Beyond persons and things : the legal status of AI systems in the EU
(2024)
Files
2024Mocanu-BeyondpersonsandthingsThelegalstatusofartificialintelligencesystemsintheEUthesis.pdf
Open Access - Adobe PDF
- 3.8 MB
Details
- Authors
- Supervisors
- Sibony , Anne-Lise
- Abstract
- Artificial Intelligence (AI) systems increasingly straddle the border between persons and things in the pursuit of agency, autonomy, and related abilities. These anthropomorphic features make them disruptive to the standard person-thing dichotomy in the law, which heralds a reconsideration of the theoretical underpinnings of juridical humanism for their legal treatment to be appropriate. This thesis argues that such reconsideration would mean reorganizing the relationship between the foundational legal categories of person and thing, since AI systems fail to neatly fit into either category. It asks the status question in relation to these systems, that is what they are to the law. To answer the status question, the thesis argues for moving beyond persons and things as the only legal status options, while acknowledging that this involves a profound change to the deep structure of law. Mindful of this, the methodology employed is rich and varied, in order to allow for a thorough critical examination of the status quo, as well as a balanced set of answers to the status question. To this end, a coherence approach is utilized, in the framework of which the best account of the relationships among the different reasons that apply to the alternatives is sought. Such an account is formulated in the thesis in the form of the so-called gradient theory of legal status, proposed as capable of accommodating the uncanny entities that AI systems seem to be. Methods such as reflective equilibrium and conceptual engineering are used to devise several new legal status options and ensure that they sit well in the larger theoretical ensemble. The analysis follows successive levels of abstraction, from the more general to the more specific, with the aim of reaching an account of legal status both fit for AI systems, and free of contradiction between the different levels. Accessible and representative examples and case-studies likely to inform and substantiate the theoretical claims are included. Moreover, inter- and cross-disciplinary insight from philosophy, science and technology studies, computer science and other fields is used to better situate and contextualize the subject, as well as to further illuminate areas of doctrinal debate by transplanting notions from other relevant fields. The underlying assumption throughout the thesis is that regardless of the answer to the status question, it must acknowledge the fact that law is artefactual. That entails the fact that the legal status of AI systems is constructed, fashioned out of both the available legal materials as well as new ones, and that it can be tailored to fit their unique characteristics, as well as cater to the socially valuable purposes that a given polity identifies as attainable in bestowing a legal status on an AI system in the first place. Thus, the gradient of legal status options available to be stabilized that is proposed here is meant to account for the varied, numerous and the diverse configurations of AI systems. This thesis argues that legal status is a matter of the utmost importance since it acts as a normative node and determines, once it has been stabilized, which law is applicable and enforceable as to the uses of AI systems and the ensuing consequences of those uses. The argument is then deconstructed by first asking why there is a status question concerning these entities in the first place. In what follows, the background assumptions and theoretical framework surrounding legal status are fleshed out, from the generic to the more granular. Next it asks what is being apprehended by the law exactly and looks at some of the history, salient features and known issues of AI, locating the source of the debate on legal status majoritarily in the accountability gap that AI systems create. Armed with that knowledge, the thesis inventories the possible answers to the status question according to the currently entrenched legal theoretical framework. In so doing, it first shows how entrenched the status quo is, by looking at the backlash to the initial EU impetus to create electronic personality for the most sophisticated AI systems, which has resulted in the current piecemeal approach to regulating AI. The recent AI Act and the accompanying liability framework are then shown to fail in bridging AI-induced liability gaps. The analysis then moves on to survey the fraught history of the debate on legal personhood for AI systems, weighing the arguments for and against the idea that have been formulated in policy, legal doctrine, but also in philosophy. A series of clarifications of the complex legal concept of personhood are considered next, when the most recent theory thereof is presented and an amended formulation of it is advanced. The thesis thus argues for the recent approach to reconceptualizing legal personhood as a bundle, which is the state of the art in legal theory, and only proposes a slight alteration of it as a gradient. It argues that such a metaphor is better suited to explain at least the AI use case. It also fits into a more general gradient theory of legal status, into the development of which it argues that more research should go. Through this adjusted lens, it goes on to analyse an in-between option for the legal status of AI systems, namely partial legal capacity, recently proposed as a solution that is neither here nor there, neither personhood, nor thinghood. It argues that accepting this half-way status would mean accepting the gradient theory of legal personhood. Finally, this thesis makes the case for legal creativity when it comes to the palette of options available as to legal status ascription. In the last section, new legal categories are introduced and analysed, inspired by several doctrinal strands coming from animal studies, sociology or science and technology studies. There it is argued that these theoretical innovations are necessary for a more coherent answer to the legal status question for the case of AI systems. Such an answer, the argument goes, has the potential to help judges decide the hard cases involving AI systems that they will undoubtedly be faced with, not to mention help lawyers argue such cases too. Perhaps most importantly however, if such a theory succeeds in painting a clearer, more complete and accurate picture of what is at stake in the ascription of legal status to AI systems, taking into account all the relevant facets of the issue, it could contribute to better balancing the interests of all those involved.
- Affiliations
Citations
Mocanu, D.-m. (2024). Beyond persons and things : the legal status of AI systems in the EU. https://hdl.handle.net/2078.5/241981