Journal of FinTech and Sustainable Finance

Articles

Vol. 2 (2026)

The Demands of Financial Education in the Face of the Markets’ “Golden Calf”: The Role of Artificial Intelligence

DOI:
https://doi.org/10.31875/2755-8398.2026.02.07
Submitted
June 23, 2026
Published
2026-06-23

Abstract

The growing pervasiveness of artificial intelligence in financial markets is exposing retail investors to unprecedented forms of cognitive and informational asymmetry, as recently underscored by ESMA’s 2025 Warning on the use of AI-driven tools in investment decisions. The purpose of this article is to investigate the regulatory and dogmatic implications of this phenomenon at the intersection of the EU AI Act (Regulation 2024/1689), the MiFID framework, and the Italian Consolidated Law on Finance (TUF). The methodology adopted is dogmatic and comparative-regulatory: the notion of “AI system” is reconstructed as an organisational rather than a merely technological entity, and is then tested against the allocation of liability along the AI value chain (Article 25 AIA), the conceptual distinction between control and authority underpinning the deployer category, and the emerging paradigm of AI-as-a-service in financial intermediation. Particular attention is devoted to fin-influencers and large diffusion models (LDMs) as new vectors of systemic risk, and to the IOSCO 2025 Report’s recommendations on supervisory convergence. The article concludes that, where algorithmic opacity erodes the traditional pillars of investor protection, financial education must be reconceived as a constitutional precondition for the effective exercise of the economic freedoms; the principle of technological neutrality, far from being a passive regulatory stance, is reinterpreted as an active duty to safeguard the human dimension of financial choice — the actus humanus — against the seductive promises of algorithmic infallibility, which the article evokes through the metaphor of the markets’ “golden calf”. Read in this key, the contribution intersects the agenda of sustainable finance: responsible AI practices, algorithmic accountability and a renewed investor awareness emerge as preconditions for the long-term resilience of capital markets and for the integration of ESG-oriented decision-making into retail investment behaviour.

References

  1. This paper reproduces, with expansions and the addition of footnotes, the report presented at the Conference on “Financial education, digital transition and artificial intelligence”, held at the University of Sassari – Department of Economic and Business Sciences – Olbia campus, on 14 May 2025.
  2. See the latest Eurispes report, https://eurispes.eu/wp-content/uploads/2025/10/eurispes_2025_i-mercati-del-lavoro-regionali-di-fronte-alle-trasformazioni-digitali.pdf, which states that 58% of Italians have never used it.
  3. M. Ferraris, Documanità, Bari-Roma, 2021; M. Ferraris, G. Saracco, Tecnosofia. Tecnologia e umanesimo per una scienza nuova, Bari-Roma, 2025.
  4. See the recent study by S.J. Westwood, The potential existential threat of large language models to online survey research, in Proceedings of the National Academy of Sciences, 2025, vol. 122, no. 47, p. 1 ff., which highlights how AI has reached the point of being able to alter public opinion polls on a large scale, bypassing every quality control, by imitating real human beings and manipulating results without leaving any trace. https://doi.org/10.1073/pnas.2518075122
  5. On this point, among many and limiting ourselves to legal scholarship, see T.E. Frosini, L’ordine giuridico del digitale. The legal order of digital, in Giur. cost., 2023, 1, p. 377 ff.
  6. See C. Coltro Campi, La nuova disciplina dell’intermediazione e dei mercati mobiliari, Torino, 1991, p. 39.
  7. D. Maffeis, I contratti del mercato finanziario, in Trattato Cicu Messineo, Milano, 2024, p. 157.
  8. See Action Plan on Building a Capital Markets Union, Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions, Brussels, 20.09.2015, COM(2015) 468 final. As is well known, this document collects the results of a public consultation conducted by the European Commission following publication of a Green Paper entitled Building a Capital Markets Union [COM(2015) 63 final], Brussels, 18 February 2015, with the aim of identifying the actions necessary to improve access to financing for all businesses and infrastructure projects in Europe, to help SMEs raise financing as easily as large enterprises, to create a single market for capital by removing obstacles to cross-border investment, and to diversify the sources of financing of the economy and reduce the cost of raising capital. On the outcome of the consultation see European Commission, Feedback Statement on the Green Paper “Building a Capital Markets Union” accompanying the document Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions “Action Plan on Building a Capital Markets Union” [SWD(2015) 184 final], Brussels, 30 September 2015.
  9. ESMA Warning – Information sheet “Using artificial intelligence to invest: what you should bear in mind”, 25 March 2025, available at https://www.consob.it/documents/d/asset-library-1912910/esma_warning_ai_20250325_it. This document forms part of a broader awareness campaign promoted by ESMA through social media platforms, with the aim of helping European citizens make more informed and safer choices in the new digital context.
  10. It is recalled, for example, that public AI tools online are under no obligation to act in the best interest of the investor or to provide advice tailored to his or her personal situation.
  11. Among the first to highlight this problem, see F. Pasquale, The Black Box Society. The Secret algorithms That Control Money and Information, Harvard, 2015; but see also, more recently, M. Kosinski, What is black box artificial intelligence (AI)?, at https://www.ibm.com/think/topics/black-box-ai.
  12. Verbatim, the 2025 ESMA warning recalls that “if something were to go wrong, you may not have access to a financial ombudsman or alternative dispute resolution mechanisms to resolve complaints”.
  13. On this point, K. Martineau, What is retrieval-augmented generation?, 22 Aug. 2023, at https://research.ibm.com/blog/retrieval-augmented-generation-RAG; S. Jeong, J. Baek, S. Cho, S. Ju Hwang, J.C. Park, Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexity, 2024, at https://arxiv.org/abs/2403.14403. Among the better-known RAG systems are Perplexity (a chatbot-based search engine using GenAI) and AI chatbots that draw on databases primarily in the legal and medical fields.
  14. See I. Aldasoro, L. Gambacorta, A. Korinek, V. Shreeti, M. Stein, Intelligent financial system: how AI is transforming finance, BIS Working Papers No. 1194, Jun. 2024, p. 7 ff., available at https://www.bis.org/publ/work1194.pdf.
  15. Indeed, the term “AI Agents” reflects AI systems endowed with ever greater autonomy. A. Chan, C. Ezell, M. Kaufmann, K. Wei, L. Hammond, et al., Visibility into AI Agents, 2024, at https://arxiv.org/abs/2401.13138, define agency as “the degree to which an AI system acts directly in the world to achieve long-horizon goals, with little human intervention or specification of how to do so. An (AI) agent is a system with a relatively high degree of agency; we consider systems that mainly predict without acting in the world, such as image classifiers, to have relatively low degrees of agency”.
  16. Suffice it to refer to the August 2025 Forrester study commissioned by AWS Marketplace (How Financial Services Leaders Are Approaching Security And Innovation. The Advancing Role Of Cloud-Native Systems And Agentic AI), available at https://pages.awscloud.com/rs/112-TZM-766/images/Forrester_AWS%20Marketplace_How%20Financial%20Services%20Leaders%20Are%20Approaching%20Security%20And%20Innovation.pdf?version=1, which shows how agentic AI is radically transforming financial services, currently based on the use of chatbots; in particular, three areas are identified in which agentic AI is transforming financial services, creating what industry experts call autonomous finance or services based on algorithms capable of taking financial decisions on customers’ behalf: AI-powered customer service (account management, processing of loan applications, dispute resolution); AI in financial operations (AI agents can analyse market conditions, adjust risk parameters in real time and optimise everything from trading strategies to compliance monitoring, not merely following rules but taking “intelligent” decisions based on continuously evolving data); hyper-personalised financial advice (an AI agent can help customers monitor their financial situation, looking for opportunities to save money, optimise investments or detect potential risks).
  17. See H. Naveed, A.U. Khan, S. Qiu, M. Saqib, S. Anwar, M. Usman, N. Akhtar, N. Barnes, A. Mian, A Comprehensive Overview of Large Language Models, 2023, at https://doi.org/10.48550/arXiv.2307.06435.
  18. By token is meant words or parts of words.
  19. These are the numerical values that control the behaviour of the model and are “tuned” through the use of the training data. The number of LLM parameters has grown over time: while version 2 of the Generative Pre-trained Transformer (GPT-2) had 1.5 billion parameters, the Pathways Language Model (PaLM) reached 540 billion parameters, triggering a competitive race (restricted exclusively to the IT big players) to build the most powerful model. The release of ChatGPT had shifted the competition from the field of mere research to that of products (also affecting market shares, stock-market valuations and corporate reputation).
  20. See N. Cristianini, Machina sapiens. L’algoritmo che ci ha rubato il segreto della conoscenza, Bologna, 2024, p. 86 ff.
  21. Among many, see C. Stryker, M. Scapicchio, What is generative AI?, at https://www.ibm.com/think/topics/generative-ai.
  22. See I. Aldasoro, L. Gambacorta, A. Korinek, V. Shreeti, M. Stein, Intelligent financial system, cit., p. 10.
  23. See S.C. Matz, J.D. Teeny, S.S. Vaid, H. Peters, G.M. Harari, M. Cerf, The potential of generative AI for personalized persuasion at scale, in Scientific Reports, 2024, 14, 4692. https://doi.org/10.1038/s41598-024-53755-0
  24. See K. Huang, X. Chen, Y. Yang, J. Ponnapalli, G. Huang, ChatGPT in Finance and Banking, in K. Huang, Y. Wang, F. Zhu, X. Chen, C. Xing (eds.), Beyond AI. ChatGPT, Web3, and the Business Landscape of Tomorrow, Cham, 2023, p. 187 ff. https://doi.org/10.1007/978-3-031-45282-6_7
  25. See C. Waghmare, Unleashing The Power of ChatGPT: A Real World Business Applications, Mumbai, 2023, p. 45 ff.; A. Santangelo, A. Sissa, M. Borghi, Critica di ChatGPT, Milano, 2025, p. 21 ff. https://doi.org/10.1007/979-8-8688-0032-0_3
  26. On this point, see A. Hüsch, D. Distelrath, T. Hüsch, Applications of GPT in Finance, Compliance, and Audit, Wiesbaden, 2024, p. 31 ff.
  27. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act).
  28. For an initial analysis, see R. Petruso, G. Smorto, Il Regolamento europeo sull’intelligenza artificiale: una prima lettura, in Nuova giur. civ. comm., 2024, 989 ff.; see also A. Mantelero, G. Resta, G.M. Riccio (eds.), Intelligenza artificiale. Commentario, Milano, 2025; O. Pollicino, F. Donati, G. Finocchiaro, F. Paolucci (eds.), La disciplina dell’intelligenza artificiale, Milano, 2025.
  29. For a diachronic overview, see G. Proietti, Definire l’indefinibile? I sistemi di intelligenza artificiale alla ricerca di un inquadramento sistematico, in Contr. impr., 2024, 882 ff.
  30. The OECD, in a 2022 recommendation document to the European Council on AI (OECD, Recommendation of the Council on Artificial Intelligence, OECD/LEGAL/0449), defines an AI system as “An AI system is a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions influencing real or virtual environments. AI systems are designed to operate with varying levels of autonomy”.
  31. See European Parliament resolution of 20 October 2020 with recommendations to the Commission on a framework of ethical aspects of artificial intelligence, robotics and related technologies (2020/2012(INL)), § 12.
  32. See recital 4: “AI is a fast-evolving family of technologies that contributes to a wide array of economic, environmental and societal benefits across the entire spectrum of industries and social activities. By improving prediction, optimising operations and resource allocation, and personalising digital solutions available for individuals and organisations, the use of AI can provide key competitive advantages to undertakings and support socially and environmentally beneficial outcomes, for example in healthcare, agriculture, food safety, education and training, media, sports, culture, infrastructure management, energy, transport and logistics, public services, security, justice, resource and energy efficiency, environmental monitoring, the conservation and restoration of biodiversity and ecosystems and climate change mitigation and adaptation”.
  33. Article 3, point 12), AIA.
  34. See Article 3, point 63), AIA: “‘general-purpose AI model’ means an AI model, including where such an AI model is trained with a large amount of data using self-supervision at scale, that displays significant generality and is capable of competently performing a wide range of distinct tasks regardless of the way the model is placed on the market and that can be integrated into a variety of downstream systems or applications, except AI models that are used for research, development or prototyping activities before they are placed on the market;” see also Article 3, point 66), AIA: “‘general-purpose AI system’ means an AI system which is based on a general-purpose AI model and which has the capability to serve a variety of purposes, both for direct use as well as for integration in other AI systems;”.
  35. On this point, see R. Petruso, G. Smorto, Il Regolamento europeo sull’intelligenza artificiale, cit., 991. It should be added that the notion of AI must be read in the perspective of the problem the AIA seeks to address, namely a problem of risk management and consequently of activating accountability mechanisms. Following a “risk-based” approach, under which the higher the risk the stricter the rules, the AIA establishes obligations for providers and operators of AI systems depending on the level of risk that the AI may generate: (i) unacceptable risk; (ii) high risk; (iii) low or minimal risk. Specific transparency obligations are also set out. On this point, see S. Orlando, Gli emendamenti alla proposta di AI Act approvati dal Parlamento europeo il 14.6.2023, in Pers. merc., 2023, 378 ff.; Id., Linguaggi di programmazione e responsabilità, in V.V. Cuocci, F.P. Lops, C. Motti (eds.), La responsabilità civile nell’era digitale (Atti della Summer school 2021), Bari, 2022, 139 ff.; Id., Regole di immissione sul mercato e «pratiche di intelligenza artificiale» vietate nella proposta di Artificial Intelligence Act, in Pers. merc., 2022, 343 ff. See also the recent monographs by A. Bertolini, Intelligenza Artificiale e Responsabilità civile. Problema, sistema, funzioni, Bologna, 2024, 88 ff.; T. De Mari Casareto dal Verme, Intelligenza artificiale e responsabilità civile. Uno studio sui criteri di imputazione, Trento, 2024, 176 ff., 207 ff. and 319 ff.
  36. F. Astone, Intelligenza artificiale e diritto civile, in V.V. Cuocci, F.P. Lops, C. Motti (eds.), La circolazione della ricchezza nell’era digitale, Atti della Summer School 2020, Pisa, 2021, 9.
  37. A. Altieri, L’Artificial Intelligence come organizzazione. Profili giuridici, Torino, 2024, p. 57 ff.
  38. J. Drexl et al., Artificial Intelligence and Intellectual Property Law. Position Statement of the Max Planck Institute for Innovation and Competition of 9 April 2021 on the Current Debate, in Max Planck Institute for Innovation and Competition Research, 2021, Paper No. 21-10, p. 18 ff.; https://doi.org/10.2139/ssrn.3822924
  39. D. Kim et al., Artificial Intelligence Systems as Inventors? A Position Statement of 7 September 2021 in view of the evolving case-law worldwide, in Max Planck Institute for Innovation and Competition Research Paper, 2021, No. 21-20, passim.
  40. Indeed, the heading of Article 25 AIA reads: “Responsibilities along the AI value chain”; on which point see below.
  41. Even from a mere reading of the definitions, one becomes aware of the actors potentially involved in the network unfolded by the AI system. Article 3, point 3), AIA, refers to the “provider”, that is a natural or legal person, public authority, agency or other body that develops or has developed an AI system or a general-purpose AI model and places that system or model on the market, or puts the AI system into service under its own name or trademark, whether for payment or free of charge. Article 3, point 4), AIA, defines the “deployer”, that is a natural or legal person, public authority, agency or other body using an AI system “under its authority”. Article 3, point 5), AIA, defines the “authorised representative”, namely a natural or legal person located or established in the Union who has received and accepted a written mandate from a provider of an AI system or general-purpose AI model in order, respectively, to fulfil and carry out on its behalf the obligations and procedures set out in the AIA. Then, point 6) sets out the concept of importer (“a natural or legal person located or established in the Union that places on the market an AI system that bears the name or trademark of a natural or legal person established in a third country”), point 7) defines “distributor” (“a natural or legal person in the supply chain, other than the provider or the importer, that makes an AI system available on the Union market”) and point 8) clarifies “operator” (“a provider, product manufacturer, deployer, authorised representative, importer or distributor”). To this set of subjects must be added the data trainers and data vendors, that is the holders of rights over the training data, and (to mention but a few) the operators of platforms and communication infrastructures, cloud service providers and API developers.
  42. Article 3, point 9), AIA: “the first making available of an AI system or a general-purpose AI model on the Union market”.
  43. Article 3, point 10), AIA: “the supply of an AI system or a general-purpose AI model for distribution or use on the Union market in the course of a commercial activity, whether in return for payment or free of charge”.
  44. Article 3, point 11), AIA: “the supply of an AI system for first use directly to the deployer or for own use in the Union for its intended purpose”.
  45. A. Zoppini, L’informazione come bene, in M. D’Auria (ed.), I problemi dell’informazione nel diritto civile, oggi. Studi in onore di Vincenzo Cuffaro, Roma, 2022, 77.
  46. On this point see N. Cristianini, Machina sapiens, cit., passim; S.J. Russell, P. Norvig, Artificial Intelligence. A Modern Approach, IV ed., Harlow, 2022, passim.
  47. See A. Altieri, L’Artificial Intelligence come organizzazione, cit., 1 ff.
  48. See, although in a different sense, R. Franceschelli, L’imprenditore nel nuovo codice civile, Torino, 1943, 105; in a not dissimilar sense F. Ferrara jr., La teoria giuridica dell’azienda, (reprint of II ed.), Milano, 1982, 113.
  49. See R. Müller-Erzbach, Deutsches Handelsrecht, Erster Teil, Tübingen, 1921, p. 65 ff.; but also F. Ferrara jr., La teoria giuridica dell’azienda, cit., 59.
  50. See, among many, G. Racugno, Azienda (voce), in Enciclopedia Treccani.it, 2015; M. Casanova, Azienda (voce), in Dig. disc. priv., sez. comm., II, Torino, 1987, 75 ff., esp. 77; G. Ferrari, Azienda (dir. priv.) (voce), in Enc. dir., IV, Milano, 1959, 680 ff.; G. Auletta, Azienda. I) Diritto commerciale (voce), in Enc. giur., Roma, 1988, 1 ff.; G.E. Colombo, L’azienda, in F. Galgano (ed.), Tratt. dir. comm. e dir. pubbl. dell’econ., III, Padova, 1979, 1 ff.; Id., Il trasferimento dell’azienda e il passaggio dei crediti e dei debiti, Padova, 1972, 12 ff.; G. Racugno, Lo “scorporo” d’azienda, Milano, 1995, 9 ff.
  51. Thus A. Iannarelli, Affitto di fondo rustico e affitto di azienda agraria, in Riv. dir. agr., 1991, 445. But see also C. Motti, Il mercato come organizzazione, in Banca, impr. e società, 1991, 455 ff., esp. 473, where the Author highlights the typically incremental dimension of organisation: “by ‘organisation’ one must not understand only the setting up and management of a series of equipment and technical services […], but rather that minimum of structures and/or discipline (as a rule, a combination of both) needed to confer on exchanges a degree of efficiency higher than that expressed by the ‘spontaneous market’, given the specificities of each form of market”.
  52. See A. Amaduzzi, L’azienda nel suo sistema e nell’ordine delle sue rilevazioni, Torino, 1969, 20.
  53. Already with regard to the “subjective” side of organisation in the field of enterprise and activity, the limits of law as an explanatory tool have been noted: on this point, particularly incisive, V. Buonocore, L’impresa, in Id. (ed.), Trattato di Diritto Commerciale, sec. I, t. 2.1, Torino, 2002, 111 ff.
  54. See in this sense N. Luhmann, Introduzione alla teoria dei sistemi (It. trans.), Lecce, 2018, esp. 89 ff. For early — albeit altogether primitive — references to organisation and cybernetics, see N. Wiener, Introduzione alla cibernetica. L’uso umano degli essere umani (It. trans.), Torino, 1966, 140 ff.
  55. N. Luhmann, Organizzazione e decisione (It. trans.), Milano, 2005, 29 ff.
  56. Ibid., 47 ff.
  57. For a more thorough analysis of the interpretive steps, see A. Altieri, L’Artificial Intelligence come organizzazione, cit., 146 ff. and 159 ff.
  58. A. Beckers, G. Teubner, Human–algorithm hybrids as (quasi-)organizations? On the accountability of digital collective actors, in J. Law Soc., 2023, 50, 100 ff.; https://doi.org/10.1111/jols.12412
  59. Beckers, G. Teubner, Three Liability Regimes for Artificial Intelligence. Algorithmic Actants, Hybrids, Crowds, Oxford et al., 2021. In particular, treating the collective actor as a network, in order to overcome the hierarchical and institutional principle, (quasi-)organisation theory tends towards a distributed collective liability, in which individual and collective orientations are simultaneously institutionalised. This theory, however, in addition to considering attribution criteria in terms of limited legal personality, certainly shifts attention towards the network, but for that very reason ends up emphasising this aspect, institutionalising it and not departing from personalist theories. https://doi.org/10.5040/9781509949366
  60. M.E. Porter, Competitive advantage: creating and sustaining superior performance, New York, 1985; O.E. Williamson, Economic Organization: Firms, Markets and Policy Control, Brighton, 1986.
  61. V. Sgroi, L’invenzione non brevettata, Milano, 1961; A. Ottolia, Big Data e innovazione computazionale, cit., p. 43 ff.; A. Bianchi, Regime del segreto e tecniche dipendenti, Torino, 1990; F. Massa Felsani, Contributo all’analisi del Know-How, Milano, 1997; M. Libertini, Le informazioni aziendali segrete come oggetto di diritti di proprietà industriale, in Riv. it. sc. giur., 2011, 2, p. 137 ff., and the further extensive bibliography there cited on trade secrets and know-how in Italian and comparative scholarship (Galli, Ghidini, Falce, Mansani, Traverso, Giudici, Moro Visconti, Arezzo, Bonelli, Gambino, Mastrelia, Ciccone-Ghini, Barbera, Magelli, Venturello, Franchini Stufler, Vanzetti, Sena, Scarpa, Banterle-Blei, Crespi, Auteri, Ottolia, Blandini, Frignani, Ruffolo, Sarti, Civello, Gualtieri, Crespi, Kastoris, Mazzacuva, Giavazzi). See also T. Aplin, Right to property and trade secrets, in C. Geiger (ed.), Research Handbook on Human Rights and Intellectual Property, Cheltenham, UK – Northampton, MA, USA, 2015, p. 421 ff.; W. van Caenegem, Trade Secrets and Intellectual Property. Breach of Confidence, Misappropriation and Unfair Competition, Alphen aan den Rijn, 2014.
  62. P. Perlingieri, L’informazione come bene giuridico, in Rass. dir. civ., 1990, 326 ff., esp. 338 ff., who uses it as a juridical paraphrase of the title of the right over the secret to justify the exercise of bundles of generally homogeneous rights by a plurality of subjects, in line with the theory of the legal good.
  63. M.S. Spolidoro, Il concetto di controllo nel codice civile e nella legge antitrust, in Riv. soc., 1995, 457 ff.
  64. L. Acciari, F. Salerno, G. Visentini, Il controllo delle imprese nella legislazione italiana. Alla ricerca di una nozione comune, Pisa, 2023.
  65. M. Lamandini, Il “controllo”. Nozioni e “tipo” nella legislazione economica, Milano, 1995, 51 ff., where, despite the effort to counter the “atomistic” tendency in theories on control, in reconstructing the type the differences are nonetheless drawn out.
  66. P. Marchetti, Note sulla nozione di controllo nella legislazione speciale, in Riv. soc., 1992, 1 ff.
  67. M. Notari, La nozione di “controllo” nella disciplina antitrust, Milano, 1996, 186.
  68. And this premise is conferred precisely by antitrust legislation which, with regard to so-called “objective” control, departs from the classical relational scheme and aims, from a competition perspective, at identifying the productive means available to an undertaking or a group of undertakings. On this point see M. Notari, La nozione di “controllo”, cit., p. 252 ff.; M. Libertini, Diritto della concorrenza dell’Unione Europea, Milano, 2014, 343 ff.
  69. See most recently the conclusions of G. Visentini, F. Salerno, Considerazioni conclusive, in L. Acciari, F. Salerno, G. Visentini (eds.), Il controllo delle imprese nella legislazione italiana. Alla ricerca di una nozione comune, Pisa, 2023, 375 ff.
  70. The vast subject arose, as is well known, in an attempt to grasp the connecting links in the theory of corporate groups: see L. Acciari, Introduzione, in L. Acciari, F. Salerno, G. Visentini (eds.), Il controllo delle imprese nella legislazione italiana. Alla ricerca di una nozione comune, Pisa, 2023, XIII ff.; but also M. Lamandini, art. 2359 c.c., in P. Abbadessa, G.B. Portale (eds.), Le società per azioni. Codice civile e norme complementari, t. 1, Milano, 2016, 743 ff., esp. 757.
  71. M. Notari, La nozione di “controllo”, cit., pp. 241-242.
  72. Ibid., p. 242.
  73. See A. Altieri, L’Artificial Intelligence come organizzazione, cit., 135 ff.
  74. Just as occurs for controlled and affiliated companies under Article 2359 of the Italian Civil Code, or as occurs in antitrust law when reconstructing market relationships.
  75. A. Zoppini, L’informazione come bene, cit., 69 ff.
  76. Even those who attempt to set out the dogmatic discourse on information through the theory of the legal good are forced to invoke the contract: see P. Perlingieri, L’informazione come bene giuridico, cit., 339-340.
  77. See also A. Zoppini, L’informazione come bene, cit., 74.
  78. The parallel as an object of inquiry between enterprise and AI systems may be permitted, since at least from a speculative point of view both phenomena seem to share a quite similar opening phase: the literature shows that in 1942, with regard to the “enterprise”, there were the same conceptual uncertainties that still surround AI today.
  79. R. Nicolò, Riflessioni sul tema dell’impresa, in S. Rodotà (ed.), Il diritto privato nella società moderna, Bologna, 1971, 412 ff. (and previously Id., Riflessioni sul tema dell’impresa e su talune esigenze di una moderna dottrina del diritto civile, in Riv. dir. comm., 1965, I, 177 ff.); contra, G. Auletta, Impresa e azienda, in Temi nap., 1958, III, 23 ff.; G. Ferri, Manuale di diritto commerciale (eds. C. Angelici and G.B. Ferri), XV ed., Torino, 2016, 33 ff.; G. Minervini, L’imprenditore. Fattispecie e statuti, Napoli, 1970, 135 ff.; G. Santini, Le teorie sull’impresa (civilisti e laburisti a confronto), in Riv. dir. civ., 1970, 422 ff.; V. Panuccio, Teoria giuridica dell’impresa, Milano, 1974, 78 ff. Some authors have to some extent developed Nicolò’s line of thought: see P. Rescigno, Per una strada sulla proprietà, in Riv. dir. civ., 1961, 61 ff.; S. Rodotà, Rapporti privati e leggi di nazionalizzazione, in Riv. dir. comm., 1969, 96 ff.; O.T. Scozzafava, Rosario Nicolò e il diritto di impresa, in Riv. dir. comm., 2008, 847 ff.; M. Tanzi, Godimento del bene produttivo e impresa, Milano, 1998, 75 ff.
  80. R. Nicolò, Riflessioni sul tema dell’impresa, cit., 415.
  81. Ibid., 419.
  82. A concept which, on the other hand, escaped the doctrine of the time and matured with the change of modern capitalism. On this point, see extensively D.S. Landes, J. Mokyr, W.J. Baumol (eds.), The Invention of Enterprise. Entrepreneurship from Ancient Mesopotamia to Modern Times, Princeton, 2010, passim (and especially the essay by J.M. Murray, Entrepreneurs and Entrepreneurship in Medieval Europe, ibid., 88 ff.); and in some respects also H. Hansmann, The Ownership of Enterprise, Cambridge-London, 1996, 11 ff.
  83. L. Enriques, D.A. Zetzsche, Corporate Technologies and the Tech Nirvana Fallacy (March 25, 2020), European Corporate Governance Institute (ECGI) – Law Working Paper No. 457/2019, Hastings Law Journal, Forthcoming, available at SSRN: https://ssrn.com/abstract=3392321. https://doi.org/10.2139/ssrn.3392321
  84. “The notion of ‘deployer’ referred to in this Regulation should be construed as any natural or legal person, including a public authority, agency or other body, using an AI system under its authority”.
  85. C.M. Bianca, Le autorità private, Napoli, 1977, 4.
  86. See S.J. Shapiro, Authority, in J.L. Coleman, K.E. Himma, S.J. Shapiro (eds.), The Oxford Handbook of Jurisprudence and Philosophy of Law, Oxford, 2004, 383 ff.;
  87. K.E. Himma, The Nature of Authority, Cambridge, 2024, 3 ff.; https://doi.org/10.1017/9781009255790
  88. J. Raz, The Problem of Authority: Revisiting the Service Conception, in Minnesota Law Review, 2006, 1003 ff.; https://doi.org/10.24926/265535.919
  89. W.R.P. Kaufman, Beyond Legal Positivism: The Moral Authority of Law, Lowell (MA, USA), 2023, 53 ff.; https://doi.org/10.1007/978-3-031-43868-4_3
  90. J.R. Graham, C.R. Harvey, M. Puri, Capital Allocation and Delegation of Decision Making Authority within Firms, in Journal of Financial Economics, 2015, vol. 115, is. 3, 449 ff. https://doi.org/10.1016/j.jfineco.2014.10.011
  91. And there is talk of control with reference to privacy. See generally S. Rodotà, Tecnologie e diritti, Bologna, 1995, 122.
  92. Although in a different context from that under examination, see A. Astone, Situazioni di fatto e schemi legali. La delimitazione della categoria, Milano, 2017, 11 ff. and, on contractual relationships, 17 ff.
  93. Who carries out a professional activity which, as is well known, in the language of the EU legislator denotes the exercise of an economic activity. Furthermore, the concept of authority as a derived power (and note the parallelism with the GDPR) is contrasted with that of control, that is the ownership of the data, which remains with the data subject, while the data controller acquires only the authority to determine the purposes thereof.
  94. But on this point, see below.
  95. The reference is to A. Romano, Le concessioni dei posti di vendita nei mercati all’ingrosso, in L’indennità di espropriazione. I mercati all’ingrosso. Atti del XVI Convegno di Studi di Scienza dell’Amministrazione, Varenna – Villa Monastero, 17-20 settembre 1970, Milano, 1972, 209 ff.; on the organisational function of stock-market markets, see C. Motti, Mercati borsistici e diritto comunitario, Milano, 1997, 47 ff., and earlier Ead., Il mercato come organizzazione, cit., 455 ff., esp. 468 ff.
  96. See O.E. Williamson, L’organizzazione economica. Imprese, mercati e controllo politico (It. trans.), Bologna, 1991, 135 ff., and the bibliography there cited.
  97. On the function that the contract assumes today, also (and above all) in a regulatory capacity, see V. Roppo, Il contratto del duemila, IV ed., Torino, 2020, passim; but see also AIA, recital no. 90 and Article 56(2)(d).
  98. On this point, with broader references, see A. Altieri, L’Artificial Intelligence come organizzazione, cit., 197 ff.
  99. For an initial confirmation in this sense, see F. Pacileo, Intelligenza Artificiale nell’impresa. Tra organizzazione e spersonalizzazione, Milano, 2025, who frames the problem in terms of “technological organisation” and “articulation of the organisation” of an entrepreneurial activity (the system’s outputs being “acts pertaining to the exercise of the enterprise”, and AI systems being identified as a criterion both for attributing effects (which would fall on the entrepreneur because they are the result of a collective and depersonalised action, comparing AI to an auxiliary of the entrepreneur) and for regulating liability (“the joint-stock enterprise that uses AI systems on a cost-benefit logic responds according to the criteria of liability for enterprise risk — depending on the case, aggravated or strict, but always depersonalised — for damages caused by the malfunctioning of the AI systems”)).
  100. Generally, P. Ferro-Luzzi, I contratti associativi, Milano, 1976; G. Auletta, Attività (dir. priv.) (voce), in Enc. dir., Vol. III, Milano, 1958, p. 981 ff.; M.S. Giannini, Attività amministrativa (voce), in Enc. dir., Vol. III, Milano, 1958, p. 988 ff.
  101. Among many, see P. Rescigno, Obbligazioni (diritto privato) (voce), in Enc. dir., Vol. XXIX, Milano, 1979, p. 133 ff., esp. p. 190; C.M. Bianca, Diritto civile. L’obbligazione, vol. 4, Milano, 1993, p. 112 ff.; more recently, A. Nicolussi, Le obbligazioni, Milano, 2021, p. 37.
  102. This observation makes it possible to deepen the discussion on the nature of the output of such an organised activity. It is well known that a fundamental distinction within obligations to do (obbligazioni di fare) relates to the object of the obligation, namely whether it consists in the performance of a work or service as a result of the activity, or, conversely, in the activity itself, irrespective of the outcome and the perfection of an opus. As this is not the place to elaborate on this aspect, also given its evident implications in terms of contractual liability, the discussion is referred to a forthcoming update of the monograph by A. Altieri, L’Artificial Intelligence come organizzazione, cit.
  103. J. Wirtz, Ch. Lovelock, Services Marketing. People, Technology, Strategy, Hackensack, NJ, 2021, p. 16 ff.: “Services are economic activities performed by one party to another. Often time-based, these performances bring about desired results to recipients, objects, or other assets. In exchange for money, time, and effort, service customers expect value from access to labor, skills, expertise, goods, facilities, networks, and systems. However, they do not normally take ownership of the physical elements involved” (p. 18); the Authors continue: “We define services as economic activities between two parties, implying an exchange of value between the seller and buyer in the marketplace. We describe services as performances that are most commonly time-based. We emphasize that purchasers buy services because they are looking for desired results. In fact, many firms explicitly market their services as ‘solutions’ to prospective customers’ needs. And finally, our definition emphasizes that while customers expect to obtain value from their service purchases in exchange for their money, time, and effort, this value comes from access to a variety of value-creating elements rather than transfer of ownership”.
  104. G. Santini, Commercio e servizi. Due saggi di economia del diritto, Bologna, 1988, p. 420: “the service does not come into relevance in its ‘static’ phase, that is as an element of the assets of the one who is to receive or render it; but rather in its ‘dynamic’ phase: it is the performance of activity, it is ‘activity’ tout court, and this implies a close link with the subject who carries it out”.
  105. G. Ferri, Le società, in Trattato di diritto civile, founded by F. Vassalli, X, Torino, 1988, p. 15.
  106. F. Nieddu Arrica, Il conferimento di prestazione d’opera e servizi nella s.r.l., Milano, 2009, p. 33 ff., who emphasises the creative-productive value of services and, on the subjective side, their attribution to intellectual professionals or entrepreneurs.
  107. Financial Stability Institute, Regulating AI in the financial sector: recent developments and main challenges, FSI Insights on policy implementation no. 63, December 2024, https://www.bis.org/fsi/publ/insights63.pdf; ESMA, Artificial intelligence in EU securities markets, ESMA50-164-6247, February 2023; IOSCO, The use of artificial intelligence and machine learning by market intermediaries and asset managers, Final Report, September 2021; FSB, Artificial intelligence and machine learning in financial services. Market developments and financial stability implications, November 2017; OECD, Artificial Intelligence, Machine Learning and Big Data in Finance, August 2021; OECD, Generative artificial intelligence in finance, December 2023; N. Linciano, V. Caivano, D. Costa, P. Soccorso, T.N. Poli, G. Trovatore, L’intelligenza artificiale nell’asset e nel wealth management, in Quaderni FinTech Consob, June 2022.
  108. See G. Serafin, Intelligenza artificiale, attività e responsabilità dell’impresa di investimento, Torino, 2025, p. 53 ff., who notes the proximity between AI systems and the provision of investment services.
  109. A. Blandini, M. Gigliotti, Fintech e innovazione digitale, prospettive applicative nella liquidazione coatta amministrativa, in Banca borsa tit. cred., 2024, 5, p. 736.
  110. It is well known that the term RegTech denotes the application of technological innovation in support of controls over regulated intermediaries; while the term SupTech refers to the use of technologies that support the external supervision of supervisory authorities. On this point, see extensively L. Enriques, Financial Supervisors and Regtech: Four Roles and Four Challenges (December 13, 2017), Revue Trimestrielle de Droit Financier 53 (2017), available at SSRN: https://ssrn.com/abstract=3087292; S. Zeranski, I.E. Sancak, Digitalisation of Financial Supervision with Supervisory Technology (SupTech) (August 1, 2020), Journal of International Banking Law & Regulation, available at SSRN: https://ssrn.com/abstract=3632053, where the Authors state that “SupTech is the name of FinTech when it is used for supervisory purposes”.
  111. The reference is to Regulation (EU) 2024/1624 of 31 May 2024 on the prevention of the use of the financial system for the purposes of money laundering or terrorist financing (the so-called AMLR), Regulation (EU) 2024/1620 of 31 May 2024 establishing the Authority for Anti-Money Laundering and Countering the Financing of Terrorism (the so-called AMLAR), and the sixth Directive (EU) 2024/1640 of 31 May 2024 on the mechanisms to be put in place by Member States for the prevention of the use of the financial system for money laundering or terrorist financing (the so-called AMLD6). See, among many, A. Minto, Navigating AML Requirements for Payment Initiation Service Providers (PISPs): The Challenging Fulfilment of Customer Due Diligence, in EBLR, 2025, p. 815 ff.
  112. See Article 5(5)(i), AMLAR (“5. The Authority shall carry out the following tasks regarding FIUs and their activities in the Member States: […] (i) it shall develop and make available to FIUs tools and services to enhance their analytical capabilities, as well as IT and artificial intelligence services and tools for the secure sharing of information, including by hosting FIU.net;”); but see also recital 9, AMLAR (“The Authority’s powers aim at enabling it to enhance AML/CFT supervision in the Union in various ways. […] Overall, the Authority should contribute to the convergence of supervisory practices and the promotion of high supervisory standards. The Authority should also coordinate and support the conduct of joint analyses by FIUs, or request the launching of joint analyses, and make available to FIUs IT and artificial intelligence services to strengthen their data analysis capabilities, as well as tools for the secure sharing of information, including by hosting FIU.net, the dedicated IT system that allows FIUs to cooperate and exchange information among themselves and, where appropriate, with their counterparts in third countries and with third parties”).
  113. See G. Schneider, La proposta di regolamento europeo sull’intelligenza artificiale alla prova dei mercati finanziari: limiti e prospettive (di vigilanza), in Resp. civ. prev., 2023, 3, p. 1023, who laments the absence of specific rules on the design and use of AI systems applied to AML/CFT.
  114. On this topic, without claim to exhaustiveness, see C. Brescia Morra, D. Colonnello, M. Gargantini, G. Sandrelli, G. Trovatore, La gamification degli investimenti finanziari, Quaderni giuridici Consob, no. 32, January 2025, p. 25 ff., p. 47 ff.; N.M.F. Faraone, Quando le piattaforme (anche social) incontrano la divulgazione finanziaria: appunti sparsi su “Fintok”, IA e funzioni di vigilanza, in Rivista di Diritto Bancario, suppl. fasc. 4, 2023, p. 163 ff.; N. Aggarwal, D.B.V. Kaye, C. Odinet, #Fintok and Financial Regulation, in Arizona State Law Journal, 2023, p. 1036 ff.; A. Canepa, Social media e fin-influencers come nuove fonti di vulnerabilità digitale nell’assunzione delle decisioni di investimento, in Rivista trimestrale di diritto dell’economia, suppl. 1/2022, p. 307 ff.; S. Guan, The Rise of the Finfluencer (December 1, 2022), 19 New York University Journal of Law and Business 489 (2023), Santa Clara Univ. Legal Studies Research Paper No. 4400042, available at SSRN: https://ssrn.com/abstract=4400042; see also D. De Filippis, La consulenza finanziaria prestata tramite gli influencers: spunti per un inquadramento della fattispecie, in ELFR, p. 106 ff.
  115. See IOSCO, Artificial Intelligence in Capital Markets, cit., p. 20 ff.
  116. See I. Demuro, Trasparenza e correttezza dell’“influencer marketing”, in Analisi Giuridica dell’Economia, 2025, pp. 12 and 13.
  117. On this point see the reflections of D. De Filippis, La consulenza finanziaria prestata tramite gli influencers, cit., p. 113 ff., who, with the proper systemic and interpretive caveats, concludes by assimilating the activity carried out via fin-influencers to financial advice. The aspect mentioned here has, in other respects, been the subject of broad debate concerning the phenomenon of robo-advice: for a synthesis, see N. Linciano, V. Caivano, D. Costa, P. Soccorso, T.N. Poli, G. Trovatore, L’intelligenza artificiale nell’asset e nel wealth management, in Quaderni FinTech Consob, no. 9, June 2022; F. Sartori, La consulenza finanziaria automatizzata: problematiche e prospettive, in Riv. trim. dir. econ., 2018, p. 256 ff.; C. Picciau, La consulenza finanziaria automatizzata, in M. Cian, C. Sandei (eds.), Diritto del Fintech, II ed., Padova, 2024, p. 417 ff.; F. Accettella, Gli strumenti di robo gestione di patrimoni e le DAO, ibid., p. 475 ff.; M.T. Paracampo, Robo-advisor, consulenza finanziaria e profili regolamentari: quale soluzione per un fenomeno in fieri?, in Riv. trim. dir. econ., suppl. to no. 4 of 2016, p. 256 ff.; Ead., La consulenza finanziaria automatizzata, in M.T. Paracampo (ed.), Fintech. Introduzione ai profili giuridici di un mercato unico tecnologico dei servizi finanziari, Torino, 2017, 127; Ead., L’adeguatezza della consulenza finanziaria automatizzata nelle linee guida dell’ESMA tra algo-governance e nuovi poteri di supervisione, in Riv. dir. bancario, 2018, p. 535 ff.
  118. On this scenario, already within reach of the major players in financial markets, see the Assogestioni White Paper (ed. R. D’Apice), AI nell’asset management: dalla visione all’azione. Strategie, policy e nuove prospettive nel risparmio gestito italiano, June 2025, available at https://www.assogestioni.it/sites/default/files/docs/assogestioni_whitepaper_ai_nellasset_management_giugno2025_ita_0.pdf.
  119. FSB (2024), The Financial Stability Implications of Artificial Intelligence, available at https://www.fsb.org/uploads/P14112024.pdf. Industry reports indicate that financial firms are experimenting with the use of LLMs for client assistance and training. Examples reported in public documentation include chatbots that allow clients to ask questions about corporate documents, news and historical prices, and that allow institutional clients to ask for the assessment of hypothetical transactions or to execute them. Another application consists in assisting customer-service employees by training LLMs on internal documents relating to policies and operations, so that customer-service employees can interact with the chatbot when responding to customer inquiries. Industry reports also indicate that firms are using LLMs to generate marketing material, including both the content of investment solicitations and investment-banking activities such as creating presentations, summarising sector knowledge, and highlighting key sales points. LLMs would be used to assist staff in generating alerts and themes for client marketing. Companies plan to use LLMs to assist in customer segmentation, profiling and personalisation for marketing. On this point, see again IOSCO, Artificial Intelligence in Capital Markets, cit., p. 27; but see also AI: Beyond the Hyperbole, Plato (Oct. 2024), available at https://static1.squarespace.com/static/6310c0b9bb63a25599f4418c/t/671a1bdff5a0a972a9a43617/1729764321806/AI+Beyond+the+Hyperbole+Final.pdf.
  120. For a brief mention, see A. Blandini, M. Gigliotti, Fintech e innovazione digitale, cit., p. 736 ff., esp. p. 737, where, in the field of SupTech, the use of AI techniques to enhance the analysis of the informational assets that can be drawn from private complaints would allow in-depth and comparative analyses to be carried out in support of, among other things, financial-education activities (see esp. p. 737, n. 78, where it is stated that “Reports, often characterised by extensive documentation that is unstructured in format and language, lend themselves to the application of advanced search techniques (text mining) and forms of automatic interpretation of the meaning of texts to extract recurring concepts and phenomena (machine learning). The use of AI technologies makes it possible to significantly reduce analysis times and improves the early identification of relevant phenomena”).
  121. For not dissimilar considerations, see I. Molenaar, D. Baten, I. Bárd, M. Stevens, Artificial Intelligence and Education. Different Perceptions and Ethical Directions, in N.A. Smuha (ed.), The Cambridge Handbook of the Law, Ethics and Policy of Artificial Intelligence, Cambridge, 2025, p. 261 ff. https://doi.org/10.1017/9781009367783.017
  122. See OECD 2020 Report on financial education, OECD/INFE 2020 International Survey of Adult Financial Literacy; see also F. Trapani, La nuova Direttiva 2023/2225/UE sul credito al consumo: note in tema di educazione finanziaria, merito di credito e servizi di consulenza sul debito, in Nuove leggi civ. comm., 2024, p. 756 ff., who (p. 757) states that “the purpose of financial education is not – and should not be – to eliminate the informational gap between consumer and professional, but only to give the consumer basic literacy in relation to credit products”.
  123. In these terms, M. Cossu, Delle scelte di investimento dei Post-Millennials, e del difficile rapporto tra analfabetismo finanziario e finanza sostenibile, in Riv. soc., 2021, 5-6, p. 1253 ff.
  124. B. Russo, sub Art. 25, in G. Martina, M. Rispoli Farina, V. Santoro (eds.), Legge capitali (5 marzo 2024, n. 21). Commentario, Torino, 2024, p. 297 ff. In particular, Article 25 of the so-called “capitals law” (Law no. 21 of 5 March 2024) amended the rules on the teaching of civic education in schools: it provided that financial, insurance and pension education should also be the subject of teaching within civic education, in a multidisciplinary perspective and including with reference to the use of new digital money-management technologies.
  125. M. Cossu, Delle scelte di investimento dei Post-Millennials, cit., pp. 1262-1263.
  126. N. Linciano, La consulenza finanziaria tra errori di comportamento e conflitti di interesse, in AGE, 2012, p. 135 ff.; see also M. Cossu, L’educazione finanziaria della “generazione Z”. Riflessioni in tempo di pandemia, in C. Costa, A. Mirone, R. Pennisi (eds.), Studi di diritto commerciale per Vincenzo di Cataldo, Vol. II, Tomo I, Torino, 2021, p. 232.
  127. See IOSCO, Artificial Intelligence in Capital Markets, cit., p. 64 ff., where, to address the risks arising from the use of AI systems in financial markets, it is observed that one of the best practicable strategies for retail investors may be to conduct “due diligence prior to deciding to invest in AI focused companies or invest with the assistance of AI technology”: probably, the service to be enhanced is to integrate AI into due diligence.
  128. See IOSCO, Artificial Intelligence in Capital Markets: Use Cases, Risks, and Challenges, Consultation Report (Board/2025/017), March 2025, https://www.iosco.org/library/pubdocs/pdf/IOSCOPD788.pdf, p. 44 ff.; but already FSB, The Financial Stability Implications of Artificial Intelligence, 14 November 2024, https://www.fsb.org/uploads/P14112024.pdf.
  129. Indeed, the vulnerability of one firm in the financial markets could simultaneously affect many firms. An event of this kind could disrupt basic institutions and vital services and potentially cause market disturbance. A failure in one AI system can have cascading effects on others, potentially leading to systemic risks and economic instability. AI systems can interact and influence each other in complex ways, creating feedback loops that can amplify risk.
  130. With effects not dissimilar from those already foreseen for so-called algorithmic trading: on this point, reference may be made to A. Altieri, La regolamentazione del trading algoritmico, tra incontinenza dei dati e abusi di mercato, in Dir. banca merc. fin., 2023, 3, p. 359 ff.
  131. See IOSCO, Artificial Intelligence in Capital Markets, cit., p. 15 ff. Specifically, IOSCO has noted that recent advances in AI are being considered by firms to support internal operations and processes through the automation of certain activities, such as coding, information extraction, classification, clustering, text summarisation, transcription, translation and drafting, and the improvement of customer communication through conversational agents (so-called chatbots). With regard to GenAI in particular, capital-market operators appear to have prioritised internal, low-risk implementations focused on enhancing internal productivity, generating insights or improving risk management, rather than customer-facing applications. This is apparent from the results of the survey conducted by the AMCC (Affiliate Member Consultative Committee). See IOSCO, Artificial Intelligence in Capital Markets, cit., Annex III, p. 71 ff.
  132. See also K. Langenbucher, Artificial Intelligence and Financial Services, in N.A. Smuha (ed.), The Cambridge Handbook of the Law, Ethics and Policy of Artificial Intelligence, Cambridge, 2025, p. 322 ff. https://doi.org/10.1017/9781009367783.020
  133. Indeed, depending on their use and diffusion, AI systems can introduce new cybersecurity threats and exacerbate existing ones, with the challenges for financial firms compounded by limited resources. Cybersecurity risks, in particular those associated with advances in AI, can be classified according to whether AI is used to perpetrate the attack, whether the attack is directed at the AI system, or whether there are AI design and implementation errors. See Workshop Report, Securing Critical Infrastructure in the Age of AI, Center for Security and Emerging Technology (Oct. 2024), available at https://cset.georgetown.edu/publication/securing-criticalinfrastructure-in-the-age-of-ai/.
  134. AI-powered tools used for malicious purposes can lower the entry barriers for bad actors, allowing them to carry out fraud, cyberattacks and other activities more cheaply, in an increasingly automated manner, and more sophisticatedly. As GenAI becomes increasingly available and its outputs more convincing and realistic, bad actors are likely to exploit it to set up schemes to defraud investors or to engage in other improper conduct linked to the financial sector. See United Nations Office on Drugs and Crime, Transnational Organized Crime and the Convergence of Cyber-Enabled Fraud, Underground Banking and Technological Innovation in Southeast Asia: A Shifting Threat Landscape (Oct. 2024), available at https://www.unodc.org/roseap/uploads/documents/Publications/2024/TOC_Convergence_Report_2024.pdf (UNODC Report), p. 9.
  135. IOSCO, Artificial Intelligence in Capital Markets, cit., p. 33 ff.
  136. On the concept of “explainability” of the algorithm see A. Frür, Transparency in the Patent System – Artificial Intelligence and the Disclosure Requirement, in R. Sikorski, Z. Zemla-Pacud (eds.), Patents as an Incentive for Innovation, Alphen aan den Rijn, 2021, p. 235; T.Y. Ebrahim, Artificial intelligence inventions & patent disclosure, in Penn. St. L. Rev., vol. 125, 2020, p. 178 ff.
  137. Among the major limitations one must mention the difficult adaptability of models to market conditions, since a model based on an AI system is probabilistic and not deterministic, with a high risk of “hallucinations”: on this point, see the New York Times article “A.I. Is Getting More Powerful, but Its Hallucinations Are Getting Worse” by Cade Metz and Karen Weise, of 6 May 2025, where it is observed that power is nothing without control; in particular, it is documented how artificial intelligences analyse increasing amounts of data but also make more mistakes — bots rely on complex mathematical systems that nevertheless cannot determine what is true and what is false. Sometimes they make things up: a phenomenon researchers call “hallucinations”, and which, far from decreasing, is rising. In the most recent AI tests it has reached 79%.
  138. On this topic, generally, T. Numerico, Big data e algoritmi. Prospettive critiche, Roma, 2021, passim. LLMs trained on data from the internet, including social media, in particular, may perpetuate or amplify the biases inherent in those data and lead to discriminatory outcomes in the provision of services. Data bias may lead to promoting the products and services offered by one service provider over potentially more affordable or more suitable products offered by a competing firm. Bias may also lead to favouring or disfavouring a particular group of investors and to exacerbating inequalities if this happens frequently. Data, and in particular alternative data, may suffer from selection bias. See IOSCO, Artificial Intelligence in Capital Markets, cit., p. 40.
  139. Big-Tech companies are currently investing massively in AI and related technologies, in intense competition for AI development. For instance, leading firms are reportedly racing to build new and enormous data centres to train and power GenAI and other applications. There is a risk of high concentration in a small number of technology providers in the financial sector, given the resource demands of AI development in terms of development costs, computing power, data access, talent and existing market penetration. The impact of the recent emergence of DeepSeek, an open-source AI model claimed to have achieved results comparable to those of leading closed-source models using a fraction of the training resources, must however also be considered: see DeepSeek-V3 Technical Report (Dec. 27, 2024), available at https://arxiv.org/html/2412.19437v1; DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (Jan. 22, 2025), available at https://arxiv.org/html/2501.12948v1.
  140. Not to mention the data used, which may be proprietary, commercial, open-source or a combination of all the above.
  141. https://www.forbes.com/sites/brianbushard/2023/02/24/workers-chatgpt-use-restricted-at-more-banks-including-goldman-citigroup/; https://www.bloomberg.com/news/articles/2023-02-24/citigroup-goldman-sachs-join-chatgpt-crackdown-fn-reports; https://www.fnlondon.com/articles/citigroup-and-goldman-sachs-join-jpmorgan-with-chatgpt-crackdown-20230224?mod=Searchresults.
  142. IOSCO, Artificial Intelligence in Capital Markets, cit., p. 42 ff.
  143. IOSCO, Artificial Intelligence in Capital Markets, cit., p. 42: “Financial products and service providers could attempt to disclaim liability for investor or market harm resulting from the use of AI systems, or could attempt to shift responsibility to others in the AI system supply chain. Depending on the facts and circumstances, there could be enforcement challenges if AI systems are used in connection with violations of law, in terms of identifying and holding accountable responsible persons, and gathering and presenting evidence due to an AI system’s complexities”.
  144. LayerX Security, Enterprise GenAI Security Report 2025, February 2025, available at https://layerxsecurity.com/blog/layerxs-enterprise-genai-security-report-2025-exposing-hidden-ai-security-blind-spots/.
  145. https://www.jpmorganchase.com/about/technology/blog/open-letter-to-our-suppliers.
  146. Ibidem.
  147. Among many, see the interesting study by M.A. Lasmar Almada, Delegating the Law of Artificial Intelligence. A Procedural Account of Technology-Neutral Regulation, Fiesole, European University Institute, 2024, p. 25 ff., who states that “Technology neutrality is, ultimately, a proposition of indifference”. On the other hand, it has been argued that technological neutrality is an incoherent ideal (see M. Thompson, The Neutralization of Harmony: The Problem of Technological Neutrality, East and West, in (2012) 18 BU J Sci & Tech L, p. 303); while others have argued that neutrality is impossible because regulation always rests on certain assumptions concerning current technologies (see M.D. Birnhack, Reverse Engineering Informational Privacy Law, in (2012) 15 Yale J L & Tech, p. 24). Even among those who consider it possible, technological neutrality is interpreted in different and at times conflicting ways (B.J. Koops, Should ICT Regulation Be Technology-Neutral?, in Id. and others (eds.), Starting Points for ICT Regulation Deconstructing Prevalent Policy One-Liners, TMC Asser Press, 2006);
  148. W.J. Maxwell, M. Bourreau, Technology Neutrality in Internet, Telecoms and Data Protection Regulation, in (2015) 21 Comp Telecomm L Rev, p. 1). https://doi.org/10.2139/ssrn.2529680
  149. Thus M.A. Lasmar Almada, Delegating the Law of Artificial Intelligence, cit., p. 26, who continues: “When we speak about indifference to technology, we mean that the policymaker is not in charge of deciding how to apply that policy to specific technological artefacts and use cases. That is, regulation is technology-neutral if it delegates the interpretation of the technical context to other actors, who might be private rule-makers, courts, administrative enforcers, or even the regulated actors themselves” (pp. 26-27); on that basis, the Author proposes two distinctive qualities of such a “procedural account”: “First, it offers an explicative definition of technology neutrality, which encompasses the core exemplars of both functional and substantive approaches to neutrality. Second, it highlights aspects of technology-neutral regulation that are not emphasized by the alternative accounts” (p. 27).
  150. Its deceptive application is denounced by A. Bertolini, Artificial intelligence does not exist! Defying the technology neutrality narrative in the regulation of civil liability for advanced technologies, in Europa dir. priv., 2022, p. 379 ff.; and now Id., Intelligenza artificiale e responsabilità civile. Problema, sistema, funzioni, Bologna, 2024, p. 160 ff., where mention is made of the “idolum neutralitis”.
  151. Among many, see G. Finocchiaro, La proposta di regolamento sull’intelligenza artificiale: il modello europeo basato sulla gestione del rischio, in Dir. inf., 2022, 2, p. 305 ff.; Ead., Intelligenza artificiale. Quali regole?, Bologna, 2024, p. 49 ff.; Ead., Diritto dell’intelligenza artificiale, Bologna, 2024, p. 1 ff.
  152. See F. Annunziata, La disciplina del mercato dei capitali, XII ed., Torino, 2023, p. 517, on the discipline of crypto-assets; in the same sense, Id., La disciplina europea del mercato delle cripto-attività (MiCAR), in Riv. soc., 2023, 5-6, p. 923 ff. See in particular European Parliament Resolution of 17 May 2017 on FinTech: the influence of technology on the future of the financial sector (2016/2243(INI)), (2018/C 307/06).
  153. Initial mentions are in G. Falcone, Tre idee intorno al c.d. «FinTech», in Riv. dir. banc., 2018, I, p. 37 ff.; see also A. Sciarrone Alibrandi, Innovazione tecnologica, regolazione e supervisione dei mercati, in V. Falce (ed.), Financial Innovation tra disintermediazione e mercato, Torino, 2021, p. 9 ff.; G. Schneider, La proposta di regolamento europeo sull’intelligenza artificiale, cit., p. 1023 ff.
  154. See W. Buczynski, F. Steffek, F. Cuzzolin, M. Jamnik, B.J. Sahakian, Hard Law and Soft Law Regulations of Artificial Intelligence in Investment Management (May 12, 2023), Cambridge Yearbook of European Legal Studies, 24 (2022), pp. 262–293, University of Cambridge Faculty of Law Research Paper 15/2024, available at SSRN: https://ssrn.com/abstract=4786682. https://doi.org/10.1017/cel.2022.10
  155. On this point, see D. Foà, Modelli di regolazione (e supervisione) per l’AI finanziaria: neutralità tecnologica, etica e tutela dell’investitore, in MediaLaws, 2024, special issue, p. 235 ff.; F. Mattassoglio, Algoritmi e regolazione. Circa i limiti del principio di neutralità tecnologica, in Rivista della Regolazione dei mercati, 2018, 2, p. 226 ff.
  156. G. Finocchiaro, Intelligenza artificiale. Quali regole?, cit., p. 49.
  157. Applying precisely those principles affirmed by UNCITRAL in the regulation on electronic signatures, where no constraints were imposed on a particular technological or commercial development, but principles were established that may remain unchanged for a certain period of time, without the constraint of technological change. See G. Finocchiaro, Intelligenza artificiale. Quali regole?, cit., pp. 49-50.
  158. See https://www.ibm.com/think/news/meet-large-database-models-ldms; but see also https://www.forbes.com/sites/ericsiegel/2025/01/13/the-rise-of-large-database-models/.
  159. And on the further applications of LLMs in finance, see Y. Li, S. Wang, H. Ding, H. Chen, Large Language Models in Finance: A Survey, 8 Jul. 2024, at https://arxiv.org/html/2311.10723v2#bib.
  160. See P. Kudva, R. Bordawekar, A. Nitsure, A Scalable Space-efficient In-database Interpretability Framework for Embedding-based Semantic SQL Queries, 1 March 2023, at https://arxiv.org/abs/2302.12178; J.L. Neves, R. Bordawekar, E. Tzortzatos, Demonstrating Semantic SQL Queries over Relational Data using the AI-Powered Database, in International Workshop on Applied AI for Database Systems and Applications (AIDB), co-located with the VLDB conference, https://aidb-workshop.github.io/aidb2019-proceeding/6-neves.pdf.
  161. https://www.ibm.com/think/news/meet-large-database-models-ldms.
  162. See A Competitiveness Compass for the EU, Communication from the Commission to the European Parliament, the European Council, the Council, the European Economic and Social Committee and the Committee of the Regions, Brussels, 29.1.2025, COM(2025) 30 final.
  163. E. Letta, Much more than a market. Speed, Security, Solidarity. Empowering the Single Market to deliver a sustainable future and prosperity for all EU Citizens, April 2024, https://www.consilium.europa.eu/media/ny3j24sm/much-more-than-a-market-report-by-enrico-letta.pdf.
  164. As repeatedly mentioned throughout this paper; and see again IOSCO, Artificial Intelligence in Capital Markets, cit., p. 64 ff.
  165. Š. Hosta, Labor e habitus. Per una comprensione teologico-morale del lavoro umano, Roma, 2024.