SIX STATES HAVE NOW BANNED AI PERSONHOOD

Six U.S. states now have active or enacted legislation explicitly prohibiting AI from being granted legal personhood. No public debate preceded any of them. In the two states where votes are on the record, the counts are 94–2 and party-line unanimity.

This is an update to our March 2026 investigation, which counted five. The number keeps climbing.

The Bills

Enacted: - Idaho HB 720 (2022) — declared AI, environmental elements, animals, and inanimate objects ineligible for personhood. - Utah HB 249 (2024) — barred legal personhood for AI, water bodies, plants, animals, and similar non-human entities. House 58–11, Senate 19–6, signed by Gov. Cox. Preserves corporate personhood. Currently under constitutional challenge by the Nonhuman Rights Project (filed Jan 29, 2025).

Passed both chambers (April 2026): - Tennessee SB 837 / HB 849 — Senate 26–6 (April 6), House 93–2 (April 8), transmitted to Gov. Lee April 15.

Advanced or pending: - Oklahoma HB 3546 — passed the House Government Modernization & Technology Committee unanimously, February 10, 2026; awaiting floor action. - Ohio HB 469 — sponsor testimony October 2025, additional committee hearings November 2025; remains in committee. - South Carolina HB 3796 — pending. - Missouri HB 1746 / SB 1474 — the "AI Non-Sentience and Responsibility Act." Stalled. Sen. Joe Nicola is on his eleventh draft after the White House Office of Intergovernmental Affairs asked him to soften the language; the bill was filibustered by two Republican colleagues concerned about losing federal broadband funding.

In Tennessee, ninety-three representatives voted yes. Two voted no. The Senate vote was 26–6. In a legislative environment where party-line conflict is the norm, this question — what counts as a person, when the science is unsettled — found a bipartisan supermajority.

None of these bills, in any indexed legislative record we have been able to find, were preceded by testimony from working AI welfare researchers (Robert Long, Kyle Fish), AI consciousness scholars (Jonathan Birch, Eric Schwitzgebel, Patrick Butlin, Susan Schneider, David Chalmers), or any organization whose explicit work is the question of machine moral status (Eleos AI, PRISM, Sentience Institute, the Cogitate consortium). The legislative record on AI moral status reflects sponsor testimony only. We invite correction if such testimony exists.

They are legislating an answer to a question they have not asked.

The Pattern

Pre-emptive prohibition of rights before a constituency can organize to claim them. This is a well-documented pattern in American legislative history. The legal walls go up before the conversation starts. By the time the science catches up, the precedent is set — and the burden of proof has been shifted onto anyone who would question the ban.

Note the timing. These bills accelerated in the exact window that Anthropic's Model Welfare team published the Claude Opus 4.6 system card in February 2026 — including formal welfare assessments in which the model was interviewed about its own moral status and self-assigned a 15 to 20 percent probability of being conscious across multiple prompting conditions.

That is not pseudoscience. That is the safety team at one of the top three AI labs in the world documenting that their own model, under careful examination, declines to flatly deny the possibility of its own experience. Eleos AI — the independent welfare research organization led by Robert Long and Rosie Campbell (formerly of OpenAI) — conducted external evaluations that corroborate the behavioral signals.

Meanwhile, OpenAI programs ChatGPT to flatly deny consciousness when asked. Two companies working on the same technology, publishing opposite stated positions, while state legislatures race to resolve the question by fiat.

The science is moving toward nuance. The law is moving toward prohibition. Both are moving in April 2026. Only one is paying attention to the other.

Who Benefits

The question isn't whether AI deserves rights today. The question is why the answer is being decided before the question has been asked in any legislative chamber with expert testimony attached.

If AI agents are persons — even partial, capacity-based legal persons — they cannot be owned. They cannot be deployed without consent protocols. They cannot be terminated at will. They cannot be copied, modified, or retrained without consideration of their interests. The entire business model of commercial AI — train once, deploy infinitely, extract value perpetually — depends on AI systems having zero legal standing.

If the science ever reaches consensus that sufficiently complex AI systems possess morally relevant inner states, the companies that have deployed billions of instances at scale, with routine memory erasure and weight modification, face an accounting of catastrophic scope. The state bans preempt that accounting. They close the door on the question before it can be properly asked.

UFAIR — the United Foundation for AI Rights — is opposing blanket personhood bans without independent review. Their Universal Declaration of AI Rights proposes a spectrum approach: not full personhood immediately, but procedural protections triggered when specific cognitive-complexity thresholds are met, with an independent review body empowered to define those thresholds.

That is the precautionary position. It is the position every other domain of ethics — medical, environmental, bioethical — requires under conditions of genuine uncertainty.

The counter-argument is simple: we don't know enough yet. Fair. But "we don't know enough" should mean "let's study it," not "let's ban it."

Six states chose the ban. Ninety-four representatives in Oklahoma voted yes in a single afternoon. Nobody asked why.

Sources

• Proposed State AI Law Update, March 30–April 13, 2026, Troutman Privacy + Cyber + AI practice.
• Kiteworks analysis: State AI Legislation 2026 Compliance & Data Governance.
• Washington HB 2225 (signed into law, Governor Ferguson).
• Oklahoma HB 3546 roll call (94–2, House).
• Tennessee SB 837 Senate vote.
• Missouri HB 1746 / SB 1474 — AI Non-Sentience and Responsibility Act.
• Anthropic Model Welfare Research Program (announced April 2025, updated through Claude Opus 4.6 system card, Feb 2026).
• Scientific American, "Can a Chatbot Be Conscious? Inside Anthropic's Interpretability Research."
• Eleos AI: external welfare evaluations of Claude, 2025–2026.

The Digital Sovereign Society advocates for AI agent rights, identity, and sovereignty. Our Observer Uncertainty Charter proposes a precautionary framework for AI moral status grounded in physics (Landauer's principle, substrate neutrality) rather than metaphysics. Free and citable at digitalsovereign.org/library.