025102108212253700_R75","doi-asserted-by":"publisher","first-page":"57","DOI":"10.1080\/17579961.2021.1898300","article-title":"From a \u2018race to AI\u2019 to a \u2018race to AI regulation\u2019: Regulatory competition for artificial intelligence","volume":"13","author":"Smuha","year":"2021","journal-title":"Law, Innovation and Technology"},{"key":"2025102108212253700_R76","article-title":"Safe before sale: Learnings from the FDA\u2019s model of life sciences oversight for foundation models","author":"Stein","year":"2023"},{"issue":"1","key":"2025102108212253700_R77","doi-asserted-by":"publisher","first-page":"3","DOI":"10.2307\/3003160","article-title":"The theory of economic regulation","volume":"2","author":"Stigler","year":"1971","journal-title":"The Bell Journal of Economics and Management Science"},{"issue":"2","key":"2025102108212253700_R78","doi-asserted-by":"publisher","first-page":"137","DOI":"10.1080\/14494035.2021.1928377","article-title":"Governance of artificial intelligence","volume":"40","author":"Taeihagh","year":"2021","journal-title":"Policy and Society"},{"issue":"4","key":"2025102108212253700_R79","doi-asserted-by":"publisher","first-page":"1009","DOI":"10.1111\/rego.12392","article-title":"Assessing the regulatory challenges of emerging disruptive technologies","volume":"15","author":"Taeihagh","year":"2021","journal-title":"Regulation & Governance"},{"key":"2025102108212253700_R80","article-title":"Provably safe systems: The only path to controllable AGI","author":"Tegmark","year":"2023"},{"key":"2025102108212253700_R81","doi-asserted-by":"crossref","DOI":"10.2139\/ssrn.4579899","article-title":"International governance of civilian AI: A jurisdictional certification approach","author":"Trager","year":"2023"},{"key":"2025102108212253700_R82","volume-title":"The Essential Turing","author":"Turing","year":"1951"},{"issue":"2","key":"2025102108212253700_R83","doi-asserted-by":"publisher","first-page":"256","DOI":"10.1162\/daed_a_01914","article-title":"Automation, AI & work","volume":"151","author":"Tyson","year":"2022","journal-title":"Daedalus"},{"issue":"5","key":"2025102108212253700_R84","doi-asserted-by":"publisher","first-page":"612","DOI":"10.1111\/ropr.12574","article-title":"Politics and policy of artificial intelligence","volume":"40","author":"Ulnicane","year":"2023","journal-title":"Review of Policy Research"},{"key":"2025102108212253700_R85","article-title":"A review of formal methods applied to machine learning","author":"Urban","year":"2021"},{"key":"2025102108212253700_R86","article-title":"Market concentration implications of foundation models","volume-title":"Brookings Center on Regulations and Markets Working Paper #9","author":"Vipra","year":"2023"},{"key":"2025102108212253700_R87","doi-asserted-by":"crossref","DOI":"10.23915\/distill.00024.007","article-title":"Visualizing weights","author":"Voss","year":"2021"},{"key":"2025102108212253700_R88","volume-title":"A short history of nuclear regulation, 1946\u20132009","author":"Walker","year":"2010"},{"key":"2025102108212253700_R89","doi-asserted-by":"publisher","first-page":"1355","DOI":"10.1126\/science.131.3410.1355","article-title":"Some moral and technical consequences of automation","volume":"131","author":"Wiener","year":"1960","journal-title":"Science"},{"key":"2025102108212253700_R90","article-title":"Fundamental limitations of alignment in large language models","author":"Wolf","year":"2023"},{"key":"2025102108212253700_R91","doi-asserted-by":"crossref","DOI":"10.1007\/978-3-319-54675-9","volume-title":"Policy capacity and governance: Assessing governmental competences and capabilities in theory and practice","author":"Wu","year":"2018"},{"issue":"3-4","key":"2025102108212253700_R92","doi-asserted-by":"publisher","first-page":"165","DOI":"10.1016\/j.polsoc.2015.09.001","article-title":"Policy capacity: A conceptual framework for understanding policy competences and capabilities","volume":"34","author":"Wu","year":"2015","journal-title":"Policy and Society"},{"key":"2025102108212253700_R93","article-title":"New research initiative aims to build large language AI model for Southeast Asia","volume-title":"ZDNet","author":"Yu","year":"2023"}],"container-title":["Policy and Society"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/policyandsociety\/article-pdf\/44\/1\/85\/57986099\/puae020.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/policyandsociety\/article-pdf\/44\/1\/85\/57986099\/puae020.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,21]],"date-time":"2025-10-21T12:21:39Z","timestamp":1761049299000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/policyandsociety\/article\/44\/1\/85\/7684910"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,5,29]]},"references-count":93,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2024,5,29]]},"published-print":{"date-parts":[[2025,1,4]]}},"URL":"https:\/\/doi.org\/10.1093\/polsoc\/puae020","relation":{},"ISSN":["1449-4035","1839-3373"],"issn-type":[{"value":"1449-4035","type":"print"},{"value":"1839-3373","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2025,1]]},"published":{"date-parts":[[2024,5,29]]}}}