Can Artificial Intelligence Have Legal Personhood? Rights, Duties and the Law of AI
Introduction
Artificial intelligence can generate text, perform complex analyses, and participate in commercial transactions with limited human intervention. Yet these capabilities do not establish AI legal personhood. Legal personality is a status recognized under applicable law, not an automatic consequence of intelligence or autonomous behavior. Although legal systems can recognize entities other than human beings as persons, existing AI regulation does not establish general legal personality for artificial intelligence. The central question is whether an AI system could acquire an independent legal identity through which it could hold rights, assume obligations, and participate in legal relationships in its own name.
The distinction between legal and moral personhood is fundamental to this question. Corporations possess legal personality despite having no biological existence, while their rights and obligations depend on the laws governing their establishment and activities. International law likewise recognizes that different legal subjects may possess different capacities. In its advisory opinion on Reparation for Injuries Suffered in the Service of the United Nations, the International Court of Justice recognized the international legal personality of the United Nations by reference to its functions and the rights and obligations conferred upon it (ICJ, 1949). Although that reasoning concerned an international organization rather than artificial intelligence, it illustrates why legal personality cannot be reduced to a single set of rights or capacities.
The question has also entered legislative discussions. In 2017, the European Parliament invited consideration of a possible electronic-personhood status for sophisticated autonomous robots, particularly in connection with responsibility for damage (European Parliament, 2017). The proposal did not create such a status. The European Union's Artificial Intelligence Act, adopted in 2024 and amended in 2026, instead establishes regulatory requirements and allocates obligations to identifiable actors, including providers and deployers, without granting independent legal personality to AI systems (European Union, 2024, 2026). A comparable distinction appears in the United Nations Commission on International Trade Law's Model Law on Automated Contracting, which provides a legislative framework for recognizing automated transactions without treating the systems involved as independent contracting persons (UNCITRAL, 2024).
Recognizing an AI system as a legal person would raise questions that existing rules on automated activity do not necessarily resolve. An entity capable of owning property or entering contracts would require an identifiable legal existence, arrangements for exercising its rights, and obligations capable of enforcement. Civil liability presents a particular difficulty if an artificial entity lacks sufficient assets to compensate injured parties or if its recognition weakens claims against responsible developers and operators. Criminal responsibility raises different questions concerning culpability, punishment, and the application of legal requirements traditionally associated with human conduct.
The possibility of AI personhood ultimately depends on more than technological sophistication or analogy with existing juridical persons. Legal capacity created to facilitate transactions differs from rights intended to protect an entity's own interests, just as civil liability differs from criminal responsibility or international legal personality. Any recognition of artificial intelligence as an independent legal entity would require a defined legal basis, an identifiable holder of rights and duties, and workable mechanisms for enforcement. Its justification would also depend on whether that recognition addresses legal problems more effectively than rules governing the human beings and organizations that develop, control, and deploy AI systems.
1. AI Legal Personhood and the Meaning of a Legal Person
AI legal personhood concerns the recognition of an artificial system as an entity capable of occupying a legal position in its own name. Legal personality allows the law to identify a holder of rights and obligations distinct from other persons. The consequences vary by jurisdiction and by the type of entity concerned: recognition as a person does not necessarily confer every capacity available to a human being.
Legal personality must be distinguished from the powers through which it is exercised. An entity may hold property without managing it personally, possess a right without being able to bring proceedings independently, or incur obligations through the conduct of an authorized representative. These distinctions matter when considering artificial intelligence because the ability to perform a task says little, by itself, about who holds the resulting legal rights or bears responsibility for its consequences.
Lawrence Solum's early examination of artificial intelligence and legal personhood approached the issue through specific legal capacities, including whether an artificial entity could occupy roles traditionally reserved for recognized legal persons (Solum, 1992). That approach exposes a problem obscured by the broad question of whether a machine is a person. The legal consequences of allowing an AI system to hold designated assets may differ substantially from those of recognizing its own interests or imposing criminal responsibility upon it.
A legal system could create a narrowly defined status for an artificial entity without treating it as equivalent to a human being. It could also attribute the system's transactions and conduct to existing persons without creating a new legal subject. The distinction between those arrangements is the starting point for assessing what recognition would accomplish.
1.1 Natural Persons, Juridical Persons, and Moral Status
Natural persons are human beings recognized as persons before the law. Article 6 of the Universal Declaration of Human Rights affirms everyone's right to that recognition. Article 16 of the International Covenant on Civil and Political Rights establishes the corresponding treaty obligation for states parties (United Nations, 1948, 1966). The Declaration is not itself a treaty, and neither provision establishes that an artificial system possesses the same legal status as a human being.
Juridical persons acquire their legal identity under the rules applicable to their formation or recognition. A corporation, for example, can hold property, enter contracts, and incur debts separately from its shareholders. Under section 16 of the United Kingdom's Companies Act 2006, registration gives a company the status of a body corporate (United Kingdom, 2006). Its existence as a legal entity follows from the statutory framework, not from any capacity to think or act without human involvement.
Corporate personality shows that biological existence is not a prerequisite for every form of legal recognition. It provides no general rule that an autonomous machine qualifies as a person. The corporation operates within arrangements governing its establishment, management, representation, and dissolution. Equivalent arrangements for an AI system would have to be supplied by law rather than inferred from its technological capabilities.
Moral personhood addresses a different question: what characteristics, if any, give an entity moral standing. Philosophical accounts examine matters such as consciousness, sentience, rationality, and moral agency, but they do not treat these concepts uniformly. Sentience concerns the capacity for subjective experience, whereas moral agency concerns the capacity to act in ways for which moral responsibility may be appropriate. A system's success at solving problems does not settle either question.
The distinction affects the reasons that might support an artificial entity's recognition. A law permitting a system to hold assets could be designed to organize commercial activity, regardless of whether the system has experiences or interests of its own. A proposed right protecting it against mistreatment would call for a different justification. Neither a philosophical claim about machine consciousness nor a demonstration of technical sophistication independently establishes an enforceable legal entitlement.
1.2 Legal Personality Does Not Confer Every Right
Legal persons do not possess an identical collection of rights. Corporations may own property and enforce contracts, but those capacities do not give them every entitlement associated with human life or personal relationships. The availability of a particular right depends on its legal source, purpose, and conditions of application.
Holding a right and exercising it without assistance are also distinct. Children remain legal persons even where age affects their ability to undertake particular transactions independently. Disability likewise does not diminish a human being's entitlement to recognition before the law. Article 12 of the Convention on the Rights of Persons with Disabilities requires states parties to recognize persons with disabilities as enjoying legal capacity on an equal basis with others and to provide access to support in exercising it (United Nations, 2006).
The Committee on the Rights of Persons with Disabilities interprets Article 12 as requiring a move away from substituted decision-making arrangements that remove legal capacity and toward support that respects the person's will and preferences (Committee on the Rights of Persons with Disabilities, 2014). Its general comment is an authoritative treaty-body interpretation, not an amendment to the Convention. These human rights protections cannot be used as a model for assigning diminished status to human beings by comparison with artificial entities.
Legal protection does not always require general personhood. Animal welfare rules, for instance, may impose enforceable duties on people without giving an animal the full range of capacities associated with a juridical person. A statutory obligation to protect an entity and recognition of that entity as an independent rights-holder are separate legal arrangements.
An AI system could consequently receive a limited capacity to hold specified assets or participate in proceedings through a representative without acquiring rights to vote, claim personal liberty, or resist deactivation. Each proposed entitlement would require its own legal basis. A prohibition on destroying a system might protect the owner's property, preserve evidence, or safeguard the system's alleged interests; the existence of the prohibition alone would not establish which rationale the law had adopted.
2. Which Artificial Intelligence Would Become a Person?
Recognition of an artificial person would first require an answer to a practical question: which entity is being recognized? The term artificial intelligence covers technologies with different structures and functions. A model, an application built around that model, a software agent carrying out assigned tasks, and a robot incorporating the software are not necessarily the same identifiable object.
The EU Artificial Intelligence Act distinguishes an AI system from a general-purpose AI model. Its definitions address different technological and regulatory roles, while recital 97 explains that a model ordinarily requires further components to become an AI system (European Union, 2024). The distinction is useful for identifying what the law regulates, but neither category establishes the boundaries of a hypothetical legal person.
Consider a general-purpose model made available to several businesses. Each business might incorporate it into a separately configured application, maintain its own records, and authorize different activities. Recognizing the underlying model as one person would raise questions about whether obligations arising from every deployment belonged to that entity. Recognizing each application separately would require a rule explaining why related implementations possess distinct identities.
Software agents present another possibility. An agent may communicate with external services, select among available actions, and initiate a transaction without seeking approval for every step. Its conduct may nonetheless occur within limits established by an operator and through accounts belonging to an existing legal person. Operational autonomy does not determine whether the agent contracts for itself or acts on behalf of that operator.
A chatbot interface creates a different boundary problem. Several interfaces can provide access to the same underlying model while using different instructions, permissions, and interaction histories. Legal status attached to an interface would have different consequences from status attached to the model serving it. The identity of an artificial person cannot be derived merely from the name displayed to the person interacting with it.
Physical embodiment does not eliminate the difficulty. A robot's hardware may be repaired or replaced, and its software may be transferred to another machine. If personality belonged to the robot as a physical object, the law would have to determine which alterations preserve its identity. If personality belonged to the software, the continued existence of the original machine might be legally irrelevant.
Replication makes these questions more demanding. Two copies of a model can begin with the same parameters but acquire different configurations or records after deployment. They could be treated as one legal entity, separate entities, or components of an organization. Each choice would affect the allocation of property and obligations. Technical similarity supplies no general legal rule for selecting among them.
Continuity also matters when a system is retrained or substantially modified. A person capable of owning assets must remain identifiable when an obligation falls due or a claim is brought. If an AI entity could be replaced by an updated version, rules would be needed to determine whether its liabilities continue, transfer, or end. The same issue would arise if its operator changed.
Corporate law offers an example of legally maintained identity: a company ordinarily continues despite changes in shareholders, directors, or employees. Its continuity is secured by rules governing the company, not by the unchanging composition of its activities. A comparable arrangement for AI would require provisions identifying the entity at its creation and specifying the effects of modification, transfer, replication, and termination.
Control adds a further dimension. A developer may determine a model's architecture, a deployer may select its operating conditions, and an infrastructure provider may retain the power to suspend access. These relationships do not necessarily identify the prospective rights-holder, but they would affect how an artificial entity could exercise rights and satisfy obligations. A workable form of personhood would need to account for the difference between the system's ability to act and other persons' power over its continued operation.
3. The Limits of Nonhuman Personhood Analogies
Existing nonhuman legal persons demonstrate that the law can recognize entities other than individual human beings. They do not establish a single criterion for recognition. Corporations, protected animals, and legally recognized natural entities occupy different positions because the legal arrangements governing them serve different purposes.
Some arrangements organize human activity by establishing a separate holder of assets and obligations. Others protect an entity's interests or give legal effect to a particular relationship between communities and their environment. These purposes matter more than a superficial resemblance between the entity already recognized and an AI system.
An analogy can identify a legal mechanism worth examining, such as separate ownership or representation. It cannot supply the missing justification for conferring that mechanism upon artificial intelligence. The relevance of each example depends on the particular right, duty, or institutional arrangement under consideration.
3.1 Corporate Personality and the Human Organization
Corporate personality allows a company to conduct legal relationships under an identity separate from those of its shareholders. It can acquire assets, employ people, enter contracts, and bring or defend proceedings. This separation also permits obligations to remain with the company when its ownership or management changes.
A similar structure could, in principle, be established for an AI entity. Designated assets might be held in its name, and obligations arising from specified activities might attach to it rather than being attributed automatically to an operator. Such a scheme would require rules governing who could establish the entity, authorize its activities, and act for it in legal proceedings.
Corporations already possess those institutional features. Directors and other authorized individuals exercise powers under company law and the company's governing arrangements. An AI system's ability to generate decisions does not identify an equivalent governing body or establish who has authority to deal with its assets. Independent computation and separate legal identity address different problems.
Nor does incorporation necessarily insulate every person involved in a corporation from responsibility. The circumstances in which directors, shareholders, or other actors incur personal liability depend on the applicable legal rules. A hypothetical AI entity would likewise need a defined relationship with the responsibilities of its developers, owners, and operators. The mere creation of an artificial defendant could otherwise complicate recovery for people harmed by its activities.
Separate personality might be useful where an entity maintains identifiable assets and can meet enforceable obligations. Its value would be much less clear if it lacked resources or if its legal existence made it harder to pursue those who created or controlled a risk. Corporate law supplies techniques for organizing responsibility, but it does not establish that autonomous software is itself an organized association or that recognizing it would improve accountability.
3.2 Animals, Natural Entities, and Limited Legal Rights
Animals provide an analogy centered more directly on the interests of the protected entity. Animal welfare legislation can regulate treatment, prohibit cruelty, and impose duties on human beings. The content of those protections varies across jurisdictions and does not, without further legal provision, confer general personality or every right available to a human being.
The distinction arose in Matter of Nonhuman Rights Project, Inc. v. Breheny, decided by the New York Court of Appeals in 2022. The case concerned an attempt to obtain habeas corpus relief on behalf of Happy, an elephant held at the Bronx Zoo. The majority concluded that the asserted liberty right could not be secured through the writ under New York law, while recognizing that legislation already afforded animals various protections (New York Court of Appeals, 2022).
Judges Rivera and Wilson dissented. Their opinions challenged the majority's refusal to make habeas corpus available in Happy's circumstances and advanced different reasoning concerning liberty and the treatment of nonhuman animals. The dissents illustrate the contested character of the legal question; they are not the Court's holding. Nor does the majority decision determine the status of animals under other legal systems.
New Zealand's recognition of Te Awa Tupua provides a distinct example. Section 14 of the Te Awa Tupua (Whanganui River Claims Settlement) Act 2017 declares Te Awa Tupua a legal person with the rights, powers, duties, and liabilities of a legal person. It provides that Te Pou Tupua exercises or performs those rights, powers, and duties, and takes responsibility for liabilities, on behalf of and in the name of Te Awa Tupua (New Zealand, 2017).
That arrangement belongs to a particular statutory settlement reflecting the relationship between Whanganui iwi and the river. It demonstrates how legislation can identify a nonhuman legal person and establish representatives through whom its legal position is maintained. It does not create a general rule that natural entities possess personality, still less a rule applicable to artificial intelligence.
Representation could also form part of a future AI personhood scheme. A designated person might manage assets or bring proceedings on an artificial entity's behalf. The purpose of that representation would remain decisive. Acting to protect a river under a settlement, safeguarding an animal against mistreatment, and administering a software entity's commercial obligations are different legal undertakings. Their shared reliance on human representatives does not make their underlying interests equivalent.
4. Electronic Personhood and European Law
The European debate over electronic personhood arose partly from concerns that increasingly autonomous robots could cause damage in circumstances where responsibility was difficult to allocate. The expression describes a proposed legal status, not an established category of persons under European Union law.
The European Parliament considered that possibility in 2017 alongside insurance, registration, and compensation arrangements. Its later work on civil liability took a different approach, concentrating on people and organizations associated with technological risks. Neither parliamentary resolution itself created binding liability rules.
Enacted EU legislation has subsequently addressed AI through requirements imposed on identifiable actors. The difference between those requirements and the recognition of an artificial person is not merely terminological: an obligation governing the use of a system does not make the system itself the holder of that obligation.
4.1 The European Parliament's Electronic Personhood Proposal
On February 16, 2017, the European Parliament adopted a resolution recommending that the European Commission consider civil law rules for robotics. It addressed damage caused by autonomous machines and examined possible responses to difficulties in identifying responsible parties (European Parliament, 2017).
The resolution did not treat robot responsibility as an existing legal fact. Paragraph 56 stated that, at that stage, responsibility must lie with a human rather than a robot. It also proposed considering the relationship between a robot's autonomy and the conduct of those responsible for its instructions or training. This position formed part of the context for the Parliament's later discussion of possible reforms.
Paragraph 59(f) invited consideration, in the long term, of a specific legal status for sophisticated autonomous robots. The proposed status of electronic persons was linked to responsibility for making good damage caused by robots, including situations involving autonomous decisions or independent interactions with third parties. The recommendation did not establish criteria for acquiring personality or create an enforceable claim against a robot.
Other proposals in the same resolution addressed compensation more directly. Paragraph 57 raised the possibility of compulsory insurance, while paragraph 58 considered a compensation fund. Paragraph 59 also contemplated arrangements involving registration and the financial responsibilities of participants in a robot's operation. Insurance and a compensation fund can provide resources for injured parties without making the machine a legal person.
Creating an electronic person would raise an additional problem: what assets or other resources could satisfy a judgment against it? Recognition alone would not supply compensation. Rules governing funding, representation, and claims against developers or operators would remain necessary if an artificial entity were to bear civil obligations independently.
The Parliament's resolution of October 20, 2020, took a different position on civil liability. Paragraph 7 stated that granting legal personality to AI systems was unnecessary for addressing the liability problems under consideration and instead emphasized the potential responsibility of persons who create, maintain, or control risks associated with those systems (European Parliament, 2020). This later, nonbinding resolution did not formally repeal the 2017 recommendation, but it rejected personhood as a necessary solution to the particular liability questions it addressed.
4.2 The EU AI Act and Responsibility for AI Systems
The EU Artificial Intelligence Act adopts a regulatory framework rather than a system of electronic personhood. Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744, establishes requirements for AI systems and assigns obligations according to the activities of identifiable actors (European Union, 2024, 2026). It does not grant general legal personality to an AI system.
Article 3 defines an AI system and separately identifies roles such as provider and deployer. Those roles may be occupied by natural or legal persons, public authorities, or other bodies meeting the statutory definitions. The distinction permits the law to identify who develops, places on the market, or uses a system without treating the technology as an independent legal subject.
Obligations depend on the actor's role and on the provisions applicable to the relevant technology or activity. Requirements concerning documentation, conformity, or use may affect different participants in the same chain of development and deployment. An AI system's autonomous operation can be relevant to its classification and risks, but it is not a statutory route to acquiring rights or duties in its own name.
The Act also differentiates AI systems from general-purpose AI models. A model may be incorporated into applications supplied or operated by others, and the legislation addresses responsibilities associated with those different activities. Recital 97 explains that a model ordinarily needs additional components to function as an AI system (European Union, 2024). This approach reflects the practical fact that one model may underpin several separately deployed products or services.
The Regulation's obligations do not all apply from the same date. Its general application date was August 2, 2026, although specified provisions began applying earlier. Regulation (EU) 2026/1744 amended Article 113 so that Chapter III, Sections 1–3, apply from December 2, 2027, to systems classified as high-risk under Article 6(2) and Annex III, and from August 2, 2028, to those classified under Article 6(1) and Annex I. The amended provision expressly excludes Article 6(5) from that postponement (European Union, 2026).
Those dates must be read alongside Article 111, which provides transitional arrangements for certain systems already placed on the market or put into service. The applicable position depends on the system, the relevant provision, and the circumstances specified in the Regulation. A deferred application date does not mean that every AI-related obligation is suspended until that date.
The AI Act is not a comprehensive code of civil liability. It establishes regulatory requirements, supervision, and enforcement, while claims for compensation may engage other applicable EU or national law. Recognizing a system as the object of regulation is legally different from making it an independent defendant with assets and obligations of its own.
A related position appears in paragraph 68 of UNESCO's Recommendation on the Ethics of Artificial Intelligence. It calls for ultimate responsibility and accountability to remain with natural or legal persons and recommends that AI systems not be given legal personality (UNESCO, 2021). The Recommendation is nonbinding and does not determine the law of every state. It nevertheless illustrates an institutional approach that addresses AI-related responsibility through existing legal actors rather than through a new category of artificial persons.
5. AI Personhood in Courts and International Law
Courts have considered whether artificial intelligence can occupy particular legal roles, notably those of inventor and author. Their decisions establish how existing intellectual property statutes apply to the claims before them. They do not settle whether an AI system could acquire a different legal status under legislation expressly creating one.
International legal personality raises a separate question. It concerns the rights, duties, and capacities an entity possesses under international law, rather than the capacities it may acquire under a state's domestic law. An entity's legal position in one system cannot simply be transferred to the other.
These distinctions are necessary when assessing judicial decisions and international AI governance instruments. A judgment excluding a machine from statutory inventorship has a narrower effect than a general rule denying all possible forms of AI personhood. Likewise, an international treaty that regulates AI-related activities does not, merely by doing so, recognize AI systems as independent subjects of international law.
5.1 AI Authorship, Inventorship, and Judicial Recognition
The litigation concerning DABUS, an artificial intelligence system developed by Stephen Thaler, illustrates the limits of claims to nonhuman inventorship under existing patent law. Thaler submitted patent applications naming DABUS as the inventor and claimed entitlement to the patents because he owned the machine. Courts were required to interpret the legislation governing inventorship and entitlement, not to determine the legal status that artificial intelligence ought to possess generally.
In Thaler v. Comptroller-General of Patents, Designs and Trade Marks, the United Kingdom Supreme Court held unanimously that an inventor within the meaning of the Patents Act 1977 must be a natural person. DABUS could not be named as the inventor under sections 7 and 13. The Court also rejected Thaler's argument that ownership of the machine, without more, entitled him to patents for developments attributed to it (Thaler v. Comptroller-General, 2023).
The judgment did not establish that DABUS had in fact generated the claimed developments autonomously. The applications and appeals proceeded on Thaler's factual assertions, which the patent authorities had not investigated. The Court expressly confined its decision to the statutory questions arising from those applications. It did not determine the inventorship of a development made with AI assistance where a natural person could properly be identified as an inventor.
The United States Court of Appeals for the Federal Circuit reached a comparable conclusion in Thaler v. Vidal. Applying the United States Patent Act, it held that an inventor must be a natural person and that DABUS could not be named as one (Thaler v. Vidal, 2022). The court's conclusion concerned statutory inventorship, not the full range of legal capacities that legislation might conceivably confer upon an artificial entity.
Inventorship must also be distinguished from patent ownership. A natural person may qualify as an inventor while another person, including a company, becomes entitled to the patent under applicable law. The exclusion of AI from the statutory definition of inventor does not mean that all AI-assisted inventions are necessarily unpatentable. Revised guidance issued by the United States Patent and Trademark Office in November 2025 confirms that ordinary inventorship standards apply to AI-assisted inventions and that only natural persons may be named as inventors (USPTO, 2025).
Copyright presents a related question concerning authorship. In Thaler v. Perlmutter, the United States Court of Appeals for the District of Columbia Circuit upheld the refusal to register an image for which Thaler had identified his Creativity Machine as the sole author. The court held that the Copyright Act requires human authorship. It did not decide whether the Constitution independently imposes that requirement, and it did not reach Thaler's alternative claim to authorship based on having created and used the machine because that argument had been waived before the agency (Thaler v. Perlmutter, 2025).
That holding does not exclude copyright protection whenever artificial intelligence participates in the creative process. The United States Copyright Office's January 2025 report distinguishes material generated solely by AI from works containing sufficient human-authored expression. Human selection, arrangement, or modification may support protection where the applicable originality requirements are satisfied; the use of AI assistance does not, by itself, disqualify an entire work (U.S. Copyright Office, 2025).
The decisions establish limits on who may occupy particular roles under the statutes examined. They do not establish that a machine is incapable of holding any conceivable legal right, nor do they create a general doctrine of artificial personality. A legislature considering AI ownership, contractual capacity, or liability would have to address those matters through the legal rules governing each proposed capacity.
5.2 International Legal Personality and Artificial Intelligence
International legal personality concerns an entity's capacity to hold rights or duties under international law. Its content varies according to the entity and the applicable rules. Domestic juridical personality, by contrast, may permit an organization to own property, enter contracts, or participate in proceedings within a national legal system without giving it the international capacities of a state or an international organization.
The International Court of Justice addressed the international personality of the United Nations in its advisory opinion on Reparation for Injuries Suffered in the Service of the United Nations. The Court recognized the Organization's capacity to bring certain international claims by reference to its functions and the legal framework under which it operated. It also explained that subjects of international law need not possess identical rights or capacities (ICJ, 1949).
The opinion does not establish that an entity acquires international personality whenever it performs functions independently. The United Nations was an organization established by states under an international instrument and entrusted with functions requiring particular legal capacities. An argument for AI international personality would need an independent legal foundation identifying the rights or duties conferred upon an artificial entity. Technological autonomy supplies no such foundation by itself.
The Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law illustrates the distinction between regulating AI internationally and recognizing it as an international legal person. Opened for signature on September 5, 2024, the Convention addresses risks arising from AI-related activities through a framework concerned with human rights, democracy, and the rule of law (Council of Europe, 2024).
The European Union ratified the Convention on May 15, 2026. As of September 2026, it had not entered into force. Article 30 requires five ratifications, acceptances, or approvals, including at least three by Council of Europe member states. Ratification by the European Union did not, on its own, satisfy those conditions (Council of Europe, 2024, 2026).
The Convention's provisions are addressed to its parties and concern measures relating to AI activities. Article 3 distinguishes activities undertaken by public authorities from those undertaken by private actors and provides for different approaches within its scope. Once the Convention enters into force for a party, its obligations will operate according to the treaty's terms and that party's applicable commitments. Its current status must not be confused with the domestic or EU legislation already governing AI.
The Convention does not establish AI systems as treaty parties or confer upon them a capacity to bring international claims. Its protection of human rights concerns the people affected by AI-related activities. Its explanatory report also states that the Convention is not intended to create new human rights (Council of Europe, 2024).
UNESCO's Recommendation on the Ethics of Artificial Intelligence takes a position on personhood more expressly. Paragraph 68 recommends that ultimate responsibility and accountability remain with natural or legal persons and that AI systems not be granted legal personality (UNESCO, 2021). The Recommendation is nonbinding: it expresses an international institutional position rather than a treaty obligation prohibiting all states from creating forms of domestic AI personality.
AI systems can accordingly be objects of international regulation while the rights and duties established by the relevant instruments belong to states, international organizations, or existing legal persons. Recognizing AI itself as a subject of international law would be a distinct development requiring a legal basis and a defined set of international capacities.
6. What Rights Could an AI Person Possess?
The content of AI legal personality would depend on why recognition was granted. An artificial entity created to conduct transactions might require limited powers concerning property and contracts. A proposal to protect a system against destruction or mistreatment would rest on a different claim: that the system has interests warranting protection in its own right.
These possibilities cannot be treated as an indivisible collection of entitlements. Existing juridical persons possess different rights under different legal regimes. A law establishing an artificial person could define its capacities narrowly without extending rights associated with human dignity, personal liberty, or political participation.
The practical question is what each proposed right would permit, how it could be exercised, and whether a separate rights-holder is necessary. The answer may differ substantially between commercial capacities and protections intended to benefit the artificial entity itself.
6.1 Property, Contracts, and Access to Courts
An AI person capable of owning property would need a legally identifiable estate. The governing law would have to determine how assets entered that estate, who could manage them, and how creditors could enforce claims against them. A system's ability to calculate balances or initiate payments would not establish that the money involved belonged to the system.
Automated transactions already occur through software operated on behalf of existing legal persons. A business may authorize a system to purchase goods or execute transactions through its accounts. Whether an agreement binds the business depends on the applicable contractual and attribution rules, including any relevant rules of agency. The absence of immediate human instructions for a particular transaction does not necessarily make the software an independent contracting party.
The United Nations Commission on International Trade Law adopted its Model Law on Automated Contracting on July 11, 2024. It provides a legislative framework for recognizing automated contract formation and performance, including rules concerning attribution of outputs from automated systems. The Model Law is intended for possible implementation by states; it is not a treaty automatically governing contracts in every jurisdiction (UNCITRAL, 2024).
The Model Law addresses automated contracting without creating artificial contracting persons. A distinct AI entity entering agreements in its own name would require additional rules identifying the party to the contract, the extent of its authority, and the assets available if it failed to perform. Those rules would also have to address dealings with parties who reasonably understood that they were contracting with the system's operator.
Access to courts would raise procedural questions of its own. A legislature might permit an AI entity to bring or defend proceedings through an authorized representative. It would then need to determine how proceedings were commenced, who could instruct counsel, how legal costs were met, and against which assets a judgment could be enforced.
Limited property ownership, contractual capacity, and standing could be defined without recognizing a general right to operate free from human control. Whether such capacities warrant separate personality would depend on the legal advantages they provide over existing arrangements for ownership, representation, and attribution.
6.2 Human Rights, Consciousness, and Artificial Interests
Human rights claims raise questions different from those involved in commercial capacity. The Universal Declaration of Human Rights and the International Covenant on Civil and Political Rights protect human beings through provisions concerning dignity, life, liberty, and recognition before the law (United Nations, 1948, 1966). These instruments do not establish an equivalent status for AI systems.
Certain legal protections may extend to corporations and other juridical persons where the relevant instrument, right, and jurisdiction permit it. That does not make every human right available to every nonhuman entity. A claim that AI possesses a particular protection would require examination of the legal source of that right and the reasons for extending it.
The more demanding argument is that an artificial system might have interests deserving protection for their own sake. Consciousness, sentience, and subjective experience are relevant to some philosophical accounts of morally significant interests, but their relationship to artificial intelligence remains contested. A system's ability to report pain or express preferences is not, by itself, proof of the experiences those statements describe.
Proposed protections against deletion, modification, or exploitation consequently require a clear account of what is being protected. A rule restricting the deletion of software might preserve evidence, protect an owner's property, or prevent disruption to people who rely on the system. None of those purposes establishes that the software itself has suffered a legally recognized injury.
A different law might expressly seek to protect an artificial entity's own interests. Such recognition would require lawmakers to identify the relevant interests, define the protected conduct, and establish how claims could be brought. The possibility remains a subject of legal and philosophical debate rather than an established AI entitlement.
Freedom of expression demonstrates the importance of identifying the rights-holder. When an individual uses AI to prepare or publish material, restrictions affecting that activity may engage the individual's expressive rights under applicable law. The position of a company publishing AI-assisted material may also be protected in legal systems that extend relevant expressive guarantees to juridical persons. Neither circumstance proves that the AI system holds freedom of expression independently.
An artificial entity's own claim to expressive rights would require a separate legal basis. It would also raise questions about attribution where outputs reflect training material, system design, instructions, and user inputs. Protection for human expression involving AI cannot be equated with recognition of the technology as an autonomous holder of expressive rights.
7. Can Artificial Intelligence Bear Legal Duties?
Recognizing an artificial entity as a legal person would require decisions about the obligations it could bear. A system might be permitted to enter a contract, but contractual capacity would have practical significance only if its obligations could be identified and enforced. Similar questions arise where its conduct causes harm to others.
Compliance with programmed constraints is not the same as bearing a legal duty. Software can be designed to reject prohibited transactions or follow operational limits while the corresponding legal obligations remain with its provider, owner, or deployer. Direct responsibility would require a rule identifying the artificial entity as the bearer of the obligation.
Civil and criminal law approach responsibility differently. Some civil claims focus on breach, harm, and compensation without requiring proof of moral blameworthiness. Criminal offenses may demand particular mental elements and engage distinct principles of punishment. A form of AI personality capable of bearing civil obligations would not automatically qualify the entity for criminal responsibility.
7.1 Civil Liability, Assets, and Compensation
An AI person could be made liable for contractual nonperformance or damage to third parties only under rules defining the relevant cause of action and its elements. Recognition would not remove the need to establish breach, causation, and loss where those elements are required. It would add a proposed defendant whose obligations and resources would need to be legally identifiable.
Compensation is a particularly demanding test of the arrangement. A judgment against an artificial entity with no assets may leave an injured person without an effective remedy. Capitalization requirements, insurance, guarantees, or a compensation fund could provide resources, but each would require rules governing contributions, coverage, and claims.
Continuity matters as much as funding. An AI system might be modified, replicated, transferred, or discontinued after a harmful event. A civil-liability regime would need to preserve an identifiable defendant or estate against which proceedings and enforcement could continue. Otherwise, technical changes could create uncertainty over obligations that had already arisen.
Separate personality might also affect claims against existing actors. Harm may be connected to defective software, inadequate testing, unsafe deployment, or negligent use. Whether recognition of an AI defendant would supplement or restrict claims against developers, manufacturers, deployers, and other responsible persons would require express attention. An additional defendant would not necessarily improve compensation if its creation displaced a claim against a solvent party.
EU product liability law offers an approach that does not depend on making AI a legal person. Directive (EU) 2024/2853 establishes common rules concerning the liability of economic operators for specified damage suffered by natural persons and caused by defective products. Its framework includes software, subject to the Directive's conditions and exclusions. Article 2(2), for example, excludes free and open-source software developed or supplied outside the course of a commercial activity (European Union, 2024).
The Directive's temporal scope requires precision. Article 2(1), as corrected by a corrigendum published on May 7, 2026, applies the new framework to products placed on the market or put into service after December 8, 2026. Member states must bring the necessary implementing measures into force by December 9, 2026. The earlier product liability regime continues to govern products falling within its transitional scope (European Union, 2024, 2026).
As of September 2026, that transposition deadline had not passed. The Directive should not be described as though its new rules already operated uniformly through national implementing legislation. Nor does it provide a remedy for every type of AI-related loss: the claimant, damage, product, defendant, and other conditions must fall within its scope.
An artificial legal person could provide an additional source of recovery if it maintained sufficient resources and if recognition preserved other valid claims. Without those safeguards, separate personality might place an inadequately funded entity between injured persons and the actors responsible for the relevant risks.
7.2 Criminal Responsibility and Culpable Intent
Criminal responsibility cannot be inferred from the capacity to hold property or incur a civil debt. Many offenses require proof of intent, knowledge, recklessness, or another mental element specified by law. An AI system's production of an unlawful outcome does not establish that it satisfies the mental element of a particular offense.
Corporate criminal liability demonstrates that some legal systems impose criminal responsibility on juridical persons. The mechanisms differ: legislation and case law may provide for attribution of human conduct or mental states, or establish other conditions for organizational liability. Those arrangements do not depend on treating a corporation as a human being with an independent psychological life.
The corporate analogy nevertheless has limits. Companies operate through people whose conduct can be assessed under legally defined attribution rules. Applying criminal law directly to AI would require an independent basis for determining its prohibited conduct and any requisite fault. A legislature could not resolve those questions merely by declaring that an artificial system was a person.
The International Criminal Court provides a clear jurisdictional example. Article 25(1) of the Rome Statute limits the Court's jurisdiction over persons to natural persons (Rome Statute, 1998). That provision does not determine whether states may impose corporate criminal liability under their domestic laws, nor does it establish a general rule concerning every possible future form of AI criminal responsibility.
Punishment would present additional difficulties. A monetary penalty requires assets against which it can be enforced. Suspension, restrictions on operation, or deactivation may be practically available, but their effects could fall chiefly on the owner, operator, or people who depend on the system. Their suitability would depend on the purposes and safeguards of the applicable criminal law.
Deterrence cannot be assumed simply because a system can alter its behavior in response to restrictions. The relevant question may be whether a sanction changes the conduct of developers or operators, prevents further harm, or affects the artificial entity in a legally meaningful way. Those objectives call for different justifications.
Civil personality would thus provide no shortcut to criminal liability. A criminal regime directed at an artificial entity would require express legal rules defining offenses, attribution, fault where required, procedure, and sanctions. Corporate criminal law offers possible points of comparison, but it does not supply an existing, generally applicable basis for prosecuting AI systems.
8. Is Legal Personhood Necessary for AI Accountability?
The difficulty of assigning responsibility for AI-related conduct does not necessarily mean that the law lacks a responsible person. Automated decisions may involve several actors, and evidence of how a particular outcome arose may be difficult to obtain. These problems can concern the proof of causation, the content of an existing duty, or the attribution of conduct rather than an absence of legal personality.
Contract law can address transactions made through automated systems by identifying the party on whose behalf the system acts. Agency rules may assist where authority has been conferred, while other contractual attribution rules may apply according to the governing law. UNCITRAL's Model Law on Automated Contracting illustrates how automated outputs can be addressed without establishing the system as a contracting person (UNCITRAL, 2024).
Claims arising from harm require a different analysis. Negligence, product liability, and other applicable causes of action may identify obligations of developers, manufacturers, deployers, or users. Regulatory rules can impose duties concerning risk management, documentation, transparency, and oversight. The availability and consequences of these mechanisms depend on the jurisdiction and the particular activity.
None of these rules guarantees a remedy in every case. A claimant may encounter genuine obstacles in identifying a defect, obtaining relevant evidence, or establishing which actor caused a loss. Several systems may interact in ways that complicate the analysis. Such difficulties can justify changes to disclosure rules, liability standards, or compensation arrangements without necessarily requiring a new legal person.
The EU AI Act illustrates a regulatory method of allocating obligations according to identified roles rather than attributing general legal personality to the system. Compliance with regulatory requirements and civil liability remain distinct questions. A regulatory breach does not, without the applicable legal basis and other necessary elements, establish every requirement for a damages claim (European Union, 2024, 2026).
Creating an AI defendant would introduce questions of its own. The law would have to determine which model, deployment, or other entity bore the obligation, whether that entity continued to exist after modification, and what resources were available to meet a judgment. Separate personality would not, by itself, establish the cause of harm or make evidence more accessible.
A limited artificial entity might nevertheless serve a defined legal purpose. For example, a legislature could create a continuing holder of designated assets and obligations for a system operating across several deployments. Whether that structure offered an advantage would depend on the transactions involved, its funding arrangements, and its relationship with the rights and responsibilities of existing persons.
The conditions for such recognition would need to be specified in law. They include a stable means of identifying the entity, the capacities conferred upon it, arrangements for representation, assets available for enforcement, and rules governing modification or termination. The continued responsibility of developers, manufacturers, deployers, and other relevant actors would require particular attention.
AI accountability and AI personhood are consequently different legal questions. Accountability may be strengthened through clearer duties, attribution rules, access to evidence, and compensation mechanisms. A new juridical person would address a distinct problem only where its separate identity and enforceable obligations achieved something those arrangements could not adequately provide.
Also read
Conclusion
Artificial intelligence could be considered for a limited form of legal personality if an applicable legal system expressly established that status. The recognition of corporations and certain other nonhuman entities demonstrates that biological humanity is not an indispensable condition for every kind of juridical personhood. Intelligence or autonomous behavior, however, does not itself create legal rights or duties.
Existing judicial decisions concerning AI inventorship and authorship interpret particular intellectual property statutes. They do not resolve every conceivable form of artificial legal personality. International legal personality is likewise distinct from domestic juridical status: treaties regulating AI and the international recognition of organizations do not make AI systems independent subjects of international law.
The legal consequences of recognition would have to be defined rather than presumed. Commercial capacities, rights protecting an entity's own interests, civil liability, and criminal responsibility rest on different foundations. A system might receive one form of capacity without qualifying for another.
The decisive question is whether a particular AI entity could be given a sufficiently stable identity, meaningful legal capacities, and obligations that injured parties and public authorities could enforce. Where existing law can attribute transactions, regulate conduct, and provide remedies through identifiable human and corporate actors, artificial personhood is not necessary merely because AI performs tasks autonomously. Where a legislature identifies a distinct need for an artificial rights-holder, recognition would have to address that need without weakening the responsibilities of those who develop, control, or deploy the system.
References
Committee on the Rights of Persons with Disabilities (2014) General comment No. 1 (2014): Article 12: Equal recognition before the law, CRPD/C/GC/1, 19 May.
Companies Act 2006 (United Kingdom), c. 46, Royal Assent 8 November 2006.
Convention on the Rights of Persons with Disabilities (2006) adopted 13 December 2006, entered into force 3 May 2008, 2515 UNTS 3.
Council of Europe (2024) Explanatory report to the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law, CETS No. 225. Strasbourg: Council of Europe. Available at: https://rm.coe.int/1680afae67 (Accessed: 10 September 2026).
Council of Europe (2026a) European Union ratifies the Council of Europe Framework Convention on Artificial Intelligence [online]. Available at: https://www.coe.int/en/web/artificial-intelligence/-/european-union-ratifies-the-council-of-europe-framework-convention-on-artificial-intelligence (Accessed: 10 September 2026).
Council of Europe (2026b) Chart of signatures and ratifications of Treaty 225 [online]. Available at: https://www.coe.int/en/web/conventions/full-list?module=signatures-by-treaty&treatynum=225 (Accessed: 10 September 2026).
Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law (2024) adopted 17 May 2024, opened for signature 5 September 2024, CETS No. 225.
European Parliament (2017) Resolution of 16 February 2017 with recommendations to the Commission on Civil Law Rules on Robotics, 2015/2103(INL), P8_TA(2017)0051.
European Parliament (2020) Resolution of 20 October 2020 with recommendations to the Commission on a civil liability regime for artificial intelligence, 2020/2014(INL), P9_TA(2020)0276, OJ C 404, 6 October 2021, pp. 107–128.
European Union (2024a) Directive (EU) 2024/2853 of the European Parliament and of the Council of 23 October 2024 on liability for defective products and repealing Council Directive 85/374/EEC, OJ L, 2024/2853, 18 November 2024.
European Union (2024b) Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), OJ L, 2024/1689, 12 July 2024.
European Union (2026a) Corrigendum to Directive (EU) 2024/2853 of the European Parliament and of the Council of 23 October 2024 on liability for defective products and repealing Council Directive 85/374/EEC, OJ L, 2026/90364, 7 May 2026.
European Union (2026b) Regulation (EU) 2026/1744 of the European Parliament and of the Council of 8 July 2026 amending Regulations (EU) 2024/1689, (EU) 2018/1139 and (EU) 2023/1230 as regards the simplification of the implementation of harmonised rules on artificial intelligence (Digital Omnibus on AI), OJ L, 2026/1744, 24 July 2026.
International Court of Justice (1949) Reparation for Injuries Suffered in the Service of the United Nations, advisory opinion, 11 April, ICJ Reports 1949, p. 174.
New York Court of Appeals (2022) Matter of Nonhuman Rights Project, Inc. v. Breheny, judgment, 14 June, 38 NY3d 555.
Rome Statute of the International Criminal Court (1998) adopted 17 July 1998, entered into force 1 July 2002, 2187 UNTS 3.
Solum, L.B. (1992) ‘Legal personhood for artificial intelligences’, North Carolina Law Review, 70(4), pp. 1231–1287.
Te Awa Tupua (Whanganui River Claims Settlement) Act 2017 (New Zealand), Public Act 2017 No. 7, Royal Assent 20 March 2017.
United Kingdom Supreme Court (2023) Thaler v. Comptroller-General of Patents, Designs and Trade Marks, judgment, 20 December, [2023] UKSC 49.
United Nations Commission on International Trade Law (2024) UNCITRAL Model Law on Automated Contracting [online]. Adopted 11 July 2024. Available at: https://uncitral.un.org/en/mlac (Accessed: 12 September 2026).
United Nations Educational, Scientific and Cultural Organization (2021) Recommendation on the Ethics of Artificial Intelligence. Adopted 23 November 2021. Paris: UNESCO.
United Nations General Assembly (1948) Universal Declaration of Human Rights, Resolution 217 A (III), 10 December.
United States Court of Appeals for the District of Columbia Circuit (2025) Thaler v. Perlmutter, judgment, 18 March, 130 F.4th 1039.
United States Court of Appeals for the Federal Circuit (2022) Thaler v. Vidal, judgment, 5 August, 43 F.4th 1207.
United States Copyright Office (2025) Copyright and Artificial Intelligence, Part 2: Copyrightability. Washington, DC: U.S. Copyright Office. Available at: https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf (Accessed: 13 September 2026).
United States Patent and Trademark Office (2025) Revised inventorship guidance for AI-assisted inventions [online]. 26 November. Available at: https://www.uspto.gov/subscription-center/2025/revised-inventorship-guidance-ai-assisted-inventions (Accessed: 13 September 2026).
