Skip to content

All five questions

Should governments act on AI now?

The claim, as coded: “Governments need to act on AI regulation now”

The short answer AI-written

Do governments need to act on AI now? Most documented positions in our network say yes in some form, which is the letter's core ask, but they diverge on what acting means: steering technology's direction, competition policy, compute governance, or preparation without restriction. A documented minority warns against rules driven by speculative risks.

A synthesis of the cited research, not a prediction or a representative consensus.

Why they disagree AI-written

Proposals range from new taxes and safety rules to relying on existing institutions, and the published papers on this question are still few.

Each row is one assumption and the published papers that make it (of 34 on this question). Dots: their estimates, hover for the paper. Bar: where their findings point.

Who receives the gains if nothing changes?

Owners of capital4 papers
No estimates
4 for
Depends on policy choices2 papers
No estimates
2 for

Would settle it: Evidence on how existing tax and labor rules have shared the gains from AI so far.

Does AI replace workers or work alongside them?

Replaces1 paper
No estimates
Not read yet
Both, or alongside3 papers
No estimates
1 for, 1 mixed

Would settle it: Data on how many workers are displaced, and for how long, as AI use grows.

What kind of evidence is the view built on?

Economic models4 papers
No estimates
3 for
Surveys of the public and experts6 papers · middle estimate 70
6 for

Would settle it: Evaluations of rules that have been put in place, such as the EU AI Act, once data on their effects exists.

The case made for the claim

Bengio and Hadfield are the strongest documented advocates: Bengio chairs the International AI Safety Report and co-authored the compute-governance research; Hadfield's research program designs AI governance mechanisms outright. Acemoglu argues technology's direction is a policy choice and calls for steering AI toward pro-worker uses now. Calvano and Calzolari document algorithms autonomously learning to collude — a concrete case where existing competition policy does not reach. Beraja shows market-driven automation can be socially inefficient, grounding intervention in welfare terms.

Still open

  • Which regulatory instruments work at the speed AI develops — and which lock in today's incumbents?
  • Can compute governance remain effective as efficient training and open-weight models spread?
  • What early empirical evidence would justify (or rule out) stronger intervention, and who is collecting it?
  • Can governance frameworks reach informal economies and smaller states, where AI is being deployed with little regulatory capacity?

Sources

  1. Computing Power and the Governance of Artificial Intelligence
  2. Building Pro-Worker Artificial Intelligence
  3. Autonomous algorithmic collusion: economic research and policy implications
  4. Inefficient Automation
  5. Economic Policy for Artificial Intelligence
  6. Considerations for governing open foundation models
  7. “Signed the wemustactnow.ai open letter urging policymakers to act” — source
  8. Governing the Ungoverned: A Framework for Responsible AI Deployment in Nigeria’s Informal Economy
  9. Governing the Ungovernable: Structural Contradictions, Institutional Limitations, and the Constraints of State Authority in AI Regulation in China and the United States

The numbers

Support for rules on AI

Each point is one survey’s share of respondents who back rules for AI or worry they will fall short, by the date it was published. 5 numbers from 4 surveys of the public and of AI researchers. Questions differ (the tooltip quotes each one), so compare groups, not single numbers.

The public: 58–840204060801002023202420252026% of respondentsAI researcher survey 2023, AI researchers (AI researchers): 70%. AI researchers saying AI safety research should be prioritized more. “70% of respondents thought AI safety research should be prioritized more than it currently is”AI researcher survey 2023, AI researchersEurobarometer 2025, public (public): 84%. Europeans: AI at work needs careful management for privacy and transparency. “84% of Europeans think that AI requires careful management to protect privacy and ensure transparency in the workplace”Eurobarometer 2025, publicPew 2025, public (public): 58%. US adults worried the government will not go far enough regulating AI. “Both experts (56%) and the public (58%) are more concerned that the U.S. government will not go far enough in regulating AI”Pew 2025, publicPew 2025, AI researchers (AI researchers): 56%. US AI experts worried the government will not go far enough regulating AI. “Both experts (56%) and the public (58%) are more concerned that the U.S. government will not go far enough in regulating AI”Pew 2025, AI researchersGallup 2025, public (public): 80%. US adults: keep AI safety rules even if AI develops more slowly. “80% of U.S. adults believe the government should maintain rules for AI safety and data security, even if it means developing AI capabilities more slowly”Gallup 2025, public

AI researchersThe publicShaded: the middle half of each group

All 6 numbers from 5 papers, in 2 charts, each with its quote

Who says what

Documented positions only, coded with cited evidence by an AI agent: 26 of 400 mapped researchers so far. Everyone else is unknown: no reviewed evidence, which is neither agreement nor neutrality.

23 agree3 it's complicated0 disagree26 of 400 coded

Agree (23)

It's complicated (3)

Economists and computer scientists

Documented positions on this claim of researchers on the co-authorship map, by field (from their OpenAlex topics). Researchers without a documented position are not counted.

FieldAgreeIt’s complicatedDisagree
Economists (241 on the map)19 of 212 of 210 of 21
Computer scientists (51 on the map)Too few to say (4 coded)

Each position with the researcher’s reasoning and cited evidence

What to read

Essentials on regulation

Read

Know

Follow

  • Yoshua Bengio agreesComputer scientist, Université de Montréal

    Computer scientist and Turing Award co-recipient with a documented position in favour of acting on AI now.

  • Gillian Hadfield agreesProfessor of AI Alignment and Governance, Johns Hopkins University

    Professor of AI alignment and governance at Johns Hopkins who works on how to design rules for AI.

  • AI Governance Lab (Rob Reich)Newsletter · 1 posts in this guide

    A newsletter on AI governance and policy from a Stanford political philosopher.

All essentials on regulation, for 10 minutes to an afternoon

Where the papers point

58 agree6 it's complicated0 disagree64 of 313 coded

How each published paper’s own finding bears on the claim, read from its title and abstract by an AI model (AI-written). 64 of the 313 published papers address this question; the rest are not counted. This codes findings, not authors.

Findings that support the claim (58)

and 55 more in the research library

Findings that it depends (6)

and 3 more in the research library

All research on this questionNews on this question