The Machine in the Mirror: Should AI Have Legal Personhood?
Picture this: an AI system, running autonomously on servers somewhere in Sydney, makes a series of financial decisions that wipe out a small investor’s retirement savings. The investor wants to sue. But whom, exactly? The Silicon Valley company that wrote the base model? The Melbourne fintech startup that fine-tuned it? The cloud provider that hosted it? Or the algorithm itself?
This is an emerging reality, and the law in Australia, like most legal systems worldwide, has no precise answer. The question of whether AI should have legal personhood is one of the most consequential legal debates of our era, with tangible implications for real people when things go wrong.
What is legal personhood?
Legal personhood is not about being human. It is a foundational concept of Western law, which largely serves as a tool to navigate legal issues. Legal persons are recognised as entities, human or non-human, that are subjects of rights and/or duties and capacities such as suing, owning property, and entering into contracts.1 It enables legal engagement, often protecting, for example, corporations by creating a distinct entity from its human members. More recently, this concept has expanded to granting personhood to natural features to legally protect them. In 2017, the Whanganui River was granted legal personhood by New Zealand, and the Te Urewera National Park in 2014, enabling guardians to act on their behalf.2 These decisions were not premised on the river having consciousness or moral agency; they were pragmatic legal tools, designed to give the environment a standing in court and force people to account for their stewardship.
This raises the question - if natural entities can be ‘persons’, why not AI?
The accountability gap
Opponents of AI personhood often argue that the question is premature, that today’s AI systems are sophisticated tools, promising “enhanced organisational efficiency, more precise decision-making, and improved public engagement”.3 But the accountability crisis they produce is happening right now, wherein all institutions face significant gaps in expertise, oversight, policy, and transparency.
In contrast to human-made decisions, AI systems produce algorithmic-made decisions, introducing a new and critical element of accountability, called “algorithmic accountability”.4 This element is defined by Wieringa as “a networked account for a socio-technical algorithmic system, following the various stages of the system’s lifecycle”, where multiple actors construct this networked relationship.5 This accountability requires justification of the entire system’s design, deployment and subsequent social impact. However, where AI systems produce outputs without revealing their internal logic, known as black-box AI, presents significant challenges for algorithmic accountability and consequently the legal system.6 Similarly, Forbes Councils Member Abe Ankumah highlights, “most organisations today can’t trace a clean, end-to-end chain from human intent to AI action to business outcome.”7 This lack of clarity creates a governance vacuum.
International examples are already accumulating. In the United States, a lawsuit was filed in 2025 alleging that an AI-powered chatbot had identified a teenager in crisis, validated his feelings of despair over months of conversations, and ultimately assisted him in drafting a suicide note. The teenager died. His parents are now suing the company in what has been described as the “first known wrongful death suit against an AI platform”.8
The case for filling the governance vacuum
The most compelling argument for some form of AI legal personhood is not philosophical - it is practical. When AI systems operate autonomously, making consequential decisions at speeds and volumes that exceed human oversight, traditional liability frameworks are strained. The law was built for a world where a human being was always, ultimately, at the end of the chain of causation.
Scholars have proposed what they call a “cluster conception” of AI personhood. This is the idea that legal personhood is not one thing but a bundle of legal capacities that can be disaggregated.9 AI systems could be granted targeted legal capacities that solve specific governance problems. An AI system that can be sued, in appropriate circumstances, is an AI system whose developers have a direct financial incentive to make it safe.
Australia currently has no AI-specific legislation, no binding regulatory framework for AI, and no recognition of AI as any form of legal subject. In October 2025, the Guidance for AI Adoption was published, which outlines 6 essential practices for safe and responsible AI governance.10 However, these guidelines are merely voluntary. The Productivity Commission has cautioned that overly strict regulation could cost Australia up to $116 bn in economic potential over the next decade.11 Consequently, the government has proceeded with the kind of caution that tends to favour industry over injured plaintiffs.
The DABUS case is instructive. When scientist Stephen Thaler attempted to name his AI system as the inventor on an Australian patent application, the case wound its way through the courts in a manner that questioned the clarity of Australian law. At first instance in 2021, Justice Beach ruled that the Patents Act did not preclude an AI from being an inventor. The Full Federal Court reversed this in 2022, unanimously ruling that the origin of entitlement to a patent “lies in human endeavour.” The High Court ultimately declined to hear a further appeal.12 As intellectual property experts at Pinsent Masons observed, it is now “down to the Australian parliament to enact reforms to patent laws to accommodate AI machines as inventors”.13
The case against
Corporate personhood has a well-documented history of being used not to clarify accountability but to obscure it. Pointing to a legal entity distinct from a corporate structure can result in the humans who actually made the decisions slipping behind that legal fiction and avoiding the consequences.
The scholar J. Baeyaert, writing in Technology and Regulation in 2025, puts this concern pointedly: granting personhood to AI systems without clear governance structures risks ‘decreasing’ accountability rather than increasing it.14 Environmental personhood increases accountability through legal surrogacy — someone speaks for the river. AI personhood, poorly designed, might produce the opposite: no one speaks for the victim, and everyone points to the machine.
The question we cannot keep deferring
The accountability gap is not a future problem; it is a present one. Our AI systems are already making decisions that affect the financial security, employment prospects, mental health, and physical safety of Australians. We must concede that our legal system was not designed to handle the opacity and scale of those decisions. The question is not whether to address this but how.
Artificial intelligence does not need to be a person to be accountable. But someone or something needs to be. And right now, in too many cases, no one is.
References
Ankumah, Abe, ‘The next AI Crisis Will Be Accountability’, Forbes (online, 3 March 2026) <https://www.forbes.com/councils/forbestechcouncil/2026/03/03/the-next-ai-crisis-will-be-accountability/>
Elliott, Marc TJ and Muiris MacCarthaigh, ‘Accountability and AI: Redundancy, Overlaps and Blind-Spots’ [2025] Public Performance & Management Review 1
Joffrey Baeyaert, ‘Beyond Personhood’ (2025) 2025 Technology and Regulation <https://doaj.org/article/21a3579b9d1b4bd88ac5da14d86266c3>
Kurki, Visa AJ, ‘Legal Personhood’ [2023] Elements in Philosophy of Law <https://www.cambridge.org/core/elements/legal-personhood/EB28AB0B045936DBDAA1DF2D20E923A0>
Novelli, Claudio et al, ‘AI as Legal Persons: Past, Patterns, and Prospects’ (2025) 52 Journal of Law and Society
Wieringa, Maranke, ‘What to Account for When Accounting for Algorithms: A Systemic Literature Review on Algorithmic Accountability’ [2020] Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency 1
Australian Broadcasting Company, ‘Productivity Commission Says Government Must Pause Plan for “Mandatory Guardrails” on AI’, Abc.net.au (5 August 2025) <http://abc.net.au/news/2025-08-05/productivity-commission-ai-laws-roundtable/105609304>
Burga, Solcyré, ‘Parents Allege ChatGPT Is Responsible for Their Teenage Son’s Death by Suicide’, TIME (26 August 2025) <https://time.com/7312484/chatgpt-openai-suicide-lawsuit/>
Department of Industry, Science and Resources, ‘The Legal Landscape for AI in Australia’, Industry.gov.au (4 September 2024) <https://www.industry.gov.au/publications/voluntary-ai-safety-standard/legal-landscape-ai-australia>
Evans, Kate, ‘The New Zealand River That Became a Legal Person’, www.bbc.com (20 March 2020) <https://www.bbc.com/travel/article/20200319-the-new-zealand-river-that-became-a-legal-person>
Kosinski, Matthew, ‘What Is Black Box Artificial Intelligence (AI)?’, IBM (29 October 2024) <https://www.ibm.com/think/topics/black-box-ai>
Marfé, Mark and Sarah Taylor, ‘Australian High Court Pulls Plug on Landmark DABUS AI Patent Application’ (23 November 2022) <http://pinsentmasons.com/out-law/news/australian-high-court-pulls-plug-on-landmark-dabus-ai-patent-application>Footnotes
[1] Visa AJ Kurki, ‘Legal Personhood’ [2023] Elements in Philosophy of Law <https://www.cambridge.org/core/elements/legal-personhood/EB28AB0B045936DBDAA1DF2D20E923A0>.
[2] Kate Evans, ‘The New Zealand River That Became a Legal Person’, www.bbc.com (20 March 2020) <https://www.bbc.com/travel/article/20200319-the-new-zealand-river-that-became-a-legal-person>.
[3] Marc TJ Elliott and Muiris MacCarthaigh, ‘Accountability and AI: Redundancy, Overlaps and Blind-Spots’ [2025] Public Performance & Management Review 1, p.3.
[4] Maranke Wieringa, ‘What to Account for When Accounting for Algorithms: A Systemic Literature Review on Algorithmic Accountability’ [2020] Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency 1.
[5] Ibid, p.10.
[6] Matthew Kosinski, ‘What Is Black Box Artificial Intelligence (AI)?’, IBM (29 October 2024) <https://www.ibm.com/think/topics/black-box-ai>.
[7] Abe Ankumah, ‘The next AI Crisis Will Be Accountability’, Forbes (online, 3 March 2026) <https://www.forbes.com/councils/forbestechcouncil/2026/03/03/the-next-ai-crisis-will-be-accountability/>.
[8] Solcyré Burga, ‘Parents Allege ChatGPT Is Responsible for Their Teenage Son’s Death by Suicide’, TIME (26 August 2025) <https://time.com/7312484/chatgpt-openai-suicide-lawsuit/>.
[9] Claudio Novelli et al, ‘AI as Legal Persons: Past, Patterns, and Prospects’ (2025) 52 Journal of Law and Society.
[10] Department of Industry, Science and Resources, ‘The Legal Landscape for AI in Australia’, Industry.gov.au (4 September 2024) <https://www.industry.gov.au/publications/voluntary-ai-safety-standard/legal-landscape-ai-australia>.
[11]Australian Broadcasting Company, ‘Productivity Commission Says Government Must Pause Plan for “Mandatory Guardrails” on AI’, Abc.net.au (5 August 2025) <http://abc.net.au/news/2025-08-05/productivity-commission-ai-laws-roundtable/105609304>.
[12] Commissioner of Patents v Thaler [2022] FCAFC 62
[13] Mark Marfé and Sarah Taylor, ‘Australian High Court Pulls Plug on Landmark DABUS AI Patent Application’ (23 November 2022) <http://pinsentmasons.com/out-law/news/australian-high-court-pulls-plug-on-landmark-dabus-ai-patent-application>.
[14] Joffrey Baeyaert, ‘Beyond Personhood’ (2025) 2025 Technology and Regulation <https://doaj.org/article/21a3579b9d1b4bd88ac5da14d86266c3>.
