{"slug": "everyone-agrees-the-ai-race-needs-to-slow-down-but-can-anyone-do-it", "title": "Everyone Agrees the AI Race Needs to Slow Down. But Can Anyone Do It?", "summary": "Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, Anthropic alignment researcher Evan Hubinger, Elon Musk and Bill Gates have publicly agreed that frontier AI development needs to be paced, with Altman saying OpenAI will give independent evaluators employee-like access to its systems. The calls for pacing came days after Anthropic researcher Jacob Coxon resigned, arguing that AI companies are locked in a race because they do not believe competitors will act responsibly. The article examines whether such mechanisms can work given the incentives of the AI race, including a feedback loop in which better models help build better models.", "body_md": "For years, the AI race has been built around one basic assumption: whoever can build the most capable systems first wins.\n\nNow some of the people leading that race are saying it has started moving too fast.\n\nAnthropic CEO [Dario Amodei](https://darioamodei.com/post/we-must-pace-the-frontier) says the frontier needs to be paced. OpenAI CEO [Sam Altman](https://x.com/sama/status/2098811563415150910) agrees and says OpenAI will give independent evaluators employee-like access to its systems. Anthropic alignment researcher [Evan Hubinger](https://x.com/EvanHub/status/2098814911078138333) says pacing is necessary if humanity is going to survive. [Elon Musk](https://x.com/elonmusk/status/2098789109980332057) says Dario is right. [Bill Gates](https://www.facebook.com/BillGates/posts/i-agree-with-dario-and-am-glad-to-see-the-support-from-leaders-of-the-other-fron/1670992314391694/) says the world needs a global plan.\n\nThat is a remarkable amount of agreement from people who normally have every reason to keep pushing.\n\nBut there is a problem.\n\nJust days before Amodei’s proposal, Anthropic researcher [Jacob Coxon](https://firethering.com/anthropic-researcher-ai-race-extinction/) resigned after arguing that the companies building these systems are locked in a race because they don’t believe their competitors will act responsibly.\n\nSo what happens when everyone agrees the race is becoming dangerous, but nobody wants to be the one who slows down first?\n\nThat is the question behind the new push to pace frontier AI. And answering it means looking beyond what the CEOs are promising to the incentives that could make those promises difficult to keep.\n\n## **Table of Contents**\n\n## **The Problem Coxon Saw Coming**\n\nJacob Coxon had already identified the problem a few days before the new calls for pacing.\n\nWhen he resigned from Anthropic, his criticism wasn’t simply that AI was becoming dangerous. He argued that Anthropic understood the stakes but was still caught in a race because it believed other companies might keep pushing ahead.\n\nThat creates a difficult incentive.\n\nA lab can believe that slowing down would make AI development safer while also believing that slowing down alone could leave it behind. If a competitor keeps improving its models, the lab that pauses could lose talent, investment, capabilities and influence.\n\nIn that situation, telling everyone to be more careful isn’t enough.\n\nSomeone has to believe that the other players will be careful too.\n\nThat is what makes the recent statements from Amodei, Altman and others worth looking at beyond the headlines. They are not only arguing that AI safety matters but also starting to talk about mechanisms that could make slowing down possible without asking one company to take the risk alone.\n\nThe question is whether those mechanisms can work once the incentives of the AI race are taken seriously.\n\n## **The Accelerator Inside the Race**\n\nThere is another reason this race is becoming harder to manage, the technology being raced over can actually help with the race itself.\n\nAI systems are already useful for writing code, running experiments, analyzing results and helping researchers explore new approaches. As those systems become more capable, more of the work involved in developing the next generation can potentially be handed to AI.\n\nThat creates a feedback loop.\n\nA better model can help build a better model, which can then help build the one after that.\n\nThe question is what happens to the amount of time humans have to understand each new jump.\n\nImagine two labs improving their systems at roughly the same pace. If one develops an AI that becomes substantially better at AI research itself, its advantage isn’t limited to the capabilities of that model. It may also have a faster way of producing the next one.\n\nNow the competitive pressure changes.\n\nFalling behind doesn’t necessarily mean having a slightly weaker model. It could mean falling behind in the ability to improve models at all.\n\nThat helps explain why the coordination problem Coxon described is so difficult. If every lab worries that another lab might discover a faster development loop, waiting can look increasingly expensive even when everyone agrees that more caution would be safer.\n\nAnd this is where the argument for pacing becomes different from simply asking researchers to be careful.\n\nThe goal would be to make sure the systems helping accelerate AI development do not move faster than the methods humans use to understand, evaluate and control them.\n\nThat is the problem Dario Amodei is trying to address. The harder question is whether a set of agreements can actually create that breathing room when every participant has an incentive to keep its lead.\n\n## **The First Test: Can Anyone Verify What Labs Are Doing?**\n\nAmodei’s first proposal is also the easiest one to understand.\n\nAnthropic would give independent evaluators ongoing, employee-like access to its AI systems and development process. They would be able to examine safety practices, investigate incidents and assess how well the models are behaving as they are trained.\n\nOpenAI says it will do the same.\n\nThat is important because one of the big problems with AI safety commitments is verification.\n\nA company can say it is taking precautions. An outside evaluator can ask whether those precautions are actually being followed.\n\nAmodei is proposing something more intrusive than a traditional audit. The evaluators would have access comparable to employees, and they would be able to publish key findings without Anthropic controlling the conclusions.\n\nThat creates a useful mechanism for accountability.\n\nBut it doesn’t answer the hardest question yet.\n\nAn evaluator can tell you whether a lab is following its safety procedures. It cannot, by itself, make competing labs agree on how quickly they should develop more capable systems.\n\nIn other words, this can make the race more visible without necessarily making the race slower and that distinction becomes important with the next two parts of Amodei’s plan.\n\n## **The Second Test: Can Competitors Actually Agree?**\n\nThe second part of Amodei’s plan is much harder.\n\nHe wants frontier AI companies in democratic countries to coordinate on common safety standards and limits on how quickly unchecked capabilities can advance.\n\nThat sounds reasonable. If every major lab follows the same rules, no company has to worry that slowing down will simply hand its advantage to a competitor.\n\nBut agreeing on the principle is easier than agreeing on the rules.\n\nWhat exactly counts as moving too fast?\n\nShould limits be tied to the size of a training run, the amount of compute used, the capabilities a model demonstrates or how much AI is being used to develop the next generation of AI?\n\nAnd who gets to decide when a model has crossed that line?\n\nThese questions become even harder because the companies involved are competing for the same customers, researchers and investment. A system that slows everyone equally may reduce the competitive problem, but only if everyone trusts the others to follow the agreement.\n\nThere is also a legal problem.\n\nAmodei acknowledges that some forms of coordination between competing companies could raise antitrust concerns. His proposal therefore involves government support, potentially giving companies a legal framework in which they can coordinate on safety without effectively agreeing to restrict competition.\n\nThat changes the nature of the proposal.\n\nThis isn’t something Anthropic and OpenAI can simply settle between themselves. It would require companies, regulators and governments to agree on what should be limited, how those limits should be measured and how violations would be handled.\n\nThe independent evaluators address one problem, whether companies are doing what they said they would do.\n\nThis part has to solve another, whether everyone is agreeing to the same rules in the first place.\n\nAnd even if the companies manage that, there is still a much bigger player that cannot be brought into the room by a corporate agreement alone.\n\n## **The Third Test: What Happens When Countries Don’t Agree?**\n\nAmodei’s third proposal is about getting governments to coordinate globally, including with China.\n\nAnd he is unusually clear about the problem.\n\nPacing only works within the limits of the lead that the United States and its allies already have. If the U.S. slows down by more than that margin while China keeps advancing, Amodei argues that CCP-associated projects could pull ahead.\n\nThat creates a contradiction at the heart of the proposal.\n\nThe reason to slow down is to give researchers more time to make increasingly capable AI safer. But the same slowdown could reduce the technological lead that Amodei believes is important for U.S. national security.\n\nHis answer isn’t to ignore that problem. He argues that the U.S. should use export controls, stronger protection against model-weight theft and restrictions on unauthorized distillation to preserve or even widen its lead while creating room to pace development.\n\nHe believes those measures could widen America’s lead over China over the next three to five years, giving democracies more breathing room to improve AI safety.\n\nBut that still leaves the larger question: what happens when the goal is to slow the frontier everywhere?\n\nAmodei proposes several levels of global cooperation with China. The easiest would involve narrow agreements around obviously dangerous uses, such as AI-assisted biological weapons. Another possibility would be requiring models to be tested for serious cyber, biological and alignment risks before release.\n\nHe sees a harder possibility further up the ladder: an agreement limiting the rate of recursive self-improvement.\n\nA full global pause sits at the very end of that ladder and Amodei is skeptical that it could happen anytime soon. The problem is verification. If one side could secretly defect from an agreement and gain a major strategic advantage, the incentive to do so would be enormous.\n\nThat brings the race problem back in a different form.\n\nA company might worry that another company will move faster. A government might worry that another country will do the same. In both cases, slowing down becomes much easier if everyone can trust that the others are slowing down too.\n\nAnd the more powerful AI becomes, the more costly that trust becomes to get wrong.\n\n## **So, Will the AI Race Actually Slow Down?**\n\nMaybe the most revealing part of this whole debate is that the people asking for a slower race are still building the systems they want to slow down.\n\nDario Amodei isn’t calling for AI development to stop. Sam Altman isn’t either. They are arguing that the frontier needs enough breathing room for safety work to catch up with capability.\n\nWhether that is possible will depend on something much harder than getting CEOs to agree in public.\n\nThey have to make slowing down feel less dangerous than continuing to race.\n\nAnd nobody has really solved that yet.\n\nFor now, the industry has something it rarely has: broad agreement that the current pace deserves a second look.\n\nThe next question is whether that agreement survives when slowing down starts to carry a real cost.", "url": "https://wpnews.pro/news/everyone-agrees-the-ai-race-needs-to-slow-down-but-can-anyone-do-it", "canonical_source": "https://firethering.com/ai-race-slowdown/", "published_at": "2026-09-13 19:15:24+00:00", "updated_at": "2026-09-13 19:21:03.474145+00:00", "lang": "en", "topics": ["ai-safety", "ai-policy", "artificial-intelligence", "ai-research"], "entities": ["Anthropic", "Dario Amodei", "OpenAI", "Sam Altman", "Evan Hubinger", "Elon Musk", "Bill Gates", "Jacob Coxon"], "alternates": {"html": "https://wpnews.pro/news/everyone-agrees-the-ai-race-needs-to-slow-down-but-can-anyone-do-it", "markdown": "https://wpnews.pro/news/everyone-agrees-the-ai-race-needs-to-slow-down-but-can-anyone-do-it.md", "text": "https://wpnews.pro/news/everyone-agrees-the-ai-race-needs-to-slow-down-but-can-anyone-do-it.txt", "jsonld": "https://wpnews.pro/news/everyone-agrees-the-ai-race-needs-to-slow-down-but-can-anyone-do-it.jsonld"}}