Pausing AI at human-genius level to solve alignment is safer than either racing to superintelligence or halting entirely

ai-regulation ethics future ai-safety ai-governance ai-policy

Share this with your representatives, journalists and friends

Arguments for

Controllable pause point

Proponents argue for scaling only up to a capability level around average human, top human expert, top human genius, or max-controllable AI, and then slowing sharply or pausing. They present that zone as the last stage where humans can still maintain control, and some suggest capability growth may plateau there rather than immediately racing onward.

In the Plan A scenario, why does the Consortium choose to steadily increase capabilities until reaching top-expert-dominating AI (TED-AI) in 2035, then slow down nearly to a pause until handing off to these AIs in 2040? [...] Scale until you’re close to max-controllable-AI, i.e. …

Eli Lifland · Jul 10, 2026 · full quote
Unlike all other actors we model, the consortium is not racing to build better AIs as fast as they possibly can. The specific strategies are discussed in more detail in the Capabilities Scaling Strategy supplement, but at a high level, the strategy is to scale smoothly to Top-Exp…

Brendan Halstead · Jul 10, 2026 · full quote
Between 2030 and 2035, we scale within the human range, to AIs that are roughly as capable as top human experts... In 2035, we pause at top-human-expert level AI in order to maintain human control. In 2040, we unpause and scale to superintelligence.

AI Futures Project · Jul 9, 2026 · full quote
First let’s talk about pause at human level — that’s a phrase from Ryan Greenblatt. So that’s when we’re at the point of time of AI engaging in AI R&D, at this point of time when things perhaps go even faster, should we at that point be trying to slow things down, even pause, sto…

Will MacAskill · Apr 22, 2026 · full quote
If the foregoing arguments are right, then scenario three (risk apprehension, self-improvement, alignment) is the most likely of the six. This suggests a concrete near-term prediction: Contrary to standard arguments, when AIs achieve average human-level capabilities, capabilities…

Peter N. Salib · May 16, 2023 · full quote

AI helps solve alignment

Supporters say human-level or near-human AI could provide a major boost for alignment and related protective work, creating a window in which AI helps make later systems safer. The pause is meant to redirect AI labor from further AI R&D toward alignment, philosophy, safety, and other major problems before any intelligence explosion.

Then, in the mid-2030s, they pause at AIs around the level of top human geniuses. These AIs are close enough to current AIs that the same alignment techniques that mostly-sort-of-work for ours might mostly-sort-of-work for them too... The plan is to spend the next ~10 years using…

Scott Alexander · Jul 9, 2026 · full quote
If our ability to make increasingly capable AI systems safe can’t benefit from such labor in a comparable way, then at least for relatively hard problem profiles, and absent large amounts of capability restraint, I think we’re likely headed for disaster. I also discuss the timing…

Joe Carlsmith · Mar 14, 2025 · full quote
I think that what we want to do as a society, if we gain confidence that we’re at the starting point of this intelligence explosion, is to redirect as much of that AI labour as we can from further AI R&D to things that could help protect us from future generations of AIs — both i…

Ajeya Cotra · Oct 20, 2025 · full quote
An alternative approach: Don't pause yet. First develop human-level AI that can help us solve the world's major problems. Don't develop superintelligence until we're on stable ground philosophically, but still take advantage of the productivity boost that AI provides. [...] This …

Michael Dickens · Nov 21, 2025 · full quote
I do think that pausing further capabilities once we have human-ish-level AIs for even just a few years while we focus on safety would massively improve the situation. This currently seems unlikely to happen.

Ryan Greenblatt · 2025 · full quote

Governed temporary slowdown

Another supporting line is that society should have the option to slow or temporarily pause frontier development when risk benchmarks are hit, resuming only after safety, alignment, security, or public-legitimacy conditions are met. On this view, a credible pause requires governance systems, societal structures, public evidence, and clear conditions for when development can continue.

We believe it would be good for the world to have the option to slow or temporarily pause frontier AI development to enable societal structures and alignment research to keep up with the advance of the technology. The Anthropic Institute will conduct research—in collaboration wit…

Anthropic Institute · Jun 2026 · full quote
The RSP aims to make it consistently the case that our model training and deployment meets a high, clearly-specified bar for safety and that there is publicly accessible evidence that we have met this bar. Roughly speaking, we run tests (‘frontier risk evaluations’) meant to asse…

Sam Bowman · Sep 3, 2024 · full quote
AI labs put out RSP commitments to stop scaling when particular capabilities benchmarks are hit, resuming only when they are able to hit particular safety/alignment/security targets. [...] For later capabilities levels, however, it is explicit in all RSPs that we do not yet know …

Evan Hubinger · Oct 14, 2023 · full quote
We call for a prohibition on the development of superintelligence, not lifted before there is (1) broad scientific consensus that it will be done safely and controllably, and (2) strong public buy-in.

Statement on Superintelligence · Sep 22, 2025 · full quote
Therefore, we call on all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4. This pause should be public and verifiable, and include all key actors. If such a pause cannot be enacted quickly, governments should step in and inst…

Pause Giant AI Experiments: An Open Letter · Mar 22, 2023 · full quote

Arguments against

Indefinite halt instead

Some critics argue that the default should be pausing frontier AI, with the possibility of never unpausing, and call for an indefinite worldwide moratorium on major training runs and large GPU clusters. On this view, a temporary pause partway through development is not a strong enough response.

The default should be pausing, and we should have the possibility of never unpausing from frontier AI development.

Holly Elmore · Feb 6, 2026 · full quote
The moratorium on new large training runs needs to be indefinite and worldwide. There can be no exceptions, including for governments or militaries. [...] Shut down all the large GPU clusters (the large computer farms where the most powerful AIs are refined). Shut down all the la…

Eliezer Yudkowsky · Apr 7, 2023 · full quote

Slowdown loses the race

Another objection is that slowing down while others continue would impose major costs in a competitive environment and may be impractical, unenforceable, or easily circumvented. These quotes argue it is better to remain at the forefront, both because AI is treated as strategically important and because frontier knowledge can help identify and mitigate risks.

I would also say the development of this technology is a national security priority, and it seems really hard to... it seems like a really big cost to impose something where we would self-consciously slow ourselves down when others are not doing that. [...] I think it’s very poss…

Dean W. Ball · Nov 21, 2025 · full quote
Regardless of what policies the U.S. adopts, the technology of large AI systems is going to continue to advance at a global level. It is far better for the U.S. to remain at the forefront of AI—advancing the state of the art, and using that knowledge to better identify and mitiga…

John Villasenor · Apr 11, 2023 · full quote
But the idea of asking AI labs to slow down just doesn't seem practical or implementable to me. Especially in this frankly competitive business environment with labs and countries trying to build advanced technologies and creating a lot of value. And the only thing worse than tha…

Andrew Ng · Apr 9, 2023 · full quote

Pauses can backfire

Opponents also argue that delaying research serves little purpose beyond reducing knowledge, that products rather than research should be regulated, and that advocates of a moratorium bear the burden of proof because a pause could itself raise catastrophic risk. Some extend this to rejecting regulatory barriers altogether and favoring aggressive development and open competition.

Should we lobby governments to impose a moratorium on AI research? Since we don’t enforce pauses on most new technologies, I hope the reader will grant that the burden of proof is on those who advocate for such a moratorium. We should only advocate for such heavy-handed governmen…

Nora Belrose · Oct 23, 2023 · full quote
Well, my first reaction to this is that calling for a delay in research and development smacks me of a new wave of obscurantism, essentially, like why slow down the progress of knowledge and science? Then there is the question of products, like, you know, I'm all for regulating p…

Yann LeCun · Apr 9, 2023 · full quote
Big AI companies should be allowed to build AI as fast and aggressively as they can [...] Startup AI companies should be allowed to build AI as fast and aggressively as they can. [...] Open source AI should be allowed to freely proliferate and compete with both big AI companies a…

Marc Andreessen · Jun 6, 2023 · full quote

Insights

Supporters propose different trigger points for slowing down, ranging from average human-level capability to top human experts, top human geniuses, or max-controllable AI.
Many pro-pause quotes explicitly distinguish their approach from a blanket stop by envisioning later unpausing, handoff, or continued development outside the most dangerous frontier race.
Several supportive quotes treat the pause as active reallocation rather than inactivity, with advanced AI labor redirected toward alignment, safety, philosophy, and broader societal preparation.
Opposition splits between people who think slowing is impractical or strategically harmful and people who prefer a much stronger indefinite moratorium.
A recurring divide is between fixed capability-level pause plans and pause plans triggered by safety benchmarks, consensus, public buy-in, or verifiable enforcement.

Supporters mainly describe a temporary stop after reaching roughly human-level or top-expert capability, using the interval to preserve human control and turn advanced AI toward alignment and related safety work. Opponents either argue that slowing would be impractical, strategically costly, or harmful to research, or else argue that anything short of an indefinite halt is too weak. The debate therefore centers on feasibility, on how much useful safety work near-human AI can do, and on whether a limited pause is the right point between racing ahead and stopping altogether.

Cast your vote:

Results (33 votes):

Total (33 votes)

For 23 (70%) Abstain 0 (0%) Against 10 (30%)

For (22)

Brendan Halstead
AI Futures Project researcher writing on AI takeoff modeling and Plan A capability trajectories.

I am this person! Add a sourced quote Report comment

votes For and says:

Unlike all other actors we model, the consortium is not racing to build better AIs as fast as they possibly can. The specific strategies are discussed in more detail in the Capabilities Scaling Strategy supplement, but at a high level, the strategy i...
more AI Verified source (Jul 10, 2026)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Eli Lifland
AI Futures Project researcher writing on AI capability scaling, controlled takeoff, and comparative AI governance plans.

I am this person! Add a sourced quote Report comment

votes For and says:

In the Plan A scenario, why does the Consortium choose to steadily increase capabilities until reaching top-expert-dominating AI (TED-AI) in 2035, then slow down nearly to a pause until handing off to these AIs in 2040? [...] Scale until you’re close...
more AI Verified source (Jul 10, 2026)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Daniel Kokotajlo
AI Futures Project founder

I am this person! Add a sourced quote Report comment

votes For and says:

If you delay the advent of superintelligence, then that gives society more time to prepare and more time to solve the various problems that it represents.
AI Verified source (Jul 9, 2026)

added 11d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
AI Futures Project
Nonprofit AI forecasting research group behind AI 2027 and Plan A; led by Daniel Kokotajlo.

I am this person! Add a sourced quote Report comment

votes For and says:

Between 2030 and 2035, we scale within the human range, to AIs that are roughly as capable as top human experts... In 2035, we pause at top-human-expert level AI in order to maintain human control. In 2040, we unpause and scale to superintelligence.
AI Verified source (Jul 9, 2026)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Scott Alexander
Author and psychiatrist

I am this person! Add a sourced quote Report comment

votes For and says:

Then, in the mid-2030s, they pause at AIs around the level of top human geniuses. These AIs are close enough to current AIs that the same alignment techniques that mostly-sort-of-work for ours might mostly-sort-of-work for them too... The plan is to ...
more AI Verified source (Jul 9, 2026)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
OpenAI
AI research organization

I am this person! Add a sourced quote Report comment

votes For and says:

One goal of such an organization should be to make it possible for the world to take coordinated action, including slowing frontier development when needed, so societal resilience, safety, and alignment can keep pace.
AI Verified source (Jun 8, 2026)

added 10d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Anthony Aguirre
Physicist; Future of Life cofounder

I am this person! Add a sourced quote Report comment

votes For and says:

We are approaching a runaway to superintelligence that could threaten our shared human future. Both publicly and privately, AI companies are recognizing that a pause or slowdown in certain developmental pathways is crucial to protect lives and liveli...
more AI Verified source (Jun 8, 2026)

added 12d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Anthropic
AI safety research company

I am this person! Add a sourced quote Report comment

votes For and says:

We believe it would be good for the world to have the option to slow or temporarily pause frontier AI development to enable societal structures and alignment research to keep up with the advance of the technology.
AI Verified source (Jun 4, 2026)

added 10d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Anthropic Institute
Anthropic policy and research institute studying AI capabilities, governance, and societal impacts.

I am this person! Add a sourced quote Report comment

votes For and says:

We believe it would be good for the world to have the option to slow or temporarily pause frontier AI development to enable societal structures and alignment research to keep up with the advance of the technology. The Anthropic Institute will conduct...
more AI Verified source (Jun 2026)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Will MacAskill
Philosopher; effective altruism cofounder; researcher at Forethought Centre for AI Strategy

I am this person! Add a sourced quote Report comment

votes For and says:

First let’s talk about pause at human level — that’s a phrase from Ryan Greenblatt. So that’s when we’re at the point of time of AI engaging in AI R&D, at this point of time when things perhaps go even faster, should we at that point be trying to slo...
more AI Verified source (Apr 22, 2026)

added 12d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Nick Bostrom
Philosopher; 'Superintelligence' author; FHI founder

I am this person! Add a sourced quote Report comment

votes For and says:

For many parameter settings, the optimal strategy would involve moving quickly to AGI capability, then pausing briefly before full deployment: swift to harbor, slow to berth.
AI Verified source (Feb 12, 2026)

added 10d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Wei Dai
Computer engineer, cryptographer, and AI alignment writer known for Crypto++ and b-money

I am this person! Add a sourced quote Report comment

votes For and says:

If AIs became strategically competent enough, they may realize that RSI is too dangerous because they're not good enough at alignment or philosophy or strategy, and potentially convince, help, or work with humans to implement an AI pause. This presen...
more AI Verified source (Feb 3, 2026)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
David Krueger
Cambridge faculty - AI alignment, DL, and existential safety. Formerly Mila, FHI, DeepMind, ElementAI, AISI.

I am this person! Add a sourced quote Report comment

votes For and says:

None of these approaches that you'll often hear mentioned—interpretability, testing or alignment—is technically adequate. We don't know how to build superintelligent AI safely. The plan is basically to roll the dice.
AI Verified source (Feb 2, 2026)

added 11d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Michael Dickens
Writer at Philosophical Multicore on explicit Bayesian reasoning and decision-making

I am this person! Add a sourced quote Report comment

votes For and says:

An alternative approach: Don't pause yet. First develop human-level AI that can help us solve the world's major problems. Don't develop superintelligence until we're on stable ground philosophically, but still take advantage of the productivity boost...
more AI Verified source (Nov 21, 2025)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Ajeya Cotra
AI safety researcher; senior research analyst at Open Philanthropy

I am this person! Add a sourced quote Report comment

votes For and says:

I think that what we want to do as a society, if we gain confidence that we’re at the starting point of this intelligence explosion, is to redirect as much of that AI labour as we can from further AI R&D to things that could help protect us from futu...
more AI Verified source (Oct 20, 2025)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Statement on Superintelligence
2025 public statement hosted by the Future of Life Institute calling for restrictions on superintelligence development.

I am this person! Add a sourced quote Report comment

votes For and says:

We call for a prohibition on the development of superintelligence, not lifted before there is (1) broad scientific consensus that it will be done safely and controllably, and (2) strong public buy-in.
AI Verified source (Sep 22, 2025)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Joe Carlsmith
Senior research analyst at Open Philanthropy; writes on AI safety, alignment, and existential risk from power-seeking AI

I am this person! Add a sourced quote Report comment

votes For and says:

If our ability to make increasingly capable AI systems safe can’t benefit from such labor in a comparable way, then at least for relatively hard problem profiles, and absent large amounts of capability restraint, I think we’re likely headed for disas...
more AI Verified source (Mar 14, 2025)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Ryan Greenblatt
Chief Scientist at Redwood Research; AI safety researcher; lead author of "Alignment faking in large language models"

I am this person! Add a sourced quote Report comment

votes For and says:

I do think that pausing further capabilities once we have human-ish-level AIs for even just a few years while we focus on safety would massively improve the situation. This currently seems unlikely to happen.
AI Verified source (2025)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Sam Bowman
AI alignment researcher at Anthropic; on leave from NYU

I am this person! Add a sourced quote Report comment

votes For and says:

The RSP aims to make it consistently the case that our model training and deployment meets a high, clearly-specified bar for safety and that there is publicly accessible evidence that we have met this bar. Roughly speaking, we run tests (‘frontier ri...
more AI Verified source (Sep 3, 2024)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Evan Hubinger
Research scientist at Anthropic working on model organisms of misalignment; former Research Fellow at MIRI.

I am this person! Add a sourced quote Report comment

votes For and says:

AI labs put out RSP commitments to stop scaling when particular capabilities benchmarks are hit, resuming only when they are able to hit particular safety/alignment/security targets. [...] For later capabilities levels, however, it is explicit in all...
more AI Verified source (Oct 14, 2023)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Peter N. Salib
Assistant Professor of Law at the University of Houston Law Center researching constitutional law, economics, and artificial intelligence.

I am this person! Add a sourced quote Report comment

votes For and says:

If the foregoing arguments are right, then scenario three (risk apprehension, self-improvement, alignment) is the most likely of the six. This suggests a concrete near-term prediction: Contrary to standard arguments, when AIs achieve average human-le...
more AI Verified source (May 16, 2023)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Pause Giant AI Experiments: An Open Letter
2023 Future of Life Institute open letter calling for a pause on AI training runs more powerful than GPT-4.

I am this person! Add a sourced quote Report comment

votes For and says:

Therefore, we call on all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4. This pause should be public and verifiable, and include all key actors. If such a pause cannot be enacted quickly, gover...
more AI Verified source (Mar 22, 2023)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate

Abstain (0)

Against (9)

Andrew Charlton
Australian Assistant Minister for Science, Technology and the Digital Economy

I am this person! Add a sourced quote Report comment

votes Against and says:

Because the point of AI safety is not to slow the future down. It is to make sure the future remains human, aligned with our values, and advancing our interests.
AI Verified source (Jul 7, 2026)

added 11d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Holly Elmore
Founder and Executive Director of PauseAI US; evolutionary biologist and AI moratorium advocate

I am this person! Add a sourced quote Report comment

votes Against and says:

The default should be pausing, and we should have the possibility of never unpausing from frontier AI development.
AI Verified source (Feb 6, 2026)

added 12d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Dean W. Ball
Author of Hyperdimensional

I am this person! Add a sourced quote Report comment

votes Against and says:

I would also say the development of this technology is a national security priority, and it seems really hard to... it seems like a really big cost to impose something where we would self-consciously slow ourselves down when others are not doing that...
more AI Verified source (Nov 21, 2025)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Nora Belrose
AI alignment and interpretability researcher; former Head of Interpretability at EleutherAI and AI pause critic.

I am this person! Add a sourced quote Report comment

votes Against and says:

Should we lobby governments to impose a moratorium on AI research? Since we don’t enforce pauses on most new technologies, I hope the reader will grant that the burden of proof is on those who advocate for such a moratorium. We should only advocate f...
more AI Verified source (Oct 23, 2023)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Marc Andreessen
General Partner at a16z (VC), co-founder of Netscape

I am this person! Add a sourced quote Report comment

votes Against and says:

Big AI companies should be allowed to build AI as fast and aggressively as they can [...] Startup AI companies should be allowed to build AI as fast and aggressively as they can. [...] Open source AI should be allowed to freely proliferate and compet...
more AI Verified source (Jun 6, 2023)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
John Villasenor
UCLA professor; Brookings senior fellow

I am this person! Add a sourced quote Report comment

votes Against and says:

Regardless of what policies the U.S. adopts, the technology of large AI systems is going to continue to advance at a global level. It is far better for the U.S. to remain at the forefront of AI—advancing the state of the art, and using that knowledge...
more AI Verified source (Apr 11, 2023)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Yann LeCun
Computer scientist, AI researcher

I am this person! Add a sourced quote Report comment

votes Against and says:

Well, my first reaction to this is that calling for a delay in research and development smacks me of a new wave of obscurantism, essentially, like why slow down the progress of knowledge and science? Then there is the question of products, like, you ...
more AI Verified source (Apr 9, 2023)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Andrew Ng
Baidu; Stanford CS faculty; founded Coursera and Google Brain

I am this person! Add a sourced quote Report comment

votes Against and says:

But the idea of asking AI labs to slow down just doesn't seem practical or implementable to me. Especially in this frankly competitive business environment with labs and countries trying to build advanced technologies and creating a lot of value. And...
more AI Verified source (Apr 9, 2023)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate
Eliezer Yudkowsky
AI researcher and writer

I am this person! Add a sourced quote Report comment

votes Against and says:

The moratorium on new large training runs needs to be indefinite and worldwide. There can be no exceptions, including for governments or militaries. [...] Shut down all the large GPU clusters (the large computer farms where the most powerful AIs are ...
more AI Verified source (Apr 7, 2023)

added 20d ago

Choose a list of delegates

to vote as the majority of them.

Unless you vote directly.

Delegate

More h/ai-safety votes

Mandate 72-hour reporting of critical AI safety incidents to a national authority

If leading AI labs adopted a windfall clause (pre-committing to broadly share profits above a high threshold), it would meaningfully reduce competitive pressure to cut safety corners

AI poses an existential threat to humanity

AI value alignment must be subject to public deliberation, not decided by developers alone

The U.S. and China should establish Mutually Assured Compute Destruction

Back