pith. sign in

arxiv: 2507.11567 · v1 · submitted 2025-07-14 · 💻 cs.CY

Consumer Law for AI Agents

Christoph Busch This is my paper

Pith reviewed 2026-05-19 04:16 UTC · model grok-4.3

classification 💻 cs.CY
keywords AI agentsconsumer lawEU consumer protectionCustobotse-commerceartificial intelligencelegal adaptationmachine decision making
0
0 comments X

The pith

AI agents that autonomously make purchases challenge the human-centric assumptions of EU consumer law.

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

The paper examines how next-generation AI agents, able to plan and carry out buying tasks with little ongoing human direction, will alter the e-commerce environment. It argues that current EU consumer rules rest on the idea that humans make consumption choices and therefore may not fit situations where machines decide. The author sketches early ideas for updating the law so it can protect users whether decisions come from people or from AI systems. Readers would care because widespread delegation to such agents could render existing safeguards ineffective or create new risks in everyday transactions.

Core claim

The advent of AI agents capable of planning and executing complex purchasing decisions with only limited human involvement will change the e-commerce landscape and challenge the premises of human-centric EU consumer law, which assumes consumption decisions are made by humans; the paper therefore offers initial considerations for a future consumer law that works for both humans and machines.

What carries the argument

The Custobot, an AI agent that plans and executes purchasing tasks autonomously with only limited human involvement, which transfers decision-making authority from consumer to machine.

Load-bearing premise

Consumers will delegate purchasing decisions to AI agents acting with only limited human involvement in the near future.

What would settle it

Evidence that current EU consumer law rules already function effectively for transactions completed by AI agents with minimal problems would show that major adaptation is not required.

read the original abstract

Since the public release of ChatGPT in November 2022, the AI landscape is undergoing a rapid transformation. Currently, the use of AI chatbots by consumers has largely been limited to image generation or question-answering language models. The next generation of AI systems, AI agents that can plan and execute complex tasks with only limited human involvement, will be capable of a much broader range of actions. In particular, consumers could soon be able to delegate purchasing decisions to AI agents acting as Custobots. Against this background, the Article explores whether EU consumer law, as it currently stands, is ready for the rise of the Custobot Economy. In doing so, the Article makes three contributions. First, it outlines how the advent of AI agents could change the existing e-commerce landscape. Second, it explains how AI agents challenge the premises of a human-centric consumer law which is based on the assumption that consumption decisions are made by humans. Third, the Article presents some initial considerations how a future consumer law could look like that works for both humans and machines.

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

1 major / 3 minor

Summary. The manuscript argues that the emergence of AI agents (termed 'Custobots') capable of planning and executing complex tasks including purchasing decisions with only limited human involvement will transform the e-commerce landscape and challenge the human-centric premises of current EU consumer law. It makes three contributions: outlining changes to e-commerce, explaining how AI agents undermine assumptions that consumption decisions are made by humans, and offering initial considerations for a future consumer law framework that accommodates both humans and machines.

Significance. If the paper's predictions about near-term delegation to autonomous AI agents hold, the work identifies important doctrinal gaps in EU consumer law (e.g., around informed consent, contract formation, and liability attribution) and provides a conceptual starting point for regulatory adaptation. The forward-looking analysis is timely given rapid AI progress, though its value depends on the realism of the adoption premise.

major comments (1)
  1. [Abstract and Section on AI agents and e-commerce landscape] Abstract and the section outlining how AI agents change the e-commerce landscape: The central claim that AI agents challenge human-centric consumer law rests on the premise that consumers 'could soon be able to delegate purchasing decisions to AI agents acting as Custobots' with only limited human involvement. The manuscript provides no technical feasibility analysis, adoption data, references to current agent limitations (e.g., reliability in contract formation or error handling), or discussion of regulatory barriers. This assumption is load-bearing; if delegation remains tethered to substantial human oversight or narrow tasks, the claimed challenge to existing EU consumer law premises does not arise.
minor comments (3)
  1. The neologism 'Custobot' is introduced without an explicit definition or etymology on first use; a brief clarifying sentence would aid readability for a mixed legal-technical audience.
  2. References to specific EU directives (e.g., Consumer Rights Directive, Unfair Commercial Practices Directive) would benefit from pinpoint citations to articles or recitals when discussing how AI agents interact with particular rules.
  3. The third contribution on future consumer law considerations remains high-level; adding even one concrete example (e.g., how a right of withdrawal might apply to an AI-initiated purchase) would strengthen the section without expanding scope.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive comments on our manuscript. The feedback correctly identifies that the central premise requires stronger contextual support. We address the major comment below and have revised the manuscript to incorporate additional grounding.

read point-by-point responses

  1. Referee: Abstract and the section outlining how AI agents change the e-commerce landscape: The central claim that AI agents challenge human-centric consumer law rests on the premise that consumers 'could soon be able to delegate purchasing decisions to AI agents acting as Custobots' with only limited human involvement. The manuscript provides no technical feasibility analysis, adoption data, references to current agent limitations (e.g., reliability in contract formation or error handling), or discussion of regulatory barriers. This assumption is load-bearing; if delegation remains tethered to substantial human oversight or narrow tasks, the claimed challenge to existing EU consumer law premises does not arise.

    Authors: We agree that the argument's force depends on the plausibility of AI agents assuming greater autonomy in purchasing decisions. As a legal analysis focused on doctrinal implications rather than a technical forecast, the manuscript does not purport to deliver a full feasibility study or empirical adoption data. To address the concern, we have revised the abstract and the relevant section to include references to recent AI agent research (e.g., advancements in planning, tool use, and multi-step execution) and a concise discussion of current limitations, including reliability issues in complex tasks, error handling, and the continued role of human oversight. We have also noted potential regulatory barriers in passing. These additions provide better context for the premise while preserving the paper's primary contribution as an examination of EU consumer law gaps. We believe the forward-looking legal analysis remains valuable even if full autonomy is not yet realized. revision: yes

Circularity Check

0 steps flagged

No significant circularity in prospective legal analysis

full rationale

The paper offers a forward-looking legal discussion on potential impacts of AI agents on EU consumer law, outlining landscape changes, challenges to human-centric assumptions, and initial reform considerations. No derivation chain, equations, fitted parameters, or self-referential definitions appear; claims rest on explicitly prospective statements about future AI capabilities rather than reducing to inputs by construction. The analysis is self-contained as a policy exploration without load-bearing self-citations or renamings of known results.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The paper rests on assumptions about future AI adoption and the current state of EU law being human-centric.

axioms (1)
  • domain assumption Consumption decisions are currently made by humans
    Basis for human-centric consumer law as stated in abstract.
invented entities (1)
  • Custobot no independent evidence
    purpose: AI agent acting as consumer purchasing decision maker
    New term introduced for the concept.

pith-pipeline@v0.9.0 · 5698 in / 942 out tokens · 32533 ms · 2026-05-19T04:16:14.522296+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

26 extracted references · 26 canonical work pages · 1 internal anchor

  1. [1]

    Custobots

    Preprint available: http://dx.doi.org/10.2139/ssrn.5187056 Abstract Since the public release of ChatGPT in November 2022, the AI landscape is undergoing a rapid transformation. Currently, the use of AI chatbots by consumers has largely been limited to image generation or question-answering language models. The next generation of AI systems, AI agents that...

    work page doi:10.2139/ssrn.5187056 2022

  2. [2]

    Operator

    Optimizing for AI Agents .......................................................................................... 7 C. How AI Agents Challenge the Premises of Consumer Law ................................................. 8 I. Beyond human-centricity: Towards an Average Consumer Test for Custobots? .............. 8 II. Information Duties: More is More? ...

    work page 2025

  3. [3]

    Custobot Economy,

    4 Scheibenreif & Raskino, supra note 3, at 1; see also Christian Twigg-Flesner & Geraint Howells, Adapting Consumer Law to New Technologies, in Roger Brownsword & Larry DiMatteo (eds) The Cambridge Handbook of the Governance of Technology: Discontent, Disconnect and Disruption (forthcoming 2025). 5 See e.g. Noam Kolt, Governing AI Agents (Feb. 11, 2025), ...

    work page doi:10.2139/ssrn.4772956 2025

  4. [4]

    13 For an overview see Melissa Heikkilä, What Are AI Agents?, MIT Tech. Rev. (Jul. 5, 2024), https://www.technologyreview.com/2024/07/05/1094711/what-are-ai-agents/. 14 Friso Bostoen & Jan Krämer, AI agents and ecosystem contestability, CERRE Issue paper (Nov. 2024),

    work page 2024

  5. [5]

    New Customer Journeys First of all, AI agents will change the way customers search for products. Today, the customer journey often begins with a search engine like Google or one of the large online marketplaces, 15 UK Department for Science, Innovation and Technology, A pro-innovation approach to AI regulation: government response (Feb. 6, 2024), https://...

    work page 2024

  6. [6]

    3, 2024), https://www.ibm.com/think/topics/ai-agents 17 OpenAI, Practices for Governing Agentic AI Systems (Dec

    16 Anna Gutowska, What are AI agents?, IBM (Jul. 3, 2024), https://www.ibm.com/think/topics/ai-agents 17 OpenAI, Practices for Governing Agentic AI Systems (Dec. 14,

    work page 2024

  7. [7]

    Do It For Me

    19 David G. W. Birch, A-Commerce Is Coming: Agentic AI And The “Do It For Me” Economy, Forbes (Feb. 6, 2025), https://www.forbes.com/sites/davidbirch/2025/02/06/a-commerce-is-coming-agentic-ai-and-the-do-it-for-me-economy/. 6 such as Amazon, which has itself become a kind of product search engine. In the near future, AI agents could serve as new access po...

    work page 2025

  8. [8]

    what we call the effects of technology are not so much features of things as they are features of social relations that employ those things

    21 Deutsche Telekom showcases app-less AI smartphone concept, Reuters (Feb. 26, 2024), https://www.reuters.com/technology/deutsche-telekom-showcases-app-less-ai-smartphone-concept-2024-02-26/ 22 Cf. Jack M. Balkin, The Path of Robotics Law, 6 Cal Law Rev 45, 49 (2015) (arguing that “what we call the effects of technology are not so much features of things...

    work page 2024

  9. [9]

    Here’s What That Means for Brands, Harvard Business Review (Feb

    24 Jur Gaarlandt, Wesley Korver & Andrew Shipilov, AI Agents Are Changing How People Shop. Here’s What That Means for Brands, Harvard Business Review (Feb. 26, 2025), https://hbr.org/2025/02/ai-agents-are-changing-how-people-shop-heres-what-that-means-for-brands. 25 See generally William Samuelson & Richard Zeckhauser, Status quo bias in decision making, ...

    work page 2025

  10. [10]

    29 Rory van Loo, Helping Buyers Beware: The Need for Supervision of Big Retail, 163 U. Pa. L. Rev. 1311, 1345-47 (2015); see also Glenn Ellison & Sara Fisher Ellison, Search Obfuscation, and Price Elasticities on the Internet, 77 Econometrica 427, 428-29 (2009) (showing that in highly commodified electronic parts markets, consumers paid 6-9% higher prices...

    work page 2015

  11. [11]

    32 UK Competition and Markets Authority, Algorithms: How they can reduce competition and harm consumers (Mar

    31 Id. 32 UK Competition and Markets Authority, Algorithms: How they can reduce competition and harm consumers (Mar. 25, 2022), https://www.gov.uk/find-digital-market-research/algorithms-how-they-can-reduce-competition-and-harm-consumers-2021-cma; Raluca Mihaela Ursu, The Power of Rankings: Quantifying the Effect of Rankings on Online Consumer Search and ...

    work page 2022

  12. [12]

    augmented consumers

    31-54. 33 Ashish Agarwal, Kartik Hosanagar, & Michael D. Smith, Location, location, location: An analysis of profitability of position in online advertising markets, 48(6) Journal of Marketing Research 1057-1073 (2011). 8 SEO or to pay for a better ranking position on an online platform. Instead, it may be necessary to adopt different strategies. Rather t...

    work page 2011

  13. [13]

    18, 2024), https://innovation.consumerreports.org/empowering-consumers-with-personal-ai-agents-legal-foundations-and-design-considerations/

    35 Dazza Greenwood, Empowering Consumers with Personal AI Agents: Legal Foundations and Design Considerations, Consumer Reports (Oct. 18, 2024), https://innovation.consumerreports.org/empowering-consumers-with-personal-ai-agents-legal-foundations-and-design-considerations/. 36 Rory van Loo, Digital Market Perfection, 117 Michigan Law Review 815, 830 (2019...

    work page 2024

  14. [14]

    reasonably well-informed, reasonably observant and circumspect

    199-220. 38 See Art. 5(2) Unfair Commercial Practices Directive 2005/29/EC. 9 consumer who is “reasonably well-informed, reasonably observant and circumspect”.39 This average consumer concept, which was initially developed by the case law of the CJEU, has since been codified by the European legislator in the secondary legislation.40 The average consumer c...

    work page 2005

  15. [15]

    40 See Recital 18 Unfair Commercial Practices Directive 2005/29/EC; see also Hanna Schebesta & Kai P

    ECR I – 4657, paras 30-31. 40 See Recital 18 Unfair Commercial Practices Directive 2005/29/EC; see also Hanna Schebesta & Kai P. Purnhagen, Island or Ocean: Empirical Evidence on the Average Consumer Concept in the UCPD, 28 European Review of Private Law 293-310 (2020). 41 See, e.g. Rossella Incardona & Cristina Poncibò, The average consumer, the unfair c...

    work page 2005

  16. [16]

    43 CJEU, Case C-646/22 Compass Banca SpA v Autorità Garante della Concorrenza e del Mercato (Nov

    38-76. 43 CJEU, Case C-646/22 Compass Banca SpA v Autorità Garante della Concorrenza e del Mercato (Nov. 14, 2024), para

    work page 2024

  17. [17]

    45 Van Loo, supra note 6, at 833; see also Christine Jolls, Behavioral Economics Analysis of Redistributive Legal Rules, 51 Vand. L. Rev. 1653, 1659 (1998); Christine Jolls, Cass Sunstein and Richard Thaler, A Behavioral Approach to Law and Economics, 50 Stan. L. Rev

    work page 1998

  18. [18]

    arXiv preprint arXiv:2303.13988 , year=

    46 Cf. Ulrike Malmendier & Stefano DellaVigna, Paying Not to Go to the Gym, 96(3) American Economic Review 694 (2006) (showing that overconfident consumers overestimate gym attendance as well as the cancellation probability of automatically renewed contracts); see also Christoph Busch & Christian Twigg-Flesner, A Roadmap for Regulating Subscriptions in th...

    work page doi:10.48550/arxiv.2303.13988 2006

  19. [19]

    52 Id. („In any case the significant variability observed between models in our study indicates that a definitive answer to the question of whether LLMs exhibit cognitivie limitations and biases cannot be drawn.“). 53 See Desai & Riedl, supra note 5, at 29; see also John J. Horton, Large language models as simulated economic agents: What can we learn from...

    work page 2023

  20. [20]

    Behavioural Economics, Consumer Policy, and Consumer Law

    221, 224-25; see also generally Peter Rott, Information Obligations and Withdrawal Rights, in Christian Twigg-Flesner (ed) The Cambridge Companion to European Union Private Law 187 (Cambridge 2010). 57 See, e.g. Samuel Issacharoff, Martin Engel & Johanna Stark, Buttons, Boxes, Ticks, and Trust: On the Narrow Limits of Consumer Choice, in Klaus Mathis (ed)...

    work page 2010

  21. [21]

    algorithmic consumers

    183; see also Omri Ben-Shahar & Carl E. Schneider, Coping with the failure of mandated disclosure 11 Jerusalem Review of Legal Studies 83-95 (2015). 59 See, e.g. Oren Bar-Gill, Defending (smart) disclosure: A comment on More than You Wanted to Know 11 Jerusalem Review of Legal Studies 75 (2015). 60 Ariel Porat and Lior Strahilevitz, Personalizing default ...

    work page 2015

  22. [22]

    Robust Physical-World Attacks on Deep Learning Models

    70 For an overview, see Camilla Crea and Alberto De Franceschi (eds), The New Shapes of Digital Vulnerability in European Private Law (Nomos 2024); see also Natali Helberger, Betül Kas, Hans-W. Micklitz, Monika Namysłowska, Laurens Naudts, Peter Rott, Marijn Sax, Michael Veale, Digital Fairness for Consumers, BEUC Report (May 2024), p. 12 (arguing that th...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1707.08945 2024

  23. [23]

    no barrier

    https://www.europeanlawinstitute.eu/fileadmin/user_upload/p_eli/Publications/ELI_Interim_Report_on_EU_Consumer_Law_and_Automated_Decision-Making.pdf. 79 See Art. 17 ELI DACC Model Rules („An algorithmic contract shall not be denied validity or enforceability solely because a digital assistant was used, irrespective of whether only one party or both partie...

    work page doi:10.48550/arxiv.2309.09640 2000

  24. [24]

    magical moment

    92 Cf. Zittrain, supra note 10 (warning of „potentially devastating consequences“ if no effective measures are introduced to contral AI agents now). 93 See Hans Schulte-Nölke, EC Law on the Formation of Contract – from the Common Frame of Reference to the ‘Blue Button’, 3(3) European Review of Contract Law, 332-349 (2007) (arguing that EU law exhibits a h...

    work page 2007

  25. [25]

    algorithmic consumers

    95 See Arts. 4-10 ELI DACC Model Rules (setting out design requirements for AI agents). 96 ELI DACC Model Rules, Introduction. 18 E. Conclusion Scholars have been arguing already for some time that the one-size-fits all model of consumer law no longer corresponds to the complexities of modern consumer markets and that a greater differentiation between dif...

    work page 2023

  26. [26]

    Bernard Marr, Is this AI’s iPhone moment?, Forbes (Jul

    98 Cf. Bernard Marr, Is this AI’s iPhone moment?, Forbes (Jul. 8, 2024), https://www.forbes.com/sites/bernardmarr/2024/07/08/is-this-ais-iphone-moment/

    work page 2024