Corporations already do. Rivers do in some countries. The question of AI personhood is closer than you think, and the answer will reshape liability, labor, and law.
In 2017, Saudi Arabia granted citizenship to Sophia, a humanoid robot made by Hanson Robotics. It was a publicity stunt. Sophia is a chatbot with a face. She cannot vote, own property, or be held liable for anything. The gesture was marketing, dressed up as geopolitics.
But the question behind the stunt keeps coming back. As AI systems grow more autonomous, more capable of independent action, and more embedded in economic life, the legal frameworks designed for humans and corporations strain to accommodate them. Who is liable when an autonomous AI agent enters a contract on your behalf and the deal goes bad? Who owns the intellectual property an AI system generates? If an AI trading bot causes a flash crash, whom do you sue?
These are practical problems that need legal answers. And "legal personhood" is the tool that legal systems have historically used when they need to assign rights and responsibilities to something that is neither a natural human nor quite a simple piece of property.
Legal personhood has never required consciousness.
The most important thing to understand about legal personhood is that it has nothing to do with being a person. It is a legal fiction, a tool for organizing rights and obligations.
Corporate personhoodThe legal doctrine that a corporation possesses some of the same legal rights as a natural person, including the right to enter contracts, sue and be sued, and own property. Established in U.S. law through a series of Supreme Court decisions beginning in the 19th century. is the clearest precedent. A corporation can own property, enter contracts, sue and be sued, and exercise certain constitutional rights. Nobody believes a corporation has feelings. The legal system grants it personhood because doing so is useful: it creates a stable entity that can be held accountable, enter into agreements, and persist beyond the lifespan of any individual owner.
Other non-human entities have received similar treatment. The Whanganui River in New Zealand was granted legal personhood in 2017, giving it the right to be represented in court. The Ganges and Yamuna Rivers received the same status in India the same year. Hindu temples in India have long been recognized as legal persons for purposes of property ownership. Ships have a form of legal personality under admiralty law.
The pattern is consistent. When a society needs a non-human entity to bear legal weight, it creates a legal person. The question is whether AI systems have reached the threshold where the legal system needs them to carry that weight.
When AI acts autonomously, current law breaks down.
Today, if an AI system causes harm, someone else takes the blame. The developer, the deployer, or the user. This works when AI is a tool that humans direct. It breaks when AI acts on its own.
Consider an AI agent authorized to negotiate contracts on behalf of a business. The agent, operating within its parameters, signs a deal that costs the business $2 million. The business owner says the AI exceeded its authority. The counterparty says a deal is a deal. The developer says the system performed as designed. Current contract law was built for humans and the entities humans control. An autonomous agent that makes independent decisions in real time fits awkwardly into every existing category.
The European UnionThe EU AI Act (adopted 2024, obligations effective August 2025) treats high-risk AI systems as regulated entities subject to oversight but stops short of granting personhood. The EU withdrew its AI Liability Directive in February 2025. tried to address this through an AI Liability Directive, then withdrew it in February 2025 after sustained industry resistance. The EU AI Act takes a different approach: it regulates AI systems based on risk level without granting them legal standing. This works for now. It will not work when AI systems begin operating as genuinely autonomous economic actors.
The UK's Law CommissionIn August 2025, the UK Law Commission published a paper examining the possibility of creating a specific legal status for AI entities, describing it as a 'radical' option worth serious consideration. went further in August 2025, publishing a paper that raised AI personhood as a serious possibility. The paper was careful to frame it as one option among several, but the fact that a body responsible for reforming English law put it on the table signals how quickly the legal landscape is shifting.
Walking through the practical consequences sector by sector.
Legal personhood for AI would mean something specific and limited. It would mean that an AI entity could hold certain rights and bear certain obligations under law, in the way that a corporation can. It would not mean granting AI the right to vote, marry, or claim human dignity.
The most likely form is what scholars call bounded legal statusA proposal to grant AI limited legal recognition in specific high-stakes domains (financial services, medical diagnostics, autonomous vehicles) while preserving human accountability. Think of it as a license, not a birth certificate.: a context-specific recognition in defined domains. An AI trading system might have legal standing in financial markets. An autonomous vehicle's decision-making system might have limited liability status on roadways. A medical AI might bear partial responsibility for diagnostic recommendations. Each domain would define the boundaries.
AI personhood creates a clear liability target when autonomous systems cause harm. It enables AI to hold assets, enter contracts, and participate in economic life with legal predictability. It distributes responsibility more accurately across the human-AI chain rather than forcing all liability onto human operators who may have limited control over an AI's specific decisions.
AI personhood becomes a liability shield for the humans who build and deploy these systems. If an AI entity bears legal responsibility, the corporation behind it can deflect blame. It creates perverse incentives: the more autonomous a system is, the less accountable its creators become. The corporate personhood analogy cuts both ways, because limited liability has already proven to be a tool for avoiding consequences.
The liability shield concern is the strongest objection. When corporations gained personhood, it enabled extraordinary economic growth. It also enabled extraordinary evasion. Corporations can declare bankruptcy, dissolve, and reform under new names. They can be fined amounts that represent rounding errors on quarterly earnings. If AI personhood follows the same pattern, an AI entity that causes harm could simply be "shut down" while its creators walk away clean.
The counterargument is design. A well-crafted AI personhood framework could require that human principals remain jointly liable for the actions of AI entities they deploy. This would preserve accountability while solving the practical problem of who stands behind an autonomous decision. The corporation analogy is instructive precisely because its flaws are known: we can build the next version to avoid the same mistakes.
I litigate product liability cases. Somebody gets hurt, we figure out who's responsible. For thirty years the chain was clean: manufacturer, distributor, retailer, user. Four links. You trace the defect, you find the link, you assign the damages. It worked because at every point a human made a decision, and decisions carry liability.
Last year I got a case where an AI procurement system at a hospital independently switched medical supply vendors based on cost optimization. The new vendor's product failed. Patient got hurt. My client built the AI. The hospital deployed it. The vendor supplied the product. The AI made the actual choice. Three humans and a machine, and the machine is the one that pulled the trigger.
I spent four months trying to figure out who to hold responsible. Under current law, the answer is unsatisfying: everyone and no one. The AI had no legal existence. It was a ghost in the chain. And I keep thinking, if that ghost had a legal identity, if it could hold insurance and bear responsibility, I could have closed that case in six weeks.
Why AI personhood might arrive through commercial pressure rather than philosophical argument.
The philosophical debate about AI consciousness may never resolve. The economic pressure for AI legal standing is already building.
AI agents are beginning to operate as independent economic actors. They negotiate prices, execute trades, manage supply chains, and allocate resources. Each of these actions creates legal exposure that current frameworks handle through workarounds: the AI's actions are legally attributed to the human or company that deployed it. But as AI systems grow more autonomous, the gap between who made the decision and who bears the legal responsibility widens.
The insurance industry is watching closely. Insurers need to assign risk, and risk requires a legal entity to attach to. If an AI system can be insured as an entity, insurers can price the risk, companies can buy coverage, and injured parties can collect. Without AI legal standing, insurance coverage for autonomous AI actions remains in legal limbo.
Intellectual property presents another pressure point. Current U.S. patent law requires a human inventor. But AI systems are generating novel designs, drug compounds, and engineering solutions. The companies that deploy these systems want patent protection for the output. Granting AI entities the ability to be named as co-inventors would solve the problem. Several jurisdictions are already debating this.
What a responsible AI personhood framework would look like.
If AI personhood arrives, the design of the framework matters more than the fact of its existence. History offers lessons in what to avoid.
A responsible framework would include several features. Joint liability, meaning the human principals who deploy an AI entity remain legally accountable alongside it, preventing the liability-shield problem. Mandatory capitalization, meaning an AI entity would be required to hold sufficient assets or insurance to cover potential damages, similar to capital requirements for banks. Transparency obligations, meaning the AI entity's decision-making processes would be subject to audit and disclosure. And sunset provisions, meaning AI legal status would expire and require renewal, preventing abandoned AI entities from persisting as legal ghosts.
The worst version of AI personhood is the one that arrives by accident, through court rulings on edge cases rather than through deliberate legislative design. The best version is the one we build on purpose, learning from every mistake the corporate form has taught us.
Idaho and Utah have already moved to preemptively block AI personhood within their borders. Whether those laws hold as AI systems become more autonomous remains to be seen. Legislative resistance to new legal categories has a poor track record against economic necessity. The question is not whether AI will gain some form of legal standing. The question is whether the framework will be designed thoughtfully or cobbled together from lawsuits.
Legal personhood is a mirror. It reflects what a society considers important enough to protect and accountable enough to hold responsible.
Corporations gained personhood because commerce needed stable entities. Rivers gained personhood because ecosystems needed advocates. If AI gains personhood, it will be because autonomous systems need a legal identity that matches their economic reality. The decision will be pragmatic, not sentimental. And the framework we build around it will say as much about our values as any constitution.
The tools are available. The precedents are clear. The economic pressure is building. What remains is the will to design the answer before the courts are forced to improvise one.
