We should stop or dramatically slow frontier AI development until governance, alignment, and societal preparedness catch up.
Overview
The pace of AI development has outstripped society’s ability to govern, understand, or align it; a deliberate slowdown is necessary.
Classification
- Geographic concentration
- Global; strongest in US and Europe; FLI letter had international reach
- Tags
- governance
- security
- Scenario type (full)
- Governance / Transition scenario
- Human position
- Sovereign; humans deliberately assert control over AI development pace
- Time horizon
- Immediate to near term (0–5 years) for the pause itself
- Discourse status
- Mainstream in public debate; contested in industry; supported by significant safety segments
Impacts
- Mechanism
- Regulatory moratorium on frontier model training above compute thresholds, international treaties, voluntary commitments, and public pressure.
- Domain impacts
- Labor & Income Delays displacement, buying time for retraining; but also delays productivity gains. Education More time for education systems to prepare; but delays AI-powered tools. Governance & Democracy Central: creates space for regulatory and institutional development. War & Security Risk of unilateral defection: if one state pauses while others continue, strategic disadvantage results. Inequality & Class Temporarily preserves existing structures; doesn’t solve underlying drivers. Culture & Art Slows cultural transformation; may create generational divide between AI-native and AI-cautious. Meaning & Purpose Preserves human relevance in the near term. Family & Reproduction Minimal direct impact; maintains status quo. Health & Longevity Delays AI-driven medical breakthroughs; some argue this has real human costs. Rights & Agency Protects existing frameworks by slowing disruption pace. Environment Reduces AI energy demand near-term; delays AI environmental solutions. Existential Survival Reduces near-term risk; may not address long-term risk if the pause is temporary.
Discourse
Key proponents
Key institutions
Notes on critique
Pop culture
Featured works
Cultural note
Notes on pop-culture references
Acceptance
- Key assumptions
- Assumes a pause is enforceable globally; assumes time gained will be used productively; assumes competitive pressures can be managed.
- Primary audiences
- AI safety community, public intellectuals, some government officials, concerned citizens
Personas
- Persona 1
- The Precautionary Activist – Organizes protests outside AI labs; lobbies for compute caps.
- Persona 2
- The Concerned Scientist – Publishes safety research and signs open letters while continuing in the field.
- Hard-believer profile
- Name & Age: Dr. Hannah Sørensen, 56. Occupation: Former AI researcher who left her faculty position to run a full-time AI safety advocacy organization; co-authored the FLI open letter. Location: Copenhagen, Denmark. Daily Life: Mornings coordinating campaign strategy with PauseAI chapters across 12 countries. Afternoons in media interviews, writing op-eds, or testifying before parliamentary committees. Weekends spent at conferences or organizing public teach-ins about AI risk. Keeps a detailed timeline of AI capability milestones that she updates weekly with growing alarm. Media Diet: AI capabilities announcements (which she tracks obsessively), FLI publications, CAIS research, various AI governance newsletters. Reads Stuart Russell and Max Tegmark as core texts. Watches mainstream media AI coverage critically for accuracy. Has read Frankenstein three times and considers it ‘the most prescient novel ever written.’ Core Conviction: We are running an uncontrolled experiment on civilization. No other technology in history has been deployed at this speed with this little governance. Every month we delay a pause, the systems become more capable and the governance gap widens. The precautionary principle exists for exactly this situation, and we are ignoring it. Social Circle: AI safety researchers, science communicators, parliamentary staff, investigative journalists. Has strained relationships with former colleagues who stayed in capabilities research. Her family is fully aligned with her mission. Biggest Fear: That by the time the world agrees a pause is necessary, it will already be too late—the systems will be too capable, too distributed, and too economically entrenched to stop. That she will have spent years shouting into the void while the labs raced ahead. Biggest Hope: A binding international treaty on frontier AI development, analogous to the Nuclear Non-Proliferation Treaty, that creates a coordinated global slowdown and redirects resources toward alignment, governance, and societal preparation.
References
Notes on canonical texts
Notes on further references
A scene from this future
The March
Outside a frontier AI lab, approximately 2027
Sarah held a sign that said PAUSE and stood in a line of three hundred people outside the glass headquarters of a company that was building God.
She was a biology teacher. She’d been here twice before. The first time, ten people came. The second, forty. Today, three hundred. Next month, she thought, maybe a thousand. Maybe more. Maybe not enough.
Inside the building, behind the glass, she could see people working. Young, bright, purposeful. They were not evil. She knew this. She had read their safety papers. They were thoughtful, careful people building the most dangerous technology in history as quickly as possible because every other lab was doing the same thing and stopping felt like losing.
She understood the logic. She also understood that understanding the logic was not the same as accepting it.
A counter-protester across the street held a sign that said: AI WILL CURE CANCER. WILL YOUR SIGN? Sarah had no answer to this. She knew that delay had costs. She knew that the cancer research being done by AI systems right now was extraordinary. She knew that every month of pause was a month of lost progress on problems that killed people.
But she also knew—in her body, in her bones, in the part of her that had evolved over three million years to assess danger—that the species was moving faster than its institutions could follow, and that the distance between capability and governance was growing, and that history did not reward civilizations that built power faster than wisdom.
At noon, a company spokesperson came out and offered them coffee. It was good coffee. He was polite. He said the company took safety very seriously. Sarah believed him. She also believed that taking safety very seriously while racing to build the most powerful technology ever created was a contradiction the coffee could not resolve.
She went home and graded biology papers. Her students were fourteen and they used AI to cheat on their homework and she used AI to catch them and the whole thing felt like a metaphor she was too tired to unpack.
She set her alarm for tomorrow’s march. Same place, same sign. The building would still be there. The people inside would still be working. The distance between what they were building and what the world was ready for would be one day wider.
Last updated 22 May 2026