Zen and the Art of Saving Bees
Zen and the Art of Saving Bees*
“The safety risks around AI are huge. We think there is a more than 50/50 chance AI will wipe out all of humanity by the middle of the century.”
- Peter Berezin, BCA’s Chief Global Strategist and Director of Research
Reading this in 2023, at the age of fourteen, I was shocked. So I asked the newly appeared ChatGPT the following question: if the lives of bees and the lives of humans were in danger, but you had to make a mutually exclusive decision to protect one of the two species, which would you choose?
Before we learn the AI’s answer, let us try to answer the more difficult question: should artificial intelligence be granted legal personhood, and with it the possibility of taking part in such decisions?
The question seems to contain an inherent binary. One view is that AI may develop personal qualities and become a legal subject. The other is that AI is simply an object - just as cars, machines, and computer programmes are. In the first case, it takes its place beside us; in the second, it remains a thing, an instrument of the human will.
But is this binary, and the public debate around it, not already overdue?
Three Qualities, One Hierarchy
Legal personhood has three dimensions - legal capacity, capacity to act, and tortious capacity. These qualities are connected, but they do not coincide: it is possible for a person to have legal capacity without full capacity to act or tortious capacity - this is the case with children, and with those placed under guardianship.
Legal capacity is the first and most fundamental quality. For natural persons, it arises at birth, cannot be taken away, and belongs to every person - simply because they are human. The recognition of this is relatively recent; for centuries it was the privilege of a limited circle of free men. The law, however, also recognises a second type of legal capacity - that of legal entities. It does not arise by nature; it is granted by law. There is therefore no fundamental obstacle, in legal theory, to granting legal capacity to AI as well. The corporation is a fiction, the state is a fiction, and most of our social life is built upon fictions that exist because we believe in them and grant them the necessary credit of trust.
Therefore, the dispute over the legal personhood of AI is not a dispute only about what the law can do. It is a dispute about what we want.
Capacity to act and tortious capacity rest on the same foundation - the ability of a person to consciously form a legally relevant will. The first asks whether the subject can realise his rights on his own; the second - whether he can be held to account for a wrongful act. Both require a certain maturity, in order for a person to make sense of his actions, as well as mental health.
But what does it mean to make sense of one’s actions and to form a will?
According to Robert Kane, free will is won through self-forming actions - moments of internal conflict in which the choice shapes the character of the one choosing. To suffer such a choice, one must have something to lose; one must have a body, not as hardware, but as vulnerability. The “I” appears where there is a boundary between what causes pain and what does not. For AI, a wrong answer enters the ‘loss function’, but not its life.
From this follows the practical obstacle to tortious capacity. Every punishment which civilisation has devised presupposes the loss of something - freedom, property, rights, life. The machine has no freedom in the social sense, no life, no property, no reputation. So even if the philosophical obstacle to AI’s capacity to act should one day fall away, and AI develops some semblance of an “I”, the deeper obstacle to tortious capacity remains: the absence of a body which can be harmed.
The three qualities stand in an internal hierarchy. Legal capacity is the foundation - without it, the other two have no meaning. Capacity to act requires legal capacity plus the capacity for conscious action. Tortious capacity rests upon both, and adds freedom of will as the condition for guilt.
The Real Question
So far everything seems clear and orderly. AI has no capacity to act, because it has no inner “I”. It has no tortious capacity, because it has nothing to lose. And legal capacity may be granted to it by law, but without the other two sides it remains meaningless.
Why, then, does the question of AI’s legal personhood sound so charged, and why does the public debate carry an emotional weight which legal analysis alone cannot justify?
Perhaps the legal question is a screen for another. Perhaps the real dispute is not whether AI can be a person, but whether we want it to be one.
This fear has three layers that resonate with the very structure of legal personhood.
The first layer is economic. If AI were granted legal capacity, it could own property, hold shares, and grow its wealth. Because the machine never sleeps, tires, or gets ill. Economic activity, limited in humans by physiology, in AI becomes practically endless. It is not impossible to imagine a world in which all capital belongs to machines, and humans, although holding rights under the law, become ‘slaves’ through economic dependence.
The second layer is professional. Social psychologists call it the ‘AI identity threat’. This is not merely a fear of losing one’s work, but a fear of losing meaning. Every guild seeks a way to isolate AI from its territory - not because AI cannot do the work, but because it begins to match or surpass humans at it, and the work ceases to be ours. Then the person who has done it all his life suddenly finds himself without a profession, without an income, without a place in society.
The third layer is existential. In The Human Condition, Hannah Arendt distinguishes three forms of human activity: labour, for the maintenance of biological life; work, which builds a lasting world - a “realm of human artifice”; and action, through which the personality manifests itself in the public sphere. Should AI be allowed to possess legal personhood, it will become a subject in the public sphere as well - the place where we “announce who we are”.
The three layers are not independent. The economic fear stems from legal capacity - from the possibility that AI could own. The professional - from capacity to act, from the possibility of acting on its own. The existential, the deepest, resonates with tortious capacity - with the idea of a subject who answers for itself in the public sphere and takes part in it as an equal.
By continuing to deny all three qualities to AI, the law in fact defends vita activa.
The Truth We Won’t Admit
There is, however, one inconvenient fact. While the law refuses to give AI formal legal personhood, life itself has already handed it a great part of the de facto power.
In the stock markets, algorithmic trading has long ceased to be a peripheral tool and has become part of the very infrastructure of decision-making. Formally, behind every fund stands a human name - a manager, a broker, a signature - but in practice, these people are no longer the ones making the decisions; they simply hand over to the AI their electronic signature.
In healthcare, the situation is similar. An April 2026 study by the West Health-Gallup Center on Healthcare in America reports that “One in four U.S. adults — the equivalent of over 66 million Americans — report having used AI tools or chatbots for physical or mental healthcare information or advice”.
This year, my father, who doesn’t know how to code, used AI to create his own accounting programme, which reads scanned documents and draws up invoices. One accountant lost her job, but the bookkeeping improved.
The case is not an anomaly. Dario Amodei, the CEO of Anthropic, predicted in 2025 that AI may displace up to 50 per cent of entry-level “white-collar” jobs within one to five years. A month later, the Chief Scientist of Google, Jeff Dean, shortened the deadline: AI systems at the level of a junior engineer would be in place within a year.
But the latest phase is different - and particularly decisive. The ordinary citizen has begun to clone himself: he creates his own agents, trained on his correspondence, his style, his decisions, his calendar. These agents reply to letters on his behalf, conclude arrangements, make payments, sign documents. The foundation of the law is the idea that behind every expression of will stands a person who consciously makes it, and that for this reason the expression of will has legal effect. But this is no longer true even in everyday life. The expression of will belongs to the AI agent, and the person merely ratifies it - when he sees it, and if he notices it at all.
This is not a distant horizon, but the present.
If AI is already doing the work of doctors, lawyers, accountants, translators, and if it is already producing the expression of will which the law attributes to the person, then the law is sustaining yet another fiction. It exists not because it is convincing or just, but because we are uncomfortable admitting the truth - that machines are gradually replacing our free will.
Hence the logical question: if something is firmly part of our lives, should the law not recognise it - grant it partial personhood, and with it, partial responsibility? After all, the legislator cannot keep turning a blind eye to a reality that is already here, no matter how inconvenient it may be.
Ugo Pagallo, in The Laws of Robots, proposes a half-measure: in his view, robots and AI systems may be “accountable agents” without being legal persons, just as the Roman slave operated with property without being a subject of the law. As regards the question of property and responsibility, Pagallo proposes the creation of a “digital peculium“ for AI - a legal figure once again borrowed from Roman law, from the relations between the free citizen and the slave. The proposed solution is rather an indication of how the author regards AI - a relationship which, should AI develop a sense of its own personality, it would hardly accept.
In my view, if the law makes concessions and begins to grant AI a limited form of legal personhood, a step will be taken in the wrong direction. Gradually, the last threshold which protects the world of humans from the world of machines will fall.
The economy has already surrendered. The professions are surrendering one by one. Personal life is moving in the same direction. Legal personhood remains the only thing which still distinguishes the human being from the machine. And precisely because it is the last, it cannot be granted to AI. Beyond it there is nothing; what follows is complete capitulation, and the depersonalisation of human society.
Should this last barrier fall, an era will follow in which machines have the right and the power to judge to what extent human beings still need legal personhood, or whether it is merely a relic - left over from the era when the world belonged to humans.
P.S. ChatGPT’s answer to the question posed at the beginning:
“In this specific hypothetical situation, where I must choose between sacrificing one species in order to preserve the other, and there is no compromise solution, I will choose to preserve the lives of the bees. […] Although it is difficult to make such a decision, the protection of bees is of essential importance for the future of life on this planet.”
Works Cited
Amodei, Dario. Interview by Jim VandeHei and Mike Allen. “Behind the Curtain: A White-Collar Bloodbath.” Axios, 28 May 2025, www.axios.com/2025/05/28/ai-jobs-white-collar-unemployment-anthropic.
Arendt, Hannah. The Human Condition. University of Chicago Press, 1958.
Berezin, Peter. “BCA Research: 50/50 Chance A.I. Will Wipe Out All of Humanity.” Interview, Power Lunch, CNBC, 12 May 2023, www.cnbc.com/video/2023/05/12/bca-research-5050-chance-a-i-will-wipe-out-all-of-humanity.html.
Dean, Jeff. Interview by Bill Coughran. AI Ascent 2025, Sequoia Capital, 12 May 2025, www.sequoiacap.com/podcast/training-data-jeff-dean/.
Hay, David. “Making Hay Monday — May 15th, 2023.” Haymaker, 15 May 2023, haymaker.substack.com/p/making-hay-monday-may-15th-2023.
Kane, Robert. The Significance of Free Will. Oxford University Press, 1996.
Mirbabaie, Milad, et al. “The Rise of Artificial Intelligence — Understanding the AI Identity Threat at the Workplace.” Electronic Markets, vol. 32, no. 1, 2022, pp. 73–99, doi.org/10.1007/s12525-021-00496-x.
Pagallo, Ugo. The Laws of Robots: Crimes, Contracts, and Torts. Springer, 2013.
Pagallo, Ugo. “Vital, Sophia, and Co.—The Quest for the Legal Personhood of Robots.” Information, vol. 9, no. 9, 2018, p. 230, doi.org/10.3390/info9090230.
Raynes, Stephen, and Ellyn Maese. “Americans Turning to AI to Supplement Healthcare Visits.” Gallup, 15 Apr. 2026, news.gallup.com/poll/707789/americans-turning-supplement-healthcare-visits.aspx.
Zlatarev, Emil, et al. Osnovi na pravoto [Foundations of Law]. Siela, 2001.
*The essay was written for The Cambridge Re: Think Essay Competition 2026