Bloomington, Indiana became the first US city to pass a resolution on AI extinction risk. Here is the liability strategy behind it, and how to bring it to your town.
Most conversations about AI safety happen in San Francisco boardrooms or federal hearing rooms far from everyday life. On this episode of For Humanity, host John Sherman sat down with two guests who took a different path. They went to a city council.
Peter Jensen, a biologist and longtime AI safety advocate, has spent years trying to get a single law passed somewhere, anywhere, that would ban superintelligence unless it can be proven safe. Dave Rollo, a biologist and a member of the Bloomington, Indiana city council for 24 years, was in a position to actually pass one. John introduced them. Within a matter of weeks, according to Rollo, Bloomington adopted the resolution nine votes to zero.
Rollo believes it may be the first city anywhere to formally take a position on the extinction risk of artificial general intelligence. Whether or not that record holds, the approach is worth studying, because it is peaceful, lawful, and built to be copied.
Rollo did not arrive at this overnight. He described first reading the case for AI risk around 2022, and then, in his words, looking hard to falsify the argument that advanced AI could threaten human survival. He says he could not find anything convincing to rule it out. He attempted a local resolution back in 2023, pulled it, and returned this year to see it adopted.
Jensen frames the ask in the simplest possible terms. The core law he campaigns for is roughly 160 words. As he put it during the episode, the choice for any community comes down to a plain question: do you want superintelligence built in your jurisdiction, or not.
Both guests were careful to separate the tools people use every day from the systems they are worried about. Jensen argued that narrow, useful AI can deliver real benefits, and that the concern is specifically about autonomous systems more capable than humans across every domain. The resolution, in their telling, is not anti-technology. It is pro-caution.
The most striking idea in the conversation is not the resolution itself. It is the mechanism the guests believe it could trigger.
Jensen, drawing on how liability has forced other industries to withdraw dangerous products, explained the logic this way. Corporate directors and officers are protected by what is called directors and officers liability insurance. If cities and states pass laws making it a legal risk to build or deliver superintelligence, Jensen argues, insurers will look at a potential harm they cannot realistically price and simply decline to cover it. Once the personal liability lands on individual directors rather than the insurer, he suggests, the market itself pushes companies to pull back.
John summarized the appeal in his own words. Laws and treaties are vital, he said, but large companies are skilled at working around them. Making unsafe AI unprofitable, and personally risky for the people signing off on it, is a different kind of pressure. As he put it, that is where the metal meets the road.
We share this as the guests presented it, not as settled fact. Whether liability plays out exactly as Jensen predicts is untested. But the underlying point stands on its own: shifting incentives may matter as much as passing rules.
Explaining existential risk to a city council is hard. The guests credited a framing from Anthony Aguirre of the Future of Life Institute, who appeared live at the Bloomington meeting to answer questions.
The idea, as Jensen relayed it, is that advanced AI has three properties: autonomy, generality, and high intelligence. You can safely combine any two of them. Combine all three, and the system becomes something no one can reliably keep under control. It is a memorable way to draw a line between the assistants people find helpful today and the frontier systems that worry safety researchers.
Rollo, speaking as a biologist, offered a second lens. He described superintelligence in ecological terms, as a potential competitor species that could outcompete humans for resources, and pointed to the concept of a critical threshold, a tipping point where one state suddenly flips into another. His conclusion was that this is a line we should not cross, precisely because it may not be reversible.
Rollo does not expect one Indiana city to change the trajectory of a global industry. He described the plan differently, as what he called a Lilliput strategy. No single town is powerful, but many towns, counties, and states acting together can create real pressure and send a signal upward to state and national representatives.
He and John both pointed to a practical truth about politics. Elected officials look for social proof before they act. Once one city has moved, the next one has cover to follow. Rollo said he is now taking the Bloomington resolution to the statehouse and to nearby university towns, using it as a template.
That is also why the guests placed hope in young people. John and Rollo discussed graduation audiences pushing back on AI messaging, and argued that college communities, worried about their jobs and their future, could become an organized and vocal constituency for stronger AI safety policy.
The most useful part of the episode was a short role play. Rollo walked through how a resident might approach their own council. Show up to the public comment period. Keep it to a few clear points. Cite the researchers who have spoken out, including figures like Stuart Russell and Roman Yampolsky. Bring the Bloomington resolution as a ready-made template. Then follow up, and try to bring a few neighbors with you.
He also offered a tip for anyone nervous about starting. If you know a council member who might sponsor it, approach that person first, because a sponsor can carry the argument to their colleagues. If you do not, put it forward anyway.
Jensen closed with a line from Stuart Russell that captures the whole effort. You do not have to be a genius, he paraphrased, to realize that building something smarter than yourself could be a problem.
This is what democratic, community-level action on AI safety can look like. Not panic, and not violence, but neighbors showing up, asking questions, and putting a formal position on the record. John challenged viewers to help pass resolutions in two more cities, and both guests offered to help anyone who wants to try.
If that is you, the path is open. Start local, stay lawful, and build the coalition city by city.
