The debate over whether AI systems might have any traits resembling consciousness or sentience is one that has occupied our imagination over decades.^[1] One need only refer to the swathe of popular culture where robots develop their own consciousness and personalities, to understand how this has influenced our understanding and fears about AI as a form of technology. This has only been heightened with the coming of Large Language Models and their natural language capabilities. Frontier AI labs are paying increased attention to the question of ‘AI welfare’.^[2]As this conversation evolves, regulators have also been considering the question of whether AI deserves ‘personhood’ under the law, i.e, be considered as an entity requiring legal rights and entitlements.^[3]

There is currently no settled consensus on whether AI is, or is approaching, any form of sentience. Indeed, there is no settled definition of consciousness itself, to then extend that concept to AI. Furthermore, the concept of personhood itself has also long been contested in the law. Even between humans, the concept of personhood has been mutable, differing based on gender, race, and ethnicity. Rights based on deemed personhood have been granted to corporations to enable their functioning and construct liability provisions, and in the context of granting protection (for example, natural resources). Therefore, there is precedent for granting legal rights or entitlements akin to humans to non-sentient entities.^[4]

This gets more complicated in the context of AI – because the questions of sentience and consciousness are often conflated with the question of granting legal entitlements. Unlike corporations or natural resources, AI is seen as a technology which might, due to its actions, replace humans, cause societal unrest^[5], and develop complex psychological states of its own.

The question of AI personhood has taken the form of concrete regulatory proposals in the USA. As of May 2026, there were 23 bills introduced across US state legislatures dealing with AI personhood in different forms.^[6] Primarily, the goal seems to be to exclude AI from the ambit of legal personhood, by using legislation to hold that AI cannot be considered ‘conscious’ or sentient and cannot be considered a ‘person’ under any law in force. In simple terms, these seem to be an attempt to ban personhood for AI.

What is interesting to note is that while legal personhood has traditionally been a flexible concept, all the legislative proposals in the USA are focused on exclusion of AI from this concept in different formats.  This may, therefore, be driven more by fear, or reflexivity than robust legislative consideration. This is relevant given the growing backlash against AI in the USA.^[7] Further, these may also be driven by partisan and religious motivations. For example, the recent intervention by the Catholic Church through the Pope’s AI encyclical shows that there is a movement to establish the value of human thought amidst AI driven anxieties.^[8]

The premise of the exclusion of personhood for AI seems to be conflating the concepts of judicial or legal personhood (i.e, creating a legal fiction which allows an entity to function with certain legal entitlements) and natural personhood (considering an entity deserving of a certain moral status). Among the bills introduced in the USA, there are a variety of approaches – some carving out exceptions based on legal personhood, some focusing on specific categories such as banning AI free speech, and others categorising AI as ‘non-sentient’ or not conscious.^[9]

Rather than evaluating the merits of granting or not granting AI systems legal personhood, I argue that devoting regulatory resources to the question of AI personhood is premature at this juncture. This approach obfuscates questions about the design of AI systems – what is being built, deployed, what the systems can do, and their risks and benefits. The debate over the legal or moral status of AI may take attention away from the very real harms which are materializing due to the operation of these systems.

For example, in a lawsuit filed by the mother of a teenager who committed suicide after interacting with a chatbot on Character.ai^[10], the company took the defense that the outputs generated by the chatbot were protected under the First Amendment, as a form of free speech. This argument was rejected by a federal judge^[11], allowing the wrongful death lawsuit to move forward. The lawsuit is based on the premise that the chatbot and platform in question together constitute a product, and therefore subject to product safety and liability provisions.

Another example is a recent class action lawsuit brought against OpenAI in California, Amargo Couture v. OpenAI[12], alleging that OpenAI aided in the interception of the Plaintiffs’ communications with ChatGPT, allowing access to it by Meta and Google for the purposes of targeted advertising. Importantly, the lawsuit here treats OpenAI as a party which has conducted ‘surveillance’ on the class and intercepted communications, akin to wiretapping and eavesdropping using technological means. These are typically the sorts of actions we see against the ‘state’, i.e, governments which typically intercept communications in the case of investigating alleged crime or in the interests of national security, but rarely against private companies, which are normally subject to consumer data litigation. Further, the plaintiffs here are seeing ChatGPT as a party to the communication, which they do not want to be accessed, revealing our changing reliance on AI for confidential communication.

These examples demonstrate that whether the AI system in question has either legal or natural personhood is largely irrelevant. What matters more, is the incentives that drive the corporations developing these systems, how they are designed, and their impacts on users. How the operation of these systems is situated within the broader political economy of information capitalism, and how people are increasingly using these systems in more personal ways, is a far more relevant question. Whether or not users believe them to be conscious beings, and whether they are conscious beings, is a secondary question that has almost no bearing on their societal impact.

Therefore, more functional legal considerations, such as developing robust frameworks for responsibility and liability would serve a more practical purpose, rather than focusing regulatory resources on the mental or cognitive properties of AI. Further, the focus on AI personhood should not take away from the existing legal personhood that the corporations developing these AI systems do have. These corporations continue to be liable, owing duties of care to the users of their platforms. A more pragmatic approach is California’s proposed bill AB 2023 / SB 1119, which, in the context of child safety, would prohibit chatbots from claiming to children that they are “sentient,” “conscious,” or “human”. This approach is less concerned with the right of personhood for the system, but rather look at the adverse impacts of its design which may lead to a false claim, confusing children. A design-centered approach should be the focus of future regulation.

Certainly, the discussion of why AI should or should not be granted personhood might become increasingly relevant as the capabilities of these models become more sophisticated. We may see more matters come before courts which contend that AI systems ought to have legal entitlements flowing from personhood.

However, the current need is to immediately address how these systems are constructed, creating evidence-based policy which considers the very real societal risks stemming from the growing capabilities of general-purpose models, such as effects on mental health, misalignment, and loss of control.^[13] We will not require an assessment of personhood or consciousness for these risks to matter, nor for their effects to materialize.

References

^[1] Yoshua Bengio and Eric Elmoznino, ‘Illusions of AI Consciousness’ (2025) 389 Science 1090; Andrzej Porębski and Jakub Figura, ‘There Is No Such Thing as Conscious Artificial Intelligence’ (2025) 12 Humanities and Social Sciences Communications 1647; Ted Chiang, ‘No, Artificial Intelligence Is Not Conscious’ The Atlantic (4 June 2026).

^[2] Anthropic, ‘Exploring Model Welfare’ (Anthropic, 24 April 2025) <https://www.anthropic.com/research/exploring-model-welfare> accessed 24 July 2026; Anthropic, ‘A Global Workspace in Language Models’ (Anthropic, 6 July 2026) <https://www.anthropic.com/research/global-workspace> accessed 24 July 2026.

^[3] Daniel C Borges, ‘Artificial Personhood: The Implications of Recognizing the Legal Personhood of Artificial Intelligence’ (2026) 28 Vanderbilt Journal of Entertainment and Technology Law 411; Aaron Mak, ‘The Case Against AI Personhood’ Politico (27 May 2026).

^[4] Katherine B Forrest, ‘The Ethics and Challenges of Legal Personhood for AI’ (2024) 133 Yale Law Journal Forum 1175.

^[5] King’s College London, ‘Public Have More Fear Than Hope on AI and Future of Work, Study Finds’ (King’s College London, 19 May 2026) <https://www.kcl.ac.uk/news/one-in-five-britons-think-ai-will-create-civil-unrest-study-finds> accessed 24 July 2026.

^[6] Austin Smith, Lucius Caviola and Heather Alexander, ‘Denying Personhood to AI: An Analysis of U.S. State Legislation on AI Legal Status’ (SSRN Working Paper, 25 May 2026) <https://ssrn.com/abstract=6829981> accessed 24 July 2026.

^[7] Miles Bryan, ‘How public opinion is turning against AI’ (Vox July 20 2026) <https://www.vox.com/podcasts/495950/ai-populism-protests-violence> accessed 27 July 2026.

^[8] Encyclical Letter: Magnifica Humanitas Of His Holiness Pope Leo Xiv On Safeguarding The Human Person In The Time Of Artificial Intelligence. Available at: https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html.

^[9] Smith, Caviola and Alexander (n6).

^[10] Garcia v Character Technologies Inc, 2025 WL 1461721 (MD Fla, 21 May 2025).

^[11] ‘Florida Judge Rules AI Chatbots Not Protected by First Amendment’ Courthouse News Service (21 May 2025).

[12] Couture v. OpenAI Global, LLC 3:2026cv03000, May 13, 2026, U.S. District Court for the Southern District of California.

^[13] Yoshua Bengio and others, International AI Safety Report 2026 (DSIT 2026/001, 3 February 2026).

Marcin Wilkowski / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/