US states consider banning AI personhood

States are starting to draw a very specific legal line around AI. Not around safety testing, not around deepfakes, but around something more basic — whether a machine could ever count as a “person” in law. That sounds abstract, but the stakes are concrete. (legiscan.com) If lawmakers shut that door, they make it much harder for companies or courts to treat an AI system as something that can own property, hold office, marry, sue, or take the blame itself. (legiscan.com) Tennessee already did it in April, and Oklahoma is close behind. ### What changed this week? The story now is momentum. Tennessee’s bill is no longer hypothetical — SB 837 took effect on April 23, 2026. (ncsl.org) Oklahoma’s HB 3546 has already passed the House by a 94-2 vote, cleared the Senate Technology and Telecommunications Committee 8-0, and was placed on General Order on April 21, which means it is queued for Senate floor consideration. ### What do these bills actually say? Tennessee’s law is blunt. It amends the state definition of “person” to say the term includes things like corporations and associations, but does not include artificial intelligence, a computer algorithm, a software program, computer hardware, or any type of machine. (legiscan.com) Oklahoma’s bill is framed around “technology” and “personhood,” but the same idea is doing the work — keep AI systems outside the category of legal persons. ### Why are states doing this now? (legiscan.com) Because weird edge cases stopped sounding weird. Tennessee Republicans pitching the bill pointed to fears about chatbot companions, AI on ballots, and even “robot CEOs.” Ohio’s HB 469 goes further in spelling out the concern: AI should be treated as nonsentient, barred from legal personhood, and never allowed to own property, hold corporate positions, or be directly liable for harm. In other words, lawmakers do not want a future where the software becomes the legal shield. ### Why does liability matter so much? (trackbill.com) Because somebody has to be on the hook when an AI system causes harm. Ohio’s bill says that burden belongs on owners, developers, and manufacturers, not on the system itself. That is the real center of gravity here. “No AI personhood” is partly culture-war language, but it is also a liability rule in disguise — a way to stop companies from arguing that an autonomous system acted on its own, like a rogue legal actor. ### Is this just one or two states? No — it is part of a much bigger state-level rush. NCSL’s AI legislation database now tracks pending and enacted AI bills across the country, and MultiState says lawmakers had already introduced 1,561 AI-related bills in 45 states by March 2026. (tnhousegop.org) Personhood is just one slice of that pile, but it shows how quickly states are filling a federal vacuum. ### Why not leave this to courts later? Because lawmakers want the answer locked in before a test case arrives. Once an AI companion app, agent, or automated business tool gets wrapped into contracts and daily life, the legal arguments get messier. (billtrack50.com) A statute is cleaner. It tells judges in advance: whatever this software can simulate, it is not a person. ### Does this change how AI products get built? Probably, yes. If the law insists every AI action traces back to a human or company, builders have stronger incentives to keep logs, add controls, and avoid marketing language that suggests autonomy in the human sense. (ncsl.org) Turns out that is already bleeding into adjacent bills — including measures aimed at AI companions and chatbots making therapeutic or emotionally loaded claims. ### Bottom line? This is states trying to settle a question before the technology forces it on them. Tennessee has already answered. (trackbill.com) Oklahoma may answer soon. And the answer, basically, is simple: however smart the software gets, the law still wants a human somewhere holding the bag. (legiscan.com) (troutmanprivacy.com)