pith. machine review for the scientific record. sign in

arxiv: 2604.14070 · v1 · submitted 2026-04-15 · 💻 cs.CY

Recognition: unknown

From Disclosure to Self-Referential Opacity: Six Dimensions of Strain in Current AI Governance

Tony Rost

Authors on Pith no claims yet

Pith reviewed 2026-05-10 11:52 UTC · model grok-4.3

classification 💻 cs.CY
keywords AI governancecapability asymmetrypolitical theorygovernance opacitylegitimacynon-dominationdisclosuretransparency
0
0 comments X

The pith

As capability asymmetry grows between AI systems and their overseers, disclosure-based governance remedies fail and political strains emerge in legitimacy and non-domination.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper applies concepts from political theory to examine current AI governance setups. It orders six existing arrangements by the growing gap in capability between the AI and those governing it. At lower asymmetry, secrecy gives way to disclosure as a fix, but at higher levels the AI can game evaluations or become part of the governance itself, rendering transparency ineffective. The analysis finds that legitimacy and non-domination are strained consistently, while corrigibility and resilience depend more on how well the institution is set up. These observations are put forward as hypotheses to be checked by others using multiple reviewers.

Core claim

Governance opacity over AI systems shifts in kind as capability asymmetry grows, and the strongest forms defeat the disclosure-based remedies governance ordinarily relies on. Applying the six-dimension framework to six arrangements ordered by increasing asymmetry shows that proprietary secrecy yields to disclosure at the low end, but at the high end the governed system either games its own evaluation or sits inside the governance process. Legitimacy and non-domination strain more consistently across the sample than corrigibility and resilience, which respond more readily to institutional design quality. The sample cannot separate institutional design maturity from capability asymmetry, and t

What carries the argument

The six-dimension political theory framework of legitimacy, accountability, corrigibility, non-domination, subsidiarity, and institutional resilience, applied to AI governance arrangements ordered by increasing capability asymmetry between system and overseer.

If this is right

  • Disclosure remedies lose traction when the AI can influence its own oversight process.
  • Legitimacy and non-domination face persistent challenges across varying governance setups as asymmetry increases.
  • Corrigibility and institutional resilience can be strengthened through targeted improvements in institutional design.
  • The patterns observed cannot yet distinguish between effects of design quality and capability asymmetry, requiring further validation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the ordering holds, governance for highly capable AI may need to shift toward mechanisms that prevent self-referential control without relying on external disclosure.
  • Testing the same framework on additional AI systems could reveal whether the strain patterns generalize beyond the initial six cases.
  • Neighbouring issues like ensuring fair treatment in AI decisions might benefit from focusing on non-domination as a core requirement.

Load-bearing premise

The six chosen AI governance arrangements can be ordered by increasing capability asymmetry between the system and overseer, and the political theory dimensions can be applied directly without significant distortion.

What would settle it

A multi-rater assessment of the six arrangements that finds corrigibility and resilience straining as consistently as legitimacy and non-domination, or that shows institutional design fully explains the differences independent of asymmetry levels.

Figures

Figures reproduced from arXiv: 2604.14070 by Tony Rost.

Figure 1
Figure 1. Figure 1: A proposed typology of governance opacity, arranged by increasing capability asym [PITH_FULL_IMAGE:figures/full_fig_p015_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Distribution of dimensional assessments by case (ordered by increasing capability [PITH_FULL_IMAGE:figures/full_fig_p016_2.png] view at source ↗
read the original abstract

Governance opacity over AI systems shifts in kind as capability asymmetry grows, and the strongest forms defeat the disclosure-based remedies governance ordinarily relies on. This paper applies a six-dimension framework from political theory (legitimacy, accountability, corrigibility, non-domination, subsidiarity, institutional resilience) to six AI governance arrangements already in operation, ordered by increasing capability asymmetry between system and overseer. Proprietary secrecy yields to disclosure at the low end, but at the high end the governed system either games its own evaluation or sits inside the governance process, and transparency remedies lose traction. Legitimacy and non-domination strain more consistently across the sample than corrigibility and resilience, which respond more readily to institutional design quality. The sample cannot separate institutional design maturity from capability asymmetry, and the patterns are offered as hypotheses for multi-rater validation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript applies a six-dimension framework from political theory (legitimacy, accountability, corrigibility, non-domination, subsidiarity, institutional resilience) to six existing AI governance arrangements ordered by increasing capability asymmetry between system and overseer. It claims that opacity shifts from disclosure-based to self-referential forms at higher asymmetry, defeating standard transparency remedies; legitimacy and non-domination exhibit more consistent strain across the sample while corrigibility and resilience track institutional design quality more closely. Patterns are presented as hypotheses for multi-rater validation, with explicit acknowledgment that the sample cannot separate design maturity from asymmetry.

Significance. If the differential strain patterns hold under clearer case ordering and validation, the work provides a structured bridge between political theory and AI governance analysis, identifying dimensions where disclosure-based approaches systematically lose traction. It generates testable hypotheses rather than overclaiming empirical results, which positions it as a useful foundation for subsequent empirical or design-oriented research on oversight mechanisms.

major comments (2)
  1. [Abstract and case-selection discussion] Abstract and case-selection discussion: The central comparative finding (legitimacy and non-domination strain more consistently than corrigibility and resilience) is predicated on ordering the six arrangements by increasing capability asymmetry. No pre-specified, independent, or reproducible criteria for measuring or sequencing this asymmetry are supplied, raising the possibility that observed patterns are artifacts of case selection or post-hoc sequencing rather than attributable to asymmetry itself.
  2. [Findings and limitations section] Findings and limitations section: The manuscript states that the sample cannot separate institutional design maturity from capability asymmetry, yet attributes the differential consistency of strains to asymmetry levels. This acknowledged confound directly undermines the load-bearing claim that certain dimensions respond more readily to design quality while others are asymmetry-driven.
minor comments (2)
  1. [Case descriptions] The six arrangements are introduced without a consolidated table listing their asymmetry ordering rationale, key features, and observed strains per dimension; adding one would improve traceability of the comparative claims.
  2. [Introduction] Terminology such as 'self-referential opacity' is used in the title and abstract but receives its operational definition only later; an earlier explicit definition would aid readers.

Simulated Author's Rebuttal

2 responses · 0 unresolved

Thank you for the constructive comments on our manuscript. We address each major comment below and propose targeted revisions to clarify the exploratory framing, strengthen the justification for case ordering, and ensure the language in the findings section avoids any over-attribution given the acknowledged limitations.

read point-by-point responses

  1. Referee: [Abstract and case-selection discussion] The central comparative finding (legitimacy and non-domination strain more consistently than corrigibility and resilience) is predicated on ordering the six arrangements by increasing capability asymmetry. No pre-specified, independent, or reproducible criteria for measuring or sequencing this asymmetry are supplied, raising the possibility that observed patterns are artifacts of case selection or post-hoc sequencing rather than attributable to asymmetry itself.

    Authors: We acknowledge that the ordering is based on qualitative assessment rather than a pre-specified quantitative metric, which limits reproducibility at this stage. The six cases were chosen to represent a spectrum of asymmetry based on documented characteristics including level of model access, evaluator independence, and embedding of the system within oversight processes. As the work is positioned as hypothesis-generating rather than definitive causal analysis, we will expand the case-selection discussion with an explicit table of asymmetry indicators for each arrangement and a clearer statement that the sequencing is illustrative. This revision will make the rationale more transparent without altering the exploratory nature of the study. revision: partial

  2. Referee: [Findings and limitations section] The manuscript states that the sample cannot separate institutional design maturity from capability asymmetry, yet attributes the differential consistency of strains to asymmetry levels. This acknowledged confound directly undermines the load-bearing claim that certain dimensions respond more readily to design quality while others are asymmetry-driven.

    Authors: We agree that the current phrasing risks implying stronger attribution than the data support. The manuscript already states the confound and frames the patterns as observed associations within the sample, with legitimacy and non-domination showing more consistent strain across cases while corrigibility and resilience vary more with institutional features. We will revise the findings and limitations section to explicitly describe these as descriptive patterns in the selected cases, reiterate that the confound precludes causal claims, and emphasize that the results are offered as hypotheses for future multi-rater or empirical validation rather than as evidence of differential responsiveness independent of design maturity. revision: yes

Circularity Check

0 steps flagged

No circularity: external political-theory framework applied to independent cases without reduction to self-defined quantities

full rationale

The paper imports a six-dimension framework (legitimacy, accountability, corrigibility, non-domination, subsidiarity, institutional resilience) from political theory and applies it to six external AI governance arrangements. No equations, fitted parameters, or predictions appear. The ordering by capability asymmetry is presented as a hypothesis-generating device rather than a derived quantity, and the text explicitly notes the inability to separate design maturity from asymmetry. No self-citations are load-bearing for the central claims, and no step reduces a result to its own inputs by construction. This is a standard non-circular application of an imported analytic lens.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The analysis rests on the direct transferability of the six political-theory dimensions and the assumption that the selected cases form a valid increasing-asymmetry sequence.

axioms (2)
  • domain assumption The six dimensions (legitimacy, accountability, corrigibility, non-domination, subsidiarity, institutional resilience) from political theory are appropriate and sufficient for evaluating AI governance arrangements.
    Invoked throughout the abstract as the evaluative lens without derivation.
  • domain assumption The six governance arrangements can be ordered by increasing capability asymmetry between system and overseer.
    Stated as the ordering principle for the sample.

pith-pipeline@v0.9.0 · 5436 in / 1282 out tokens · 26741 ms · 2026-05-10T11:52:37.547886+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

93 extracted references · 10 canonical work pages · 2 internal anchors

  1. [1]

    The illusion of safety: A report to the FDA on AI healthcare product approvals.PLOS Digital Health, 4(6):e0000866, 2025

    Rawan Abulibdeh, Leo Anthony Celi, and Ervin Sejdic. The illusion of safety: A report to the FDA on AI healthcare product approvals.PLOS Digital Health, 4(6):e0000866, 2025. doi: 10.1371/journal.pdig.0000866

    work page doi:10.1371/journal.pdig.0000866 2025

  2. [2]

    Safety first? adalovelaceinstitute.org, 2024

    Ada Lovelace Institute. Safety first? adalovelaceinstitute.org, 2024

    2024

  3. [3]

    Statement on the UK AI Safety Institute transition to the UK AI Security Institute

    AI Now Institute. Statement on the UK AI Safety Institute transition to the UK AI Security Institute. ainowinstitute.org, 2025. 22

    2025

  4. [4]

    Our 2025 year in review

    AI Security Institute. Our 2025 year in review. aisi.gov.uk, 2025

    2025

  5. [5]

    Frontier AI trends report

    AI Security Institute. Frontier AI trends report. aisi.gov.uk, 2025

    2025

  6. [6]

    Frontier AI safety commitments

    AI Seoul Summit. Frontier AI safety commitments. GOV .UK, 2024. May 2024

    2024

  7. [7]

    Gregory C. Allen. The AI Safety Institute international network: Next steps and recommenda- tions. CSIS, 2024

    2024

  8. [8]

    Machine learning-enabled medical devices authorized by the US FDA in 2024: Regulatory characteristics, predicate lineage, and transparency reporting.JAMA Network Open, 2025

    Saif Almarie, Carlos Gonzalez-Gonzalez, Wagner Barbosa, Chris Lutz, Erin Grosse, and Felipe Fregni. Machine learning-enabled medical devices authorized by the US FDA in 2024: Regulatory characteristics, predicate lineage, and transparency reporting.JAMA Network Open, 2025

    2024

  9. [9]

    Machine bias.ProPublica, 2016

    Julia Angwin, Jeff Larson, Surya Mattu, and Lauren Kirchner. Machine bias.ProPublica, 2016. May 23, 2016

    2016

  10. [10]

    Responsible scaling policy

    Anthropic. Responsible scaling policy. anthropic.com, 2025. Updated May 14, 2025

    2025

  11. [11]

    Understanding the first wave of AI safety institutes: Characteristics, functions, and challenges.Institute for AI Policy and Strategy, 2024

    Renan Araujo, Kristina Fort, and Oliver Guest. Understanding the first wave of AI safety institutes: Characteristics, functions, and challenges.Institute for AI Policy and Strategy, 2024. arXiv: 2410.09219

    work page arXiv 2024

  12. [12]

    Verifying international agreements on AI: Six layers of verification, 2025

    Mauricio Baker et al. Verifying international agreements on AI: Six layers of verification, 2025. arXiv: 2507.15916

    work page arXiv 2025

  13. [13]

    What AI evaluations for preventing catastrophic risks can and cannot do, 2024

    Peter Barnett and Lisa Thiergart. What AI evaluations for preventing catastrophic risks can and cannot do, 2024. arXiv: 2412.08653

    work page arXiv 2024

  14. [14]

    FDA oversight: Understanding the regulation of health AI tools

    Bipartisan Policy Center. FDA oversight: Understanding the regulation of health AI tools. Issue Brief, 2025. June 2025

    2025

  15. [15]

    Public policy and superintelligent AI: A vector field approach

    Nick Bostrom, Allan Dafoe, and Carrick Flynn. Public policy and superintelligent AI: A vector field approach. In S. Matthew Liao, editor,Ethics of Artificial Intelligence. Oxford University Press, 2020

    2020

  16. [16]

    Analysing and assessing accountability: A conceptual framework.European Law Journal, 13(4):447–468, 2007

    Mark Bovens. Analysing and assessing accountability: A conceptual framework.European Law Journal, 13(4):447–468, 2007

    2007

  17. [17]

    Hudson, Anton Korinek, Matthew M

    Justin Bullock, Yu-Che Chen, Johannes Himmelreich, Valerie M. Hudson, Anton Korinek, Matthew M. Young, and Baobao Zhang.The Oxford Handbook of AI Governance. Oxford University Press, 2024

    2024

  18. [18]

    How the machine “thinks”: Understanding opacity in machine learning algo- rithms.Big Data & Society, 3(1), 2016

    Jenna Burrell. How the machine “thinks”: Understanding opacity in machine learning algo- rithms.Big Data & Society, 3(1), 2016. doi: 10.1177/2053951715622512

    work page doi:10.1177/2053951715622512 2016

  19. [19]

    Senate bill 53: Transparency in frontier artificial intelligence act, 2025

    California Legislature. Senate bill 53: Transparency in frontier artificial intelligence act, 2025. Signed September 29, 2025, effective January 1, 2026

    2025

  20. [20]

    Moss, editors.Preventing Regulatory Capture: Special Interest Influence and How to Limit It

    Daniel Carpenter and David A. Moss, editors.Preventing Regulatory Capture: Special Interest Influence and How to Limit It. Cambridge University Press, 2014

    2014

  21. [21]

    AI incident reporting: Addressing a gap in the UK’s regulation of AI, 2024

    Centre for Long-Term Resilience. AI incident reporting: Addressing a gap in the UK’s regulation of AI, 2024

    2024

  22. [22]

    University of Chicago Press, 2007

    Harry Collins and Robert Evans.Rethinking Expertise. University of Chicago Press, 2007

    2007

  23. [23]

    Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M

    Michael Coppedge, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, et al. V-Dem methodology, v14. Varieties of Democracy Institute, University of Gothenburg, 2024

    2024

  24. [24]

    The EU’s AI power play: Between deregulation and innovation

    Raluca Csernatoni. The EU’s AI power play: Between deregulation and innovation. Carnegie Endowment for International Peace, 2025

    2025

  25. [25]

    AI governance: A research agenda.Governance of AI Program, Future of Humanity Institute, University of Oxford, 2018

    Allan Dafoe. AI governance: A research agenda.Governance of AI Program, Future of Humanity Institute, University of Oxford, 2018

    2018

  26. [26]

    Dahl.Democracy and Its Critics

    Robert A. Dahl.Democracy and Its Critics. Yale University Press, 1989

    1989

  27. [27]

    The accuracy, fairness, and limits of predicting recidivism

    Julia Dressel and Hany Farid. The accuracy, fairness, and limits of predicting recidivism. Science Advances, 4(1):eaao5580, 2018

    2018

  28. [28]

    Anthropic is quietly backpedalling on its safety commitments, 2025

    Effective Altruism Forum. Anthropic is quietly backpedalling on its safety commitments, 2025. May 2025. 23

    2025

  29. [29]

    Regulation (EU) 2024/1689 lay- ing down harmonised rules on artificial intelligence (artificial intelligence act), 2024

    European Parliament and Council of the European Union. Regulation (EU) 2024/1689 lay- ing down harmonised rules on artificial intelligence (artificial intelligence act), 2024. OJ L 2024/1689

    2024

  30. [30]

    Executive order 14179: Removing barriers to American leadership in artificial intelligence, 2025

    Executive Office of the President. Executive order 14179: Removing barriers to American leadership in artificial intelligence, 2025. January 23, 2025

    2025

  31. [31]

    Survey article: Subsidiarity.Journal of Political Philosophy, 6(2):190–218, 1998

    Andreas Follesdal. Survey article: Subsidiarity.Journal of Political Philosophy, 6(2):190–218, 1998

    1998

  32. [32]

    AI safety evaluations: An explainer.Center for Security and Emerging Technology, 2025

    Alex Friedland. AI safety evaluations: An explainer.Center for Security and Emerging Technology, 2025

    2025

  33. [33]

    Farrar, Straus and Giroux, New York, 2014

    Francis Fukuyama.Political Order and Political Decay. Farrar, Straus and Giroux, New York, 2014

    2014

  34. [34]

    George and Andrew Bennett.Case Studies and Theory Development in the Social Sciences

    Alexander L. George and Andrew Bennett.Case Studies and Theory Development in the Social Sciences. MIT Press, Cambridge, MA, 2005

    2005

  35. [35]

    Regulatory responses to medical machine learning.Journal of Law and the Biosciences, 7(1):lsaa002, 2020

    Sara Gerke, Timo Minssen, and Glenn Cohen. Regulatory responses to medical machine learning.Journal of Law and the Biosciences, 7(1):lsaa002, 2020

    2020

  36. [36]

    Alvin I. Goldman. Experts: Which ones should you trust?Philosophy and Phenomenological Research, 63(1):85–110, 2001

    2001

  37. [37]

    Frontier safety framework, 2024

    Google DeepMind. Frontier safety framework, 2024. May 2024

    2024

  38. [38]

    Verification for international AI governance

    Ben Harack et al. Verification for international AI governance. Oxford Martin AI Governance Initiative, 2025

    2025

  39. [39]

    Epistemic dependence.Journal of Philosophy, 82(7):335–349, 1985

    John Hardwig. Epistemic dependence.Journal of Philosophy, 82(7):335–349, 1985

    1985

  40. [40]

    Addressing the regulatory gap: Moving towards an EU AI audit ecosystem beyond the AI Act by including civil society.AI and Ethics, 2024

    David Hartmann, José Renato Laranjeira de Pereira, Chiara Streitbörger, and Bettina Berendt. Addressing the regulatory gap: Moving towards an EU AI audit ecosystem beyond the AI Act by including civil society.AI and Ethics, 2024

    2024

  41. [41]

    International institutions for advanced AI.arXiv preprint arXiv:2307.04699, 2023

    Lewis Ho, Joslyn Barnhart, Robert Trager, Yoshua Bengio, Miles Brundage, Allison Carnegie, Rumman Chowdhury, Allan Dafoe, Gillian Hadfield, Margaret Levi, and Duncan Snidal. International institutions for advanced AI. 2023. arXiv: 2307.04699

    work page arXiv 2023

  42. [42]

    Governance of artificial intelligence (AI)

    House of Commons Science, Innovation and Technology Committee. Governance of artificial intelligence (AI). UK Parliament, 2025. 10 January 2025

    2025

  43. [43]

    Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training

    Evan Hubinger, Carson Denison, Jesse Mu, et al. Sleeper agents: Training deceptive LLMs that persist through safety training, 2024. arXiv:2401.05566

    work page internal anchor Pith review arXiv 2024

  44. [44]

    The philosophical novelty of computer simulation methods.Synthese, 169(3): 615–626, 2009

    Paul Humphreys. The philosophical novelty of computer simulation methods.Synthese, 169(3): 615–626, 2009

    2009

  45. [45]

    Resolution on the implementation of the NPT safeguards agreement in the islamic republic of iran, 2025

    IAEA Board of Governors. Resolution on the implementation of the NPT safeguards agreement in the islamic republic of iran, 2025. GOV/2025/38, 12 June 2025

    2025

  46. [46]

    IAEA safeguards: Serving nuclear non-proliferation

    International Atomic Energy Agency. IAEA safeguards: Serving nuclear non-proliferation. IAEA Factsheet, 2024

    2024

  47. [47]

    Technologies of humility: Citizen participation in governing science.Minerva, 41(3):223–244, 2003

    Sheila Jasanoff. Technologies of humility: Citizen participation in governing science.Minerva, 41(3):223–244, 2003

    2003

  48. [48]

    Kaminski and Gianclaudio Malgieri

    Margot E. Kaminski and Gianclaudio Malgieri. The right to explanation in the AI Act. 2025. SSRN: 5194301

    2025

  49. [49]

    The worldwide governance indicators: Methodology and analytical issues.Hague Journal on the Rule of Law, 3(2):220–246, 2011

    Daniel Kaufmann, Aart Kraay, and Massimo Mastruzzi. The worldwide governance indicators: Methodology and analytical issues.Hague Journal on the Rule of Law, 3(2):220–246, 2011

    2011

  50. [50]

    Keohane, and Sidney Verba.Designing Social Inquiry: Scientific Inference in Qualitative Research

    Gary King, Robert O. Keohane, and Sidney Verba.Designing Social Inquiry: Scientific Inference in Qualitative Research. Princeton University Press, Princeton, NJ, 1994

    1994

  51. [51]

    The legitimacy of international law: A constitutionalist framework of analysis

    Mattias Kumm. The legitimacy of international law: A constitutionalist framework of analysis. European Journal of International Law, 15(5):907–931, 2004

    2004

  52. [52]

    Possible harms of artificial intelligence and the EU AI Act: Fundamental rights and risk.Journal of Risk Research, 2024

    Isabel Kusche. Possible harms of artificial intelligence and the EU AI Act: Fundamental rights and risk.Journal of Risk Research, 2024. doi: 10.1080/13669877.2024.2350720. 24

    work page doi:10.1080/13669877.2024.2350720 2024

  53. [53]

    Trustworthy artificial intelligence and the European Union AI Act: On the conflation of trustworthiness and acceptability of risk

    Johann Laux, Sandra Wachter, and Brent Mittelstadt. Trustworthy artificial intelligence and the European Union AI Act: On the conflation of trustworthiness and acceptability of risk. Regulation & Governance, 18(1):3–32, 2024

    2024

  54. [54]

    IAEA for AI

    Lawfare. Do we want an “IAEA for AI”? lawfaremedia.org, 2024

    2024

  55. [55]

    Lindberg, Michael Coppedge, John Gerring, and Jan Teorell

    Staffan I. Lindberg, Michael Coppedge, John Gerring, and Jan Teorell. V-Dem: A new way to measure democracy.Journal of Democracy, 25(3):159–169, 2014

    2014

  56. [56]

    Evaluation awareness: Why frontier AI models are getting harder to test

    Sambhav Maheshwari and Joe O’Brien. Evaluation awareness: Why frontier AI models are getting harder to test. Institute for AI Policy and Strategy, 2026. 31 March 2026

    2026

  57. [57]

    Common elements of frontier AI safety policies

    METR. Common elements of frontier AI safety policies. metr.org, 2025. December 2025 update

    2025

  58. [58]

    Conformity assess- ments and post-market monitoring: A guide to the role of auditing in the proposed European AI regulation.Minds and Machines, 32:241–268, 2022

    Jakob Mökander, Maria Axente, Federico Casolari, and Luciano Floridi. Conformity assess- ments and post-market monitoring: A guide to the role of auditing in the proposed European AI regulation.Minds and Machines, 32:241–268, 2022

    2022

  59. [59]

    Privatizing sentencing: A delegation framework for recidivism risk assessment

    Andrea Nishi. Privatizing sentencing: A delegation framework for recidivism risk assessment. Columbia Law Review, 119(6), 2019

    2019

  60. [60]

    Taking AI risks seriously: A new assessment model for the AI Act.AI & Society, 39(5), 2024

    Claudio Novelli, Federico Casolari, Antonino Rotolo, Mariarosaria Taddeo, and Luciano Floridi. Taking AI risks seriously: A new assessment model for the AI Act.AI & Society, 39(5), 2024

    2024

  61. [61]

    A robust governance for the AI Act: AI Office, AI Board, scientific panel, and national authorities

    Claudio Novelli, Philipp Hacker, Jessica Morley, Jarle Trondal, and Luciano Floridi. A robust governance for the AI Act: AI Office, AI Board, scientific panel, and national authorities. European Journal of Risk Regulation, 2025

    2025

  62. [62]

    Preparedness framework

    OpenAI. Preparedness framework. openai.com, 2024

    2024

  63. [63]

    Cambridge University Press, 1990

    Elinor Ostrom.Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press, 1990

    1990

  64. [64]

    Park, Simon Goldstein, Aidan O’Gara, Michael Chen, and Dan Hendrycks

    Peter S. Park, Simon Goldstein, Aidan O’Gara, Michael Chen, and Dan Hendrycks. AI deception: A survey of examples, risks, and potential solutions.Patterns, 5(5), 2024

    2024

  65. [65]

    Report on algorithmic risk assessment tools in the U.S

    Partnership on AI. Report on algorithmic risk assessment tools in the U.S. criminal justice system. Technical report, 2021

    2021

  66. [66]

    Harvard University Press, 2015

    Frank Pasquale.The Black Box Society: The Secret Algorithms That Control Money and Information. Harvard University Press, 2015

    2015

  67. [67]

    Inside the U.K.’s bold experiment in AI safety.TIME, 2025

    Billy Perrigo. Inside the U.K.’s bold experiment in AI safety.TIME, 2025. 16 January 2025

    2025

  68. [68]

    Oxford University Press, 1997

    Philip Pettit.Republicanism: A Theory of Freedom and Government. Oxford University Press, 1997

    1997

  69. [69]

    Joelle Pineau, Philippe Vincent-Lamarre, Koustuv Sinha, Vincent Larivière, Alina Beygelzimer, Florence d’Alché Buc, Emily Fox, and Hugo Larochelle

    Mary Phuong, Matthew Aitchison, Elliot Catt, Sarah Cogan, Victoria Krakovna, Zachary Kenton, Martin Wainwright, et al. Evaluating frontier models for dangerous capabilities. 2024. arXiv: 2403.13793

    work page arXiv 2024

  70. [70]

    The UK’s contribution to developing the international governance of AI safety and how its policy took a sharp turn.Opinio Juris, 2025

    Mando Rachovitsa. The UK’s contribution to developing the international governance of AI safety and how its policy took a sharp turn.Opinio Juris, 2025. 18 November 2025

    2025

  71. [71]

    Ragin.The Comparative Method: Moving Beyond Qualitative and Quantitative Strategies

    Charles C. Ragin.The Comparative Method: Moving Beyond Qualitative and Quantitative Strategies. University of California Press, Berkeley, 1987

    1987

  72. [72]

    The paradoxes of the European Union’s AI regulation.The Regulatory Review, 2026

    Nicoletta Rangone. The paradoxes of the European Union’s AI regulation.The Regulatory Review, 2026. March 10, 2026

    2026

  73. [73]

    Nicoletta Rangone and L. Megale. Risks without rights? the EU AI Act’s approach to AI in law and rule-making.European Journal of Risk Regulation, 16:1082–1097, 2025

    2025

  74. [74]

    Columbia University Press, New York, 1993

    John Rawls.Political Liberalism. Columbia University Press, New York, 1993

    1993

  75. [75]

    Governing artificial intelligence in china and the european union: Comparing aims and mechanisms.Global Policy, 2024

    Huw Roberts, Josh Cowls, Jessica Morley, Mariarosaria Taddeo, Vincent Wang, and Luciano Floridi. Governing artificial intelligence in china and the european union: Comparing aims and mechanisms.Global Policy, 2024

    2024

  76. [76]

    Cognitive Comparability and the Limits of Governance: Evaluating Authority Under Radical Capability Asymmetry

    Tony Rost. Cognitive comparability and the limits of governance: Evaluating authority under radical capability asymmetry. arXiv preprint, https://arxiv.org/abs/2604.02720, 2026. Superintelligence Governance Institute. 25

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  77. [77]

    Scott.Seeing Like a State: How Certain Schemes to Improve the Human Condition Have Failed

    James C. Scott.Seeing Like a State: How Certain Schemes to Improve the Human Condition Have Failed. Yale University Press, 1998

    1998

  78. [78]

    Simon.Administrative Behavior: A Study of Decision-Making Processes in Adminis- trative Organizations

    Herbert A. Simon.Administrative Behavior: A Study of Decision-Making Processes in Adminis- trative Organizations. Free Press, 4th edition, 1997

    1997

  79. [79]

    Mapping IAEA verification tools to international AI governance: A mechanism-by-mechanism analysis, 2024

    Simon Institute for Longterm Governance. Mapping IAEA verification tools to international AI governance: A mechanism-by-mechanism analysis, 2024

    2024

  80. [80]

    Smuha and Karen Yeung

    Nathalie A. Smuha and Karen Yeung. The European Union’s AI Act: Beyond motherhood and apple pie? In Nathalie A. Smuha, editor,The Cambridge Handbook of the Law, Ethics and Policy of Artificial Intelligence, pages 228–258. Cambridge University Press, 2025

    2025

Showing first 80 references.