Pith. sign in

REVIEW 3 major objections 9 minor 46 references

First review of agentic AI governance finds a field with no rules yet

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · glm-5.2

2026-07-09 04:53 UTC pith:EJP3GNVL

load-bearing objection First systematic review of agentic AI governance literature — useful synthesis but methodology excludes the policy documents where governance actually lives the 3 major comments →

arxiv 2607.07612 v1 pith:EJP3GNVL submitted 2026-07-08 cs.CY cs.AI

Towards Agentic AI Governance: A Preliminary Assessment

Mubarak Raji , Masooda Bashir This is my paper
classification cs.CY cs.AI
keywords agenticgovernanceevolvingpreliminaryreviewsystemsacceleratedadaptive
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper presents what it claims is the first systematic literature review on the governance of agentic AI — AI systems that autonomously plan, decide, and execute multi-step tasks toward goals with minimal human oversight. Through an analysis of 21 peer-reviewed articles, the authors argue that agentic AI has distinctive features (autonomy, adaptability, environmental interaction, multi-agent coordination, and persistent memory) that existing AI governance frameworks were not designed to address. They identify two competing lenses in the literature for thinking about these systems: one treats agentic AI as an autonomous goal-pursuit tool, the other treats it through moral-agency theory, asking who is the principal and who bears liability when an AI agent acts on someone's behalf. The paper synthesizes privacy concerns, stakeholder roles, and the relevance of Singapore's Model AI Governance Framework for Agentic AI (released January 2026, the world's first such framework), and lays out preliminary groundwork for a governance roadmap. The central claim is that agentic AI is sufficiently different from generative and traditional AI that it requires its own governance architecture rather than being folded into existing instruments like the EU AI Act or GDPR, which were written for systems that wait for human instructions rather than systems that act on their own.

Core claim

The paper's central finding is that scholarly discussion of agentic AI governance splits into two classification approaches — autonomous goal-pursuit (what can the system do on its own?) and moral agency (who is responsible when it does it?) — and that neither approach, nor any existing regulatory instrument, adequately covers the full risk surface created by systems that plan and act without continuous human supervision. The authors catalog eight attributes that distinguish agentic AI from prior AI generations (adaptability, autonomy, goal complexity, environmental interaction, learning capability, workflow optimization, multi-agent systems, and temporal coherence), and show that these map勉

What carries the argument

The two classification lenses — autonomous goal-pursuit and moral agency — serve as the paper's organizing device. The autonomous goal-pursuit lens, grounded in Russell and Norvig's definition of agent autonomy, catalogues eight system attributes. The moral-agency lens, grounded in common-law principal-agent doctrine, asks how liability, disclosure, and loyalty apply when the 'agent' is software. Singapore's Model AI Governance Framework for Agentic AI (January 2026) provides the only concrete governance artifact examined, organized around four dimensions: assess and bound risks upfront, make humans meaningfully accountable, implement technical controls, and enable end-user responsibility.

Load-bearing premise

The paper assumes that 21 peer-reviewed articles capture the relevant governance landscape for agentic AI, excluding 33 non-peer-reviewed works. But because agentic AI governance is nascent and rapidly evolving, much of the substantive thinking likely lives in policy documents, industry frameworks, and preprints that were filtered out — meaning the review may miss the most practically relevant governance work, including the Singapore framework itself which was discussed but (

What would settle it

The paper's central claim that agentic AI is 'sufficiently different' to warrant its own governance framework would be weakened if future empirical work shows that the risks, failure modes, and accountability gaps of agentic AI are not structurally different from those of generative AI or traditional autonomous systems — i.e., if the eight attributes catalogued turn out to be matters of degree rather than kind, and existing frameworks handle them adequately through minor amendments.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If agentic AI requires its own governance framework, organizations deploying autonomous AI agents today are operating in a liability vacuum — existing privacy law and AI ethics principles cover data collection and model behavior but not the chain of autonomous actions an agent takes on a user's behalf.
  • The two classification lenses produce different governance designs: goal-pursuit governance would focus on bounding what an agent can do technically, while moral-agency governance would focus on who is liable — and these may conflict when a system's autonomous actions exceed what any principal explicitly authorized.
  • Singapore's framework, being the only formal agentic AI governance model, may become a de facto international reference point, similar to how the EU's GDPR influenced global privacy law — but it also means a single jurisdiction is shaping the baseline architecture for a global technology.
  • The paper's finding that moral-agency discussion is limited to common-law jurisdictions means governance thinking currently excludes civil-law, religious-law, and customary-law perspectives — a gap that could fragment agentic AI governance across legal traditions.

Where Pith is reading between the lines

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

  • If the most substantive governance thinking for agentic AI lives in policy documents, industry frameworks, and grey literature (as the paper's own discussion of Singapore's framework suggests), then restricting a review to 21 peer-reviewed articles may systematically exclude the very sources where governance is actively being constructed — meaning the review captures the academic conversation but
  • The tension between the two classification lenses may not be resolvable within a single framework: goal-pursuit governance assumes the system is a tool to be bounded, while moral-agency governance assumes the system is an actor to be held responsible — these are different ontological commitments that could require parallel governance tracks rather than a unified one.
  • If agentic AI deployment grew from 11% to 42% of organizations in three quarters of 2025 (as the paper reports), governance frameworks drafted in 2026 may already be addressing yesterday's deployment patterns rather than tomorrow's — suggesting that governance for agentic AI may need to be adaptive by design rather than static.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 9 minor

Summary. This paper presents a systematic literature review of agentic AI governance, identifying two classification approaches (autonomous goal-pursuit and moral agency), synthesizing privacy concerns, stakeholder roles, and discussing Singapore's Model AI Governance Framework for Agentic AI. The review covers 21 peer-reviewed articles filtered from an initial set of approximately 3,000 papers. The paper claims to be the first systematic review on this topic and aims to lay groundwork for future governance frameworks.

Significance. The paper addresses a timely and important gap at the intersection of AI governance and agentic systems. The identification of two distinct classification lenses (autonomous goal-pursuit vs. moral agency) is a useful organizing contribution, as is the observation that the moral agency discussion is heavily skewed toward common law jurisdictions. The synthesis of agentic AI attributes in Table 2 and the discussion of the Singapore framework provide a reference point for future work. However, the methodological execution limits the strength of these contributions.

major comments (3)
  1. §3 (Method): The systematic review methodology has a structural tension with the paper's framing. The inclusion criteria restrict the corpus to 21 peer-reviewed articles, excluding 33 non-peer-reviewed works. However, the paper discusses Singapore's Model AI Governance Framework for Agentic AI (§4.3, ref [46]) and OpenAI's 'Practices for Governing Agentic AI Systems' (ref [37]) as substantive governance findings — neither of which would have passed the peer-review inclusion criterion. The paper thus synthesizes academic commentary about governance while excluding the actual governance instruments it discusses. This does not invalidate the thematic findings, but the framing as a 'systematic review of the governance landscape' overstates what the methodology captures. The authors should either (a) reframe the contribution as a review of peer-reviewed academic literature on agentic AI治理, or
  2. §3 (Method): The filtering pipeline from ~3,000 papers to 21 lacks reproducibility details. No inter-rater reliability information is provided (it is unclear whether multiple reviewers participated in screening), the specific exclusion criteria applied at each stage are not fully operationalized, and the search strategy uses only three closely related phrases ('agentic AI governance,' 'governance of agentic AI,' 'governance of autonomous agentic AI') which may miss relevant work using alternative terminology (e.g., 'AI agent accountability,' 'autonomous AI regulation'). For a paper whose central claim is to be the first systematic review, these methodological gaps are load-bearing. At minimum, the authors should report the number of reviewers, any agreement metrics, and justify the search term breadth.
  3. §4.1, Table 1: Of the 21 included papers, 19 are classified as 'Conceptual/Theoretical' and only one (Vu et al. [43]) is labeled 'Empirical/Qualitative.' This homogeneity means the synthesis reflects conceptual arguments rather than empirical evidence about governance practice. The paper does not adequately address how this composition affects the generalizability of its findings. The authors should explicitly acknowledge this as a limitation in §5 and temper claims about the governance 'landscape' accordingly.
minor comments (9)
  1. §2.1: ChatGPT is stated to have been 'released to the public in 2021' — the public release was November 2022. Please correct.
  2. Table 1: Two different papers are both cited as reference [31] (Pawar and Raheem & Hossain). This creates ambiguity throughout the paper when [31] is cited. Please renumber.
  3. §2.2: The phrase 'strategically synthesis those data' should read 'strategically synthesize those data.'
  4. §4.1: The text states 'we curate a list of features or attributes of an agentic AI in Table 2 below' but Table 2 is titled 'Attributes of Agentic AI' — the caption and text should use consistent terminology.
  5. §4.1, Table 2: The 'Literature' column for each attribute lists author names, but the mapping is inconsistent — some attributes list authors whose works 'can still be inferred from the collective reading' (as acknowledged in the text below the table). The criteria for inclusion in this column should be clarified.
  6. §4.3: The paper states Singapore's framework was 'released at the World Economic Forum in Davos on January 22, 2026' and the introduction repeats this. Given the paper was accepted at AIR-RES 2026, the temporal relationship between the framework release and the paper's search window (2020–2025) should be clarified — the framework postdates the search corpus.
  7. §1: The citation '(Raji et al., 2025) [32]' uses an author-date format inconsistent with the numbered citation style used elsewhere. Please standardize.
  8. §5: The limitation section is brief and does not mention the narrow search terms or the absence of inter-rater reliability, which are notable methodological limitations.
  9. References: Several arXiv preprints are cited (e.g., [26], [36]) despite the stated exclusion of non-peer-reviewed work. If these were included, the inclusion criteria description in §3 should be updated; if they are cited only for background context, this should be clarified.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for a careful and constructive reading of our manuscript. The referee raises three major comments concerning: (1) a structural tension between our systematic review methodology (peer-reviewed corpus only) and our discussion of non-peer-reviewed governance instruments (Singapore's Model AI Governance Framework, OpenAI's practices), (2) insufficient reproducibility details in the filtering pipeline (inter-rater reliability, operationalized exclusion criteria, search term breadth), and (3) the homogeneity of our corpus (19 of 21 papers are conceptual/theoretical) and its implications for generalizability. We address each point below and indicate where revisions will be made.

read point-by-point responses

  1. Referee: §3 (Method): Structural tension between systematic review methodology (peer-reviewed corpus only) and discussion of non-peer-reviewed governance instruments (Singapore MGF, OpenAI practices). The framing as a 'systematic review of the governance landscape' overstates what the methodology captures.

    Authors: The referee correctly identifies a genuine tension in our methodology. Our systematic review corpus is restricted to 21 peer-reviewed articles, yet we discuss Singapore's Model AI Governance Framework for Agentic AI (ref [46]) and OpenAI's 'Practices for Governing Agentic AI Systems' (ref [37]) as substantive governance instruments — neither of which would have passed our peer-review inclusion criterion. We agree that this creates a structural inconsistency between what our methodology captures and what our paper discusses. We will adopt the referee's option (a): we will reframe the contribution as a systematic review of peer-reviewed academic literature on agentic AI governance, and explicitly position the Singapore framework and OpenAI practices as contextual reference points discussed alongside — but not derived from — the systematic review corpus. We will add clarifying language in the abstract, introduction, and §4.3 to make this distinction transparent. We believe this reframing is appropriate and sufficient because the Singapore framework is genuinely the only existing dedicated agentic AI governance instrument and omitting it entirely would weaken the paper's practical relevance; however, we agree it must be clearly distinguished from the systematic review findings. revision: yes

  2. Referee: §3 (Method): Filtering pipeline from ~3,000 papers to 21 lacks reproducibility details. No inter-rater reliability information, exclusion criteria not fully operationalized, search strategy uses only three closely related phrases which may miss relevant work using alternative terminology.

    Authors: The referee is correct that our methodology section does not provide sufficient reproducibility detail. We will make the following revisions: (1) We will report the number of reviewers — both authors participated in screening and title/abstract review, with disagreements resolved through discussion. We will state this explicitly. (2) We will add inter-rater agreement information where applicable, though we acknowledge that because this was a two-author process with consensus resolution rather than independent dual coding with formal metrics (e.g., Cohen's kappa), we cannot retroactively compute a reliability statistic. We will be transparent about this rather than misrepresent the process. (3) We will more fully operationalize the exclusion criteria at each filtering stage, providing clearer definitions of what constituted exclusion at each step. (4) Regarding search term breadth, we acknowledge the referee's point that our three search phrases are closely related and may miss work using alternative terminology such as 'AI agent accountability' or 'autonomous AI regulation.' We will add these and related terms to our search strategy and conduct a supplementary search to check whether additional relevant papers were missed. We will report the results of this supplementary search. If additional papers are identified, we will assess whether they change our thematic findings. We note as a limitation that the novelty of 'agentic AI' as a distinct term means that some relevant work may use adjacent terminology, and we will acknowledge this explicitly. revision: yes

  3. Referee: §4.1, Table 1: Of 21 included papers, 19 are 'Conceptual/Theoretical' and only one (Vu et al.) is 'Empirical/Qualitative.' This homogeneity means the synthesis reflects conceptual arguments rather than empirical evidence about governance practice. The paper does not adequately address how this composition affects generalizability of findings.

    Authors: The referee's observation is factually correct: 19 of 21 papers are conceptual or theoretical, and only Vu et al. [43] represents empirical qualitative work. We agree that this composition means our synthesis primarily reflects conceptual arguments and normative proposals rather than empirical evidence about governance in practice. We will address this in two ways: (1) We will add an explicit limitation in §5 acknowledging that the corpus is overwhelmingly conceptual/theoretical, which means our findings represent the state of scholarly discourse rather than empirical evidence about how agentic AI governance operates in practice. (2) We will temper claims throughout the paper that imply our review captures the governance 'landscape' — reframing to reflect that we capture the academic discourse landscape. (3) We will add a note in §5 that the scarcity of empirical studies is itself a finding, pointing to a significant gap in the literature that future research should address. We note that this homogeneity is partly a function of the recency of agentic AI as a distinct concept — empirical governance studies require deployed systems and observed practice, which is only beginning to emerge. This does not excuse the lack of acknowledgment, and we will remedy that. revision: yes

Circularity Check

0 steps flagged

No significant circularity; one minor self-citation used for a factual claim, not load-bearing

full rationale

This is a literature review paper, not a derivation or prediction paper. It does not fit the circularity patterns (self-definitional, fitted-input-called-prediction, uniqueness-imported, ansatz-smuggled, renaming-known-result) because it makes no first-principles claims, fits no parameters, and derives no predictions from its own definitions. The one self-citation (Raji et al. 2025, ref [32]) is used in Section 1 for a factual statement about data protection enforcement mechanisms ('breach of data privacy laws can expose violators to different kinds of enforcement, including administrative, civil, criminal, or financial sanctions'). This is a supporting citation for a background fact, not a load-bearing premise that the paper's central argument reduces to. The paper's main contributions — identifying two classification approaches (autonomous goal-pursuit and moral agency), synthesizing privacy concerns, and discussing the Singapore framework — are grounded in external sources (Bent, Kolt, Lior, Navaie, Singapore MGF, etc.) that are independent of the authors. The thematic synthesis is descriptive, not derivational. The skeptic's concern about excluding grey literature is a methodological scope limitation (correctness risk), not a circularity problem: the paper does not define its inputs in terms of its outputs or vice versa. Score 1 reflects the one minor self-citation that is non-load-bearing.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 0 invented entities

No free parameters (this is a qualitative review). Axioms are domain assumptions about source selection and categorization. No invented entities. The ledger is thin because the paper is a literature synthesis rather than a derivation or model.

axioms (4)
  • domain assumption Peer-reviewed publications are more valid and reliable than non-peer-reviewed sources for governance research
    Section 3: 33 papers excluded for lack of formal peer review, citing Kelly et al. [17]. This assumption removes policy documents, industry frameworks, and grey literature that are primary sources in governance research.
  • domain assumption Agentic AI is a distinct category from generative AI and traditional AI, warranting separate governance treatment
    Section 3: 389 papers on generative AI, conversational AI, and traditional AI were excluded. The paper's central claim depends on agentic AI being categorically distinct.
  • domain assumption Software-based agentic AI governance differs from governance of physical autonomous devices
    Section 3: 552 papers on autonomous vehicles, military equipment, and drones were excluded because 'their governance differs from that of software-based AI.'
  • domain assumption The principal-agent framework from common law is an appropriate lens for analyzing AI agent liability
    Section 4.1B: The moral agency analysis relies entirely on common law doctrine. The paper acknowledges this limitation in Section 5 but the framework is load-bearing for the liability discussion.

pith-pipeline@v1.1.0-glm · 16313 in / 2456 out tokens · 479901 ms · 2026-07-09T04:53:34.938347+00:00 · methodology

0 comments
read the original abstract

Artificial intelligence is rapidly evolving from generative systems to agentic AI capable of autonomously planning and executing tasks. Widely characterized as the Year of Agentic AI, 2025 marked accelerated development and deployment, introducing new ethical and governance challenges. This paper presents a systematic review of the emerging literature on agentic AI governance. Our analysis identifies features that distinguish agentic AI from traditional systems and why it warrants targeted governance attention. We synthesize prevailing governance priorities, proposed mechanisms, and stakeholder roles shaping this evolving domain. As an initial scholarly effort, this review lays the preliminary groundwork for developing a structured roadmap to guide responsible and adaptive agentic AI governance.

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

46 extracted references · 46 canonical work pages · 1 internal anchor

  1. [1]

    A high-throughput and secure coded blockchain for IoT,

    Acharya, D.B. et al.: Agentic AI: Autonomous Intelligence for Complex Goals —A Comprehensive Survey. IEEE Access. 13, 18912 –18936 (2025). https://doi.org/10.1109/ACCESS.2025.3532853

    work page doi:10.1109/access.2025.3532853 2025

  2. [2]

    et al.: AI governance: a systematic literature review

    Batool, A. et al.: AI governance: a systematic literature review. AI Ethics. 5, 3, 3265 –3279 (2025). https://doi.org/10.1007/s43681-024-00653-w

    work page doi:10.1007/s43681-024-00653-w 2025

  3. [3]

    Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society

    Bent, B.: The Term ‘Agent’ Has Been Diluted Beyond Utility and Requires Redefinition. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society. 8, 1, 403–413 (2025). https://doi.org/10.1609/aies.v8i1.36558

    work page doi:10.1609/aies.v8i1.36558 2025

  4. [4]

    et al.: AI governance: themes, knowledge gaps and future agendas

    Birkstedt, T. et al.: AI governance: themes, knowledge gaps and future agendas. Internet Research. 33, 7, 133–167 (2023). https://doi.org/10.1108/INTR-01-2022-0042

    work page doi:10.1108/intr-01-2022-0042 2023

  5. [5]

    et al.: Harms from Increasingly Agentic Algorithmic Systems

    Chan, A. et al.: Harms from Increasingly Agentic Algorithmic Systems. In: Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency. pp. 651 –666 Association for Computing Machinery, New York, NY, USA (2023). https://doi.org/10.1145/3593013.3594033

    work page doi:10.1145/3593013.3594033 2023

  6. [6]

    et al.: Visibility into AI Agents

    Chan, A. et al.: Visibility into AI Agents. In: Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency. pp. 958 –973 Association for Computing Machinery, New York, NY, USA (2024). https://doi.org/10.1145/3630106.3658948

    work page doi:10.1145/3630106.3658948 2024

  7. [7]

    In: Handbook on the Ethics of Artificial Intelligence

    Chesterman, S.: From Ethics to Law: Why, When, and How to Regulate AI. In: Handbook on the Ethics of Artificial Intelligence. pp. 113 –127 Edward Elgar (2024). https://doi.org/https://doi.org/10.4337/9781803926728

    work page doi:10.4337/9781803926728 2024

  8. [8]

    Michigan Journal of Race and Law

    Coleman, D.: Digital Colonialism: The 21st Century Scramble for Africa through the Extraction and Control of User Data and the Limitations of Data Protection Laws. Michigan Journal of Race and Law. 24, 2, 417 –439 (2019). https://doi.org/https://doi.org/10.36643/mjrl.24.2.digital

    work page doi:10.36643/mjrl.24.2.digital 2019

  9. [9]

    Television & New Media

    Couldry, N., Mejias, U.A.: Data Colonialism: Rethinking Big Data’s Relation to the Contemporary Subject. Television & New Media. 20, 4, 336 –349 (2019). https://doi.org/10.1177/1527476418796632

    work page doi:10.1177/1527476418796632 2019

  10. [10]

    American Advanced Journal for Emerging Disciplinaries (AAJED) ISSN: 3067-

    Dodda, A.: AI Governance and Security in Fintech: Ensuring Trust in Generative and Agentic AI Systems. American Advanced Journal for Emerging Disciplinaries (AAJED) ISSN: 3067-

    work page

  11. [11]

    Policy Soc

    Gahnberg, C.: What rules? Framing the governance of artificial agency. Policy Soc. 40, 2, 194–210 (2021). https://doi.org/10.1080/14494035.2021.1929729

    work page doi:10.1080/14494035.2021.1929729 2021

  12. [12]

    Journal of Computer Science and Technology Studies

    Garg, V.: Designing the Mind: How Agentic Frameworks Are Shaping the Future of AI Behavior. Journal of Computer Science and Technology Studies. 7, 5, 182 –193 (2025). https://doi.org/10.32996/jcsts.2025.7.5.24

    work page doi:10.32996/jcsts.2025.7.5.24 2025

  13. [13]

    Gartner: Gartner Predicts 40% of Enterprise Apps Will Feature Task -Specific AI Agents by 2026, Up from Less Than 5% in 2025, https://www.gartner.com/en/newsroom/press - releases/2025-08-26-gartner-predicts-40-percent-of-enterprise-apps-will-feature-task- specific-ai-agents-by-2026-up-from-less-than-5-percent-in-2025, last accessed 2026/01/12

    work page 2026

  14. [14]

    Minds & Machines

    Hagendorff, T.: The Ethics of AI Ethics: An Evaluation of Guidelines. Minds & Machines. 30, 1, 99–120 (2020). https://doi.org/10.1007/s11023-020-09517-8

    work page doi:10.1007/s11023-020-09517-8 2020

  15. [15]

    Hosseini, S., Seilani, H.: The role of agentic AI in shaping a smart future: A systematic review. Array. 26, 100399 (2025). https://doi.org/10.1016/j.array.2025.100399. Towards Agentic AI Governance 15

    work page doi:10.1016/j.array.2025.100399 2025

  16. [16]

    et al.: AI Agents and Agentic Systems: A Multi -Expert Analysis

    Hughes, L. et al.: AI Agents and Agentic Systems: A Multi -Expert Analysis. Journal of Computer Information Systems. 65, 4, 489 –517 (2025). https://doi.org/10.1080/08874417.2025.2483832

    work page doi:10.1080/08874417.2025.2483832 2025

  17. [17]

    et al.: Peer Review in Scientific Publications: Benefits, Critiques, & A Survival Guide

    Kelly, J. et al.: Peer Review in Scientific Publications: Benefits, Critiques, & A Survival Guide. EJIFCC. 25, 3, 227–243 (2014)

    work page 2014

  18. [18]

    https://doi.org/10.2139/ssrn.4772956

    Kolt, N.: Governing AI Agents, https://papers.ssrn.com/abstract=4772956, (2025). https://doi.org/10.2139/ssrn.4772956

    work page doi:10.2139/ssrn.4772956 2025

  19. [19]

    KPMG: Agent Deployment Accelerates as Organizations Build Confidence Through Early Wins, https://kpmg.com/us/en/media/news/q3-ai-pulse.html, last accessed 2026/01/01

    work page 2026

  20. [20]

    Mitchell Hamline Law Review

    Lior, A.: AI Entities as AI Agents: Artificial Intelligence Liability and the AI Respondeat Superior Analogy. Mitchell Hamline Law Review. 46, 5, (2020)

    work page 2020

  21. [21]

    MarketsandMarkets: Agentic AI Market Share, Forecast | Growth Analysis by 2032, https://www.marketsandmarkets.com/Market-Reports/agentic-ai-market-208190735.html, last accessed 2026/01/12

    work page 2032

  22. [22]

    IEEE Intelligent Systems

    Murugesan, S.: The Rise of Agentic AI: Implications, Concerns, and the Path Forward. IEEE Intelligent Systems. 40, 2, 8–14 (2025). https://doi.org/10.1109/MIS.2025.3544940

    work page doi:10.1109/mis.2025.3544940 2025

  23. [23]

    Presented at the AAAI 2025 Workshop on AI Governance: Alignment, Morality, and Law March 3 (2025)

    Nasim, I.: Governance in Agentic Workflows: Leveraging LLMs as Oversight Agents. Presented at the AAAI 2025 Workshop on AI Governance: Alignment, Morality, and Law March 3 (2025)

    work page 2025

  24. [24]

    AI Magazine

    Navaie, K.: From rights to runtime: Privacy engineering for agentic AI. AI Magazine. 46, 4, e70036 (2025). https://doi.org/10.1002/aaai.70036

    work page doi:10.1002/aaai.70036 2025

  25. [25]

    et al.: Law-Following AI: Designing AI Agents to Obey Human Laws

    O’Keefe, C. et al.: Law-Following AI: Designing AI Agents to Obey Human Laws. Fordham Law Review. 94, 1, 57–129 (2025)

    work page 2025

  26. [26]

    https://doi.org/10.48550/arXiv.2601.17191

    Okolo, C.T., Raji, M.: The Global Majority in International AI Governance, http://arxiv.org/abs/2601.17191, (2026). https://doi.org/10.48550/arXiv.2601.17191

    work page doi:10.48550/arxiv.2601.17191 2026

  27. [27]

    The Future Society (2025)

    Oueslati, A., Staes-Polet, R.: Ahead of the Curve: Governing AI Agents under the EU AI Act. The Future Society (2025)

    work page 2025

  28. [28]

    et al.: Responsible artificial intelligence governance: A review and research framework

    Papagiannidis, E. et al.: Responsible artificial intelligence governance: A review and research framework. The Journal of Strategic Information Systems. 34, 2, 101885 (2025). https://doi.org/10.1016/j.jsis.2024.101885

    work page doi:10.1016/j.jsis.2024.101885 2025

  29. [29]

    International Journal of Humanities and Information Technology

    Pawar, A.: Ethical and Governance Challenges of Agentic AI. International Journal of Humanities and Information Technology. 7, 03, 76–82 (2025)

    work page 2025

  30. [30]

    PricewaterhouseCoopers: PwC’s AI Agent Survey, https://www.pwc.com/us/en/tech - effect/ai-analytics/ai-agent-survey.html, last accessed 2026/01/12

    work page 2026

  31. [31]

    In: 2025 IEEE International Conference on Electro Information Technology (eIT)

    Raheem, T., Hossain, G.: Agentic AI Systems: Opportunities, Challenges, and Trustworthiness. In: 2025 IEEE International Conference on Electro Information Technology (eIT). pp. 618–624 (2025). https://doi.org/10.1109/eIT64391.2025.11103638

    work page doi:10.1109/eit64391.2025.11103638 2025

  32. [32]

    et al.: An Assessment of the Enforcement Mechanisms in African Data Protection Laws

    Raji, M. et al.: An Assessment of the Enforcement Mechanisms in African Data Protection Laws. African Journal on Privacy & Data Protection. 2, 1, 1 –40 (2025). https://doi.org/10.29053/ajpdp.v2i1.0002

    work page doi:10.29053/ajpdp.v2i1.0002 2025

  33. [33]

    Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society

    Riedl, M.O., Desai, D.R.: AI Agents and the Law. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society. 8, 3, 2189 –2198 (2025). https://doi.org/10.1609/aies.v8i3.36705

    work page doi:10.1609/aies.v8i3.36705 2025

  34. [34]

    Pearson, Boston (2022)

    Russell, S.J., Norvig, P.: Artificial intelligence: a modern approach. Pearson, Boston (2022)

    work page 2022

  35. [35]

    2, 4, 325–336 (2025)

    Saloustrou, C.: Agentic AI Regulation under the AI Act: Journal of AI Law and Regulation. 2, 4, 325–336 (2025). https://doi.org/10.21552/aire/2025/4/5. Towards Agentic AI Governance 16

    work page doi:10.21552/aire/2025/4/5 2025

  36. [36]

    Agentic Inequality

    Sharp, M. et al.: Agentic Inequality, http://arxiv.org/abs/2510.16853, (2025). https://doi.org/10.48550/arXiv.2510.16853

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2510.16853 2025

  37. [37]

    et al.: Practices for Governing Agentic AI Systems

    Shavit, Y. et al.: Practices for Governing Agentic AI Systems. OpenAI. (2023)

    work page 2023

  38. [38]

    et al.: Critically Comparing Global AI Regulations

    Silva, E.R.B. et al.: Critically Comparing Global AI Regulations. (2025). https://doi.org/10.24251/HICSS.2025.537

    work page doi:10.24251/hicss.2025.537 2025

  39. [39]

    GW Law Faculty Publications & Other Works

    Solove, D.: A Taxonomy of Privacy. GW Law Faculty Publications & Other Works. (2006)

    work page 2006

  40. [40]

    Florida Law Review

    Solove, D.: Artificial Intelligence and Privacy. Florida Law Review. 77, 1, 1 (2025)

    work page 2025

  41. [41]

    Urban Info

    Tiwari, A.: Conceptualising the emergence of Agentic Urban AI: from automation to agency. Urban Info. 4, 1, 13 (2025). https://doi.org/10.1007/s44212-025-00079-7

    work page doi:10.1007/s44212-025-00079-7 2025

  42. [42]

    United Nations (2024)

    United Nations: Governing AI for Humanity. United Nations (2024). https://doi.org/10.18356/9789211067873

    work page doi:10.18356/9789211067873 2024

  43. [43]

    et al.: Agentic Business Process Management: Practitioner Perspectives on Agent Governance in Business Processes

    Vu, H. et al.: Agentic Business Process Management: Practitioner Perspectives on Agent Governance in Business Processes. In: ER, M. et al. (eds.) Business Process Management: Responsible BPM Forum, Process Technology Forum, Educators Forum. pp. 29–43 Springer Nature Switzerland, Cham (2026). https://doi.org/10.1007/978-3-032-02936-2_3

    work page doi:10.1007/978-3-032-02936-2_3 2026

  44. [44]

    et al.: Artificial intelligence: A powerful paradigm for scientific research

    Xu, Y. et al.: Artificial intelligence: A powerful paradigm for scientific research. Innovation (Camb). 2, 4, 100179 (2021). https://doi.org/10.1016/j.xinn.2021.100179

    work page doi:10.1016/j.xinn.2021.100179 2021

  45. [45]

    PublicAffairs, New York, NY (2019)

    Zuboff, S.: The age of surveillance capitalism: the fight for a human future at the new frontier of power. PublicAffairs, New York, NY (2019)

    work page 2019

  46. [46]

    Singapore Model AI Governance Framework for Agentic AI, https://www.imda.gov.sg/ - /media/imda/files/about/emerging-tech-and-research/artificial-intelligence/mgf-for-agentic- ai.pdf

    work page