Auditing of AI: Legal, Ethical and Technical Approaches
Open accessThe DOI matches its Crossref record; 176 of the 176 references it cites were found in a registry or an index; 10 of the 102 sources its claims cite were checked for support.
Abstract
AI auditing is a rapidly growing field of research and practice. This review article, which doubles as an editorial to Digital Society’s topical collection on ‘Auditing of AI’, provides an overview of previous work in the field. Three key points emerge from the review. First, contemporary attempts to audit AI systems have much to learn from how audits have historically been structured and conducted in areas like financial accounting, safety engineering and the social sciences. Second, both policymakers and technology providers have an interest in promoting auditing as an AI governance mechanism. Academic researchers can thus fill an important role by studying the feasibility and effectiveness of different AI auditing procedures. Third, AI auditing is an inherently multidisciplinary undertaking, to which substantial contributions have been made by computer scientists and engineers as well as social scientists, philosophers, legal scholars and industry practitioners. Reflecting this diversity of perspectives, different approaches to AI auditing have different affordances and constraints. Specifically, a distinction can be made between technology-oriented audits, which focus on the properties and capabilities of AI systems, and process-oriented audits, which focus on technology providers’ governance structures and quality management systems. The next step in the evolution of auditing as an AI governance mechanism, this article concludes, should be the interlinking of these available—and complementary—approaches into structured and holistic procedures to audit not only how AI systems are designed and used but also how they impact users, societies and the natural environment in applied settings over time.
Overview
CoChat Summary
TL;DR
This review/editorial surveys AI auditing as a multidisciplinary governance mechanism, arguing that it should learn from financial, safety, and social-science audits and combine technology-oriented and process-oriented approaches. It concludes that the next step is to interlink complementary auditing tools into structured, holistic procedures, while remaining realistic about auditing’s conceptual, technical, economic, and institutional limits.
Key findings
- Contemporary AI auditing can learn from established audit domains: financial accounting, safety engineering, and social-science audit studies. The paper notes financial auditing is a large industry with an estimated market size of over $110 bn.
- Both policymakers and technology providers have incentives to promote AI auditing: top-down regulatory pressure and bottom-up corporate risk-management pressure. Academic researchers are positioned to study the feasibility and effectiveness of different procedures.
- AI auditing is multidisciplinary and can be divided into narrow, technology-oriented audits focused on AI system outputs/properties and broad, process-oriented audits focused on providers’ governance structures and quality management systems. These approaches are complementary and should be interlinked.
- The paper cites examples such as Bertrand and Mullainathan’s audit study, where résumés with white-sounding names were 50% more likely to receive callbacks than those with black-sounding names, and Dawson et al.’s estimate that fixing a bug can cost up to 15 times more depending on when it is found.
Methods
- Design: a review article that doubles as an editorial for *Digital Society*’s topical collection on “Auditing of AI.” It provides an overview of previous work rather than presenting new empirical research.
- Data/sample: prior academic literature, policy documents, standards, guidance, and real-world examples from financial auditing, safety auditing, social-science audit studies, and AI auditing.
- Analysis: conceptual and narrative synthesis. It distinguishes narrow vs. broad conceptions of AI auditing and technical, legal, and ethical approaches, then introduces six articles in the topical collection. The paper states “Materials and Method Not applicable.”
Limitations
- The paper states that AI auditing, even when following best practices, is subject to conceptual, technical, economic, and institutional limitations; some can be addressed by policy or technology, while others are intrinsic.
- It states that claims about the limitations and affordances of AI auditing as a governance mechanism have yet to be substantiated by empirical research, and that well-established standards for AI auditing have yet to emerge.
- Evident from the text: as a review/editorial, it does not use a systematic review protocol, formal search strategy, or primary data collection; its conclusions are synthetic and illustrative rather than empirically tested.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
8 Valid · 15 Not valid · 57 Not checked · 62 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“The AI systems auditing literature is also multidisciplinary in that it harbours contributions from many academic disciplines, including computer science (Adler et al., 2018;Kearns et al., 2018), systems engineering (Dennis et al., 2016;Leveson, 2011), law (Laux et al., 2021;Selbst, 2021), media and communication studies (Bandy & Diakopoulos, 2019;Sandvig et al., 2014), social science (Metaxa et”
§ Auditing of AI: Legal, Ethical and Technical Approaches > 4.1 The AI Auditing Literature
What the sources say
- Supportedcites Adler et al., 2018 source
Adler et al. is cited for the computer-science contribution, not for the full list of disciplines. The Conceptual Context section (Section II) explicitly places black-box auditing within machine learning and computer science at large. (LLM, conf=0.93)
- Supportedcites Dennis et al., 2016 source
Dennis et al. is cited alongside Leveson for the systems-engineering contribution, not for the entire list of disciplines. Section 2, “Background and related work,” describes verification of hybrid agent systems as drawing on hybrid control systems, agent-based programming, and verification approaches; Section 4.1 describes the resulting system architecture. (LLM, conf=0.89)
- Supportedcites Laux et al., 2021 source
The citation to Laux et al. is attached to “law” as an example of a discipline contributing to AI-systems auditing literature, not to the full list of disciplines. The Introduction and section 4.1 discuss algorithmic audits in relation to the DSA’s legal and regulatory rules, so the paper fits that example. (LLM, conf=0.92)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“By assessing the capabilities and limitations of AI systems prior to deployment, auditing helps technology providers identify and mitigate risks before harm occurs (Floridi & Strait, 2020;Wilson et al., 2021).”
§ Auditing of AI: Legal, Ethical and Technical Approaches > 3.2 The Role of AI Auditing in Corporate Governance
What the sources say
- Partly supportedcites Floridi & Strait, 2020 sourceExperimental — accuracy not yet measured
Section 2.4 says technology firms use ethical reviews to identify potential harms and actionable ways to mitigate them. The paper also discusses pre-deployment assessments and testing technologies for ethical weaknesses, but it does not specifically describe auditing the capabilities and limitations of AI systems as the means of doing so (sections 3.2 and the concluding discussion; page numbers unavailable). (LLM, conf=0.84)
- Not checked yetcites Wilson et al., 2021 source
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Furthermore, by providing a basis on which technology providers can be held accountable, audits incentivise investments in adequate risk management (Shen et al., 2021).”
§ Auditing of AI: Legal, Ethical and Technical Approaches > 3.2 The Role of AI Auditing in Corporate Governance
What the sources say
- Partly supportedcites Shen et al., 2021 sourceExperimental — accuracy not yet measured
Section 7.5, “Design for Incentivization” (page unavailable), says public scrutiny can motivate companies to find and remediate harmful algorithmic behavior. It also says policy or legal interventions are needed to ensure external accountability. It does not say that audits themselves provide a basis for accountability or incentivize investment in adequate risk management. (LLM, conf=0.85)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“To start with, there is a significant body of literature calling for AI systems to be audited (Diakopoulos, 2015;Sandu et al., 2022;Sandvig et al., 2014).”
§ Auditing of AI: Legal, Ethical and Technical Approaches > 4.4 Who Audits the Auditors?
What the sources say
- Not checked yetcites Diakopoulos, 2015 source
- Supportedcites Sandu et al., 2022 source
The Introduction (section 1) states that 'various attempts have been made to audit AI algorithms (Bandy 2021)' and attributes distinct audit conceptions to several authors (Sandvig et al. 2014 expert testing; Koshiyama et al. 2021 ethical audit frameworks; Raji et al. 2020 on internal auditors' role), and it declares a literature review 'aimed at gaining insights into the elements which are important for the audit of AI algorithms'. The Abstract and section 6 frame the paper itself as a call for rigorous algorithm audit to build public trust and to meet EU regulatory pressure. As a broad background statement, the claim that a body of literature calls for AI systems to be audited is treated as such in our copy; the only nuance is that the paper characterises much of that literature as auditing attempts/frameworks rather than explicit 'calls'. (LLM, conf=0.85)
- Not checked yetcites Sandvig et al., 2014
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Reflecting that convergence, previous research has developed procedures to audit AI systems for transparency and explainability (Cobbe et al., 2021;Mittelstadt, 2016), bias and fairness (Bartley et al., 2021;Raji & Buolamwini, 2019;Morina et al., 2019) and accountability (Busuioc, 2021;Metcalf et al., 2021).”
§ Auditing of AI: Legal, Ethical and Technical Approaches > 4.3 Technical, Legal and Ethics-Based Approaches
What the sources say
- Not checked yetcites Cobbe et al., 2021
- Not checked yetcites Bartley et al., 2021
- Partly supportedcites Busuioc, 2021 sourceExperimental — accuracy not yet measured
The Conclusion (page unavailable) discusses model certification, independent testing, audit trails, and impact assessments as tools that could bolster algorithmic accountability. But it describes these as proposals, mostly not yet adopted; it does not develop procedures for auditing AI systems. Busuioc therefore supports the accountability-audit topic, but not the claim that this research developed such procedures. (LLM, conf=0.84)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
45 more claims · 1 with their cited sources checked
Create a free account to see all 142 claims and to verify the 92 references still open.
- It may state two separate claims as one. 30
- Probably not a substantial claim. 78
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“This lack of safeguards enables audited entities to search for auditors who are more likely to provide a favorable opinion or whose audits are less stringent, known as opinion shopping in the financial sector [15,39,41].”
From A Framework for Assurance Audits of Algorithmic Systems (2024) · cites this paper as [41]
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Auditing of AI: Legal, Ethical and Technical Approaches > 4.2 Narrow vs Broad Conceptions of Auditing of AI Systems
“a major challenge for both narrow and broad conceptions of AI audits remains how to define their material, i.e., how to classify AI systems in ways that enable effective and proportionate governance and oversight (Mökander et al, 2023a).”
§ Auditing of AI: Legal, Ethical and Technical Approaches > References
“123. Mökander, J., Schuett, J., Kirk, H. R., & Floridi, L. (2023a). Auditing large language models: A three- layered approach. AI and Ethics. https:// doi. org/ 10. 1007/ s43681-023-00289-2 124. Mökander, J., Sheth, M., Watson, D. S., et al. (2023b). The switch, the ladder, and the matrix: Mod- els for classifying AI systems.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Auditing of AI: Legal, Ethical and Technical Approaches > 3.1 Auditing as a Mechanism to Implement Legislation
“Similarly, the UK Information Commissioner's Office (ICO, 2020) has issued guidance on how to audit AI systems.”
§ Auditing of AI: Legal, Ethical and Technical Approaches > References
“81. Information Commissioner's Office (ISO). (2020). Guidance on the AI auditing framework: Draft guidance for consultation.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Auditing of AI: Legal, Ethical and Technical Approaches > 4.3 Technical, Legal and Ethics-Based Approaches
“Legal scholars have debated about when and how the above-listed regulations apply to AI systems (Durante & Floridi, 2021;Edwards & Veale, 2018;Pentland, 2019;Wachter et al., 2017).”
§ Auditing of AI: Legal, Ethical and Technical Approaches > References
“44. Durante, M., & Floridi, L. (2022). A legal principles-based framework for AI liability regulation.”
Citations
Cites
Language Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372 · cited 2×
“The release of ChatGPT has drawn public attention to the capacity of LLMs – such as OpenAI’s GPT -3 (Brown et al., 2020) and Google’s LaMDA (Thoppilan et al., 2022) – to generate human-like text based on the input provided to them.”
Kaugeranna, R., Kaugeranna, E., & Aion, (. S. 4. 6. (2023) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/lipics.giscience.2023.43 · Cited by 13,690
The global landscape of AI ethics guidelines
Jobin, A., Ienca, M., & Vayena, E. (2019) · Nature Machine Intelligence · doi:10.1038/s42256-019-0088-2 · Cited by 6,220 · cited 1×
“While these guidance documents vary in language (Jobin et al., 2019), they converge on a limited set of principles (Floridi & Cowls, 2019).”
BackgroundFloridi, L., Cowls, J., Beltrametti, M., Chatila, R., Chazerand, P., Dignum, V., Luetge, C., Madelin, R., Pagallo, U., Rossi, F., et al. (2018) · Minds and Machines · doi:10.1007/s11023-018-9482-5 · Cited by 4,774
Bertrand, M., & Mullainathan, S. (2004) · American Economic Review · doi:10.1257/0002828042002561 · Cited by 4,504
Capitalism, Socialism and Democracy
Schumpeter, J. A. (2013) · Routledge eBooks · doi:10.4324/9780203202050 · Cited by 4,425
The Audit Society: Rituals of VerificationPreprint
Power, M. (1999) · RePEc: Research Papers in Economics · Cited by 3,647
Barocas, S., & Selbst, A. D. (2014) · doi:10.2139/ssrn.2477899 · Cited by 3,048 · cited 2×
“Audits can provide answers to both kinds of question, as two real-world examples illustrate.”
Capitalism, Socialism and Democracy.
Hawtrey, R. G., & Schumpeter, J. (1944) · Economica · doi:10.2307/2549943 · Cited by 2,916
Model Cards for Model Reporting
Mitchell, M., Wu, S., Zaldivar, A., Barnes, P., Vasserman, L., Hutchinson, B., Spitzer, E., Raji, I. D., & Gebru, T. (2019) · Proceedings of the Conference on Fairness, Accountability, and Transparency · doi:10.1145/3287560.3287596 · Cited by 2,681 · cited 1×
“…by auditors for specific tasks, including detecting bias in AI systems (Saleiro et al., 2018; Sokol et al., 2022), documenting how AI systems are designed (Gebru et al., 2021; Mitchell et al., 2019), and simulating or monitoring their behaviour in real-world settings (Akpinar et al., 2022).”
Background
Research works printed but not linked: 7
- Allford, L., & Carson, P. (2015). Safety practice safety, health, and environment audits with selected case histories. In Loss Prevention Bulletin (241). Retrieved July 20, 2023, from www.researchgate.net/publication/307978324
- Brown, R. G. (1962). Changing audit objectives and techniques. The Accounting Review, 37(4), 696–703. Retrieved July 20, 2023, from https://www.proquest.com/docview/1301318804
- Cosserat, G. W. (2004). Modern auditing (2nd ed.). John Wiley & Sons, Ltd.
- Dawson, M., Burrell, D. N., Rahim, E., & Brewster, S. (2010). Integrating software assurance into the software development life cycle (SDLC) meeting department of defense (DOD) demands. Journal of Information Systems Technology and Planning, 3(6), 49–53. Retrieved July 20, 2023, from www.academia.edu/22484322
- Gupta, K. (2004). Comtemporary auditing. McGraw Hill.
- Klinect, J., Murray, P., Merritt, A., & Helmreich, R. (2003). Line operation safety audits: Definition and operating characteristics (pp. 663–668). Proceedings of the 12th International Symposium on Aviation Psychology.
- Mittelstadt, B. (2016). Auditing for transparency in content personalization systems. International Journal of Communication, 10, 4991–5002. Retrieved July 20, 2023, from www.researchgate.net/publication/309136069
Cited by
A Comprehensive Survey of Deep Learning Approaches in Image Processing
Τρίγκα, Μ., & Dritsas, E. (2025) · Sensors · doi:10.3390/s25020531 · Cited by 207
Policy advice and best practices on bias and fairness in AI
Álvarez, J. M., Colmenarejo, A. B., Elobaid, A., Fabbrizzi, S., Fahimi, M., Ferrara, A., Ghodsi, S., Mougan, C., Papageorgiou, I. T., Lobo, P. R., et al. (2024) · Ethics and Information Technology · doi:10.1007/s10676-024-09746-w · Cited by 160
“Algorithm auditing (Koshiyama et al., 2021), and specifically AI auditing (Mökander, 2023), is a concept coined to seek for the development of auditing frameworks on research and in practice .”
Black-Box Access is Insufficient for Rigorous AI Audits
Casper, S., Ezell, C., Siegmann, C., Kolt, N., Curtis, T. L., Bucknall, B., Haupt, A., Wei, K., Scheurer, J., Hobbhahn, M., et al. (2024) · The 2024 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3630106.3659037 · Cited by 86
“External evaluations of AI systems are emerging as a key component of AI oversight [10, 11, 19, 25, 32, 151, 160, 165, 177, 190–192, 194, 196, 201, 247, 249, 252, 272, 274, 276, 279, 284, 293, 294, 324, 333] and governance frameworks [61, 85, 217, 311, 314].”
A Framework for Assurance Audits of Algorithmic Systems
Lam, K., Lange, B., Blili-Hamelin, B., Davidovic, J., Brown, S., & Hasan, A. (2024) · The 2024 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3630106.3658957 · Cited by 75
ISO/IEC quality standards for AI engineering
Oviedo, J., Rodríguez, M., Trenta, A., Cannas, D., Natale, D., & Piattini, M. G. (2024) · Computer Science Review · doi:10.1016/j.cosrev.2024.100681 · Cited by 78
Metadata
- Authors
- Mökander, J.
- Affiliations
- Princeton University (US)
- Corresponding author
- Mökander, Jakob
- ORCID
- Mökander, Jakob
- Venue
- Digital Society
- Published
- 2023-11-08
- Volume
- 2
- Issue
- 3
- ISSN
- 2731-4650 (ISSN-L)
- 2731-4669
- OpenAlex
- W4388499666
- Semantic Scholar
- d33b084c5afed7b32591a0d8a43e8bf8756125cc
- Cited by
- 152
- References
- 167
- Citations by year
- 2026: 69 · 2025: 53 · 2024: 29 · 2023: 1
- FWCI
- 38.06
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0 (tdm)
- creativecommons.org/licenses/by/4.0 (vor)
- Locations
- Digital Society (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Funders
- Center for Information Technology Policy (CITP), Princeton University
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Social Sciences › Law › Digitalization, Law, and Regulation
- Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
- Keywords
- artificial intelligence in auditing, AI governance, quality management systems, history of auditing, societal impact of AI
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-07, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-07 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Auditing of AI: Legal, Ethical and Technical Approaches. https://app.cochat.ai/papers/10.1007/s44206-023-00074-y
- The paper
- Mökander, J. (2023). Auditing of AI: Legal, Ethical and Technical Approaches. Digital Society. https://doi.org/10.1007/s44206-023-00074-y
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.