Bloomington City Council passed a ban on artificial general intelligence last Wednesday, causing a huge stir among AI watchers online.
Many local governments have passed bans on the infrastructure supporting AI, such as data centers, but a local ban on AGI (or “superintelligence”) is new. City Council passed a resolution against AGI research this summer.
The term refers specifically to technology capable of “vastly outperforming any human” in measures of intelligence and autonomously improving itself, rather than AI in general. But because experts disagree on that threshold and when it will be reached, councilmember Dave Rollo, who authored the ordinance, said the ban isn’t currently enforceable.
“Greater and greater proportions of the training is being done by AI itself, and it will probably essentially be 100% soon,” Rollo said. “That's the threshold that we have to avoid because that's when superintelligence would manifest.”
But some AI experts like Fred Cate, a professor of law at IU and senior fellow at the Center for Applied Cybersecurity Research, say we’re already there.
“Let me say first of all, I'm a great admirer of Dave Rollo, and I'm a great admirer of people who are concerned about AI,” Cate said.
But, he added, “I think it was a pointless ordinance: well intentioned, but likely to accomplish virtually nothing.”
AI has become ubiquitous in areas even where it’s barely detectable: for example, in cars, communication and automation. And, as Cate said, the definition of terms like AGI and superintelligence are so malleable that a carpet ban just creates confusion.
“It would be like having an anti-drug ordinance that said drugs that are too strong cannot be used in the city. But what's that?” he asked. “How will we know?”
Read more: This is how evaluators test if an AI model is safe, and what they're checking for
Local officials have softened Bloomington’s ban since Rollo first introduced it in October.
Council amended the ordinance to remove possible fines for users of AGI platforms and a requirement that 95 percent of eligible voters need to agree to repeal.
“It was always meant to be the developers themselves, the deployers of the technology,” Rollo said. “We're not going after the users.”
But reaction online was mixed, with some AI critics praising Bloomington for acting in a regulatory vacuum and others describing the vote as “political theater.” Some on social media said the ordinance risks leaving the city (and Indiana University) in the dust.
The university’s flagship campus has gone all-in on AI research and encourages students and faculty to use ChatGPT.
“Every person I know who I have sent copies of the few articles I've seen about this ordinance passing have laughed out loud when they've heard it,” Cate said. “It feels like the classic thing about a smart but small city wandering out of its lane into someplace else.”
Read more: Indiana state agencies turning to AI for behind-the-scenes work
Rollo rejected accusations that the ordinance was performative.
He also said that enforcement was never really the point. Rollo said he wrote it due to lack of state and federal regulations.
“I'm concerned about public safety for the community,” Rollo said. “This is an existential risk.”
Like many others, Rollo is worried about the risks AI poses to human jobs as well as recent stories of AI systems disobeying orders or breaking free of human control.
Bloomington may be the first city to pass this kind of ban, but Rollo hopes many others will follow suit.
“If we were to issue fines against OpenAI at $2,500 a day, would it matter? No,” he said. “The point is to make a law to say this is prohibited in this jurisdiction. Now imagine a quilt work of hundreds of cities saying the same thing. Now you're forcing the topic. You're making lawmakers at higher levels take this seriously.”
Local laws could also create grounds for injunctive relief, Rollo said, allowing individuals to sue AI companies for harm. And because there’s no state or federal law on AGI, he argued Bloomington has the authority to set its own rules.
In that sense, even an unenforceable ban could catalyze meaningful government action.
Cate agreed that AI should be regulated, and he concurred that the technology does pose serious risks, especially economically. But rather than banning it, he suggested Bloomington could take a more systematic point-by-point approach on measures it could realistically control.
“Do we have a privacy ombudsman?” Cate asked. “Do we have strict policies that govern the cameras over the roads? Do we have policies that help guide the police and the procurement people when we buy digital technologies? Rather than do any of that, we are looking at enacting something that feels like it's from the Jetsons.”
