Ohio’s AI personhood ban risks outlawing the future

The Ohio legislature is to be commended for taking AI personhood seriously, but we argue the current proposed bill is overbroad.

In his recent article, Nick Evans quotes Ohio state lawmaker Thaddeus Clagget as saying, “(i)t makes no difference the ability of a donkey to speak, that does not make the donkey a human, okay?”

As directors of the Laboratory for the Future of Citizenship, which studies how technology is reshaping the legal and philosophical foundations of personhood and belonging, we agree. But we have concerns about the approach Ohio House Bill 469 takes to AI safety, not only for Ohio, but as precedent for other jurisdictions.

If passed, it would be one of the most sweeping and consequential pieces of AI legislation in the country because it would outlaw not only human-AI marriage, but also rule on a question of metaphysics — whether it is possible for beings made of metal and silicon to have feelings.

It would go even further than the recently-passed law in Utah, which declares that no AI can ever be a person, no matter how person-like it may become.

What’s in Ohio’s proposal banning AI personhood

The stakes are high.

According to a recent survey conducted by the Center for Democracy and Technology, 19% of around 1000 9th to 12th graders surveyed said that they or someone they knew had been in a “romantic relationship” with an AI.

To be clear, we agree that humans should not be allowed to marry large language models like ChatGPT.

We also agree that it is unlikely that today’s AI systems are conscious or sentient (one of us is a philosophy professor, and the other is a human rights lawyer. We have thought about these questions quite a bit).

The problem is that Ohio’s bill doesn’t stop there.

Ohio H.B. 469 defines “artificial intelligence” so broadly that it could apply to “any software, machine, or system capable of simulating human-like cognitive functions, including learning or problem solving.”

This is far-reaching.

Consider the future of medicine: neural implants that use AI to restore memory in people with Alzheimer’s, or prosthetic limbs driven by machine-learning algorithms interpreting neural signals.

Under the bill’s current language, such systems could qualify as “artificial intelligence.”

If part of a person’s cognition or motor control is mediated by AI, does that make them partly AI?

Would Ohio’s law mean that a veteran with a bionic limb or a patient with an AI-driven brain implant couldn’t legally marry?

The drafters of the bill surely didn’t intend that outcome, but that’s the point.

Overbroad definitions make bad law. Instead of limiting its scope to what currently exists and is known about AI, the law attempts to legislate the unknown and unknowable future of a tech that just a few years ago would have sounded like impossible science fiction.

Imagine if someone had written a law in the 1700s setting a forever speed limit of 10 miles per hour for all wheeled transport. That might have made sense for carts, but not so much for airplanes.

Yet that’s what this bill attempts to do for artificial intelligence.

Consider animals.

The law asserts that animals are not people. That is why you cannot marry them.

It is not necessary to add that animals don’t have feelings (as we philosophers would put it: that they aren’t “sentient”, or “conscious”). It is a question of fact whether animals have feelings, not a matter for the law to decide.

And we don’t need to deny that animals can have feelings to explain why it isn’t ok (and shouldn’t be legal) for humans to marry them.

Yet, bill H.B. 469 declares that no AI systems may ever be sentient or conscious.

Contrast H.B. 469 with the model rules on declaring a human to be brain dead, The Uniform Determination of Death Act, which contains numerous scientifically established indicators that must be met and is frequently reviewed and updated with the latest findings on consciousness in the human brain.

If the Ohio bill passes, no amount of testing or indicators could ever be enough to establish AI consciousness, no matter what might develop in the future.

None of this means we should open the door to exploitative or absurd forms of “AI marriage.”

Laws can and should protect vulnerable people from predatory relationships with software, as well as from corporate manipulation of emotionally dependent users.

But a sound approach would define the problem narrowly, for example, banning contracts or ceremonies purporting to create legal marriage with text-based auto-regressive LLMs, rather than issuing sweeping, metaphysical bans that will age as badly as 18th-century laws about airplanes.

We commend Ohio in taking the initiative to lead in AI governance, but what we need is humility.

Lawmakers should invite scientists, ethicists, and legal scholars into the drafting process and build in a requirement for periodic review as the technology evolves. AI is moving fast, but we must not panic.

Premature legislation can cause problems in the future, just as inadequate legislation may.