Across multiple countries and pollsters, large majorities say they want stronger rules on frontier AI, including on catastrophic risk. The puzzle is not public appetite. The puzzle is why that appetite has not yet become law with teeth.
Across countries and across the political spectrum, large majorities consistently favor caution, regulation, and even slowing development. Public resistance is real; the puzzle is why that mandate has not yet moved policy, when it once did.
What the polling consistently shows
Surveys across multiple countries and pollsters have found remarkably stable patterns. Large majorities express concern about the risks of ASI, including catastrophic risks. Majorities support government regulation of AI, often by wide margins. Substantial shares (frequently majorities) say they would support slowing down AI development, and many favor the idea of international coordination or agreements to manage it. These findings recur across the political spectrum, making AI caution one of the less polarized issues in an otherwise divided politics.
- Concern is broad. Majorities in many countries report worry about ASI's risks, and a meaningful share take catastrophic scenarios seriously.
- Regulation is popular. Support for government oversight of AI consistently outweighs opposition, often heavily.
- Slowing down has support. Contrary to the 'no one wants to slow AI down' narrative, sizeable majorities in some surveys back slowing development to get safety right.
- It crosses party lines. Unlike most high-salience issues, AI caution attracts support across the political spectrum, a rare basis for durable policy.
Why the mandate hasn't moved policy
If the public broadly supports action, why has so little binding regulation followed? The answer is not opposition but salience. For most people, AI risk is a real but low-priority concern, something they endorse when asked, but not an issue they vote on, contact representatives about, or organize around. It competes for attention with immediate economic worries, and it lacks the visible, mobilizing triggers that drive political action on other issues. A latent majority that is not activated exerts little pressure.
There is also an organization gap. Concentrated interests that favor rapid, unregulated development (well-funded, well-connected, and highly motivated) are far better organized than the diffuse public majority that favors caution. In the absence of a strong, visible movement translating public concern into political demand, policymakers hear the organized voice more clearly than the larger but quieter one. This is a familiar dynamic: diffuse majorities routinely lose to concentrated minorities not because they are outnumbered but because they are outorganized.
Why this is grounds for optimism
The strategic implication is encouraging. Movements that must first persuade a hostile public face a long, uncertain battle. ASI governance does not: the persuasion is largely done. The task is the more tractable one of raising salience and building organization, turning broad, latent agreement into visible, sustained political demand. Advocacy makes the issue concrete and urgent, and gives the existing majority a way to be counted.
The public is not the obstacle to ASI governance. The public is already broadly on side. The gap is between what people believe and what they are mobilized to demand, and closing that gap is exactly what a movement is for.
From latent majority to political force
Democratic governance of AI does not require changing what people think; it requires helping what they already think matter politically. The consistent polling is a foundation to build on: it means that a treaty or binding regulation would not be imposed against the public will but would express it. The unfinished work is to raise the issue's priority, to organize the diffuse majority into a force policymakers cannot ignore, and to ensure that when the political moment comes (as it will, whether through a warning event or the steady advance of the technology) there is an activated constituency ready to demand action. The mandate exists. Turning it into policy is the task, and it starts from a far stronger position than most assume.