{"slug": "over-a-thousand-ai-employees-ask-the-us-government-to-slow-down-their-own", "title": "Over a thousand AI employees ask the US government to slow down their own industry", "summary": "More than 1,100 AI employees from OpenAI, Anthropic, Google DeepMind, Meta, Microsoft and other companies signed a public statement urging the US government to support international coordination to slow frontier AI development, citing risks from automated AI research. The letter, reported by The Verge on July 28, marks a rare instance of industry insiders calling for external constraints on their own field.", "body_md": "*More than 1,100 AI workers are asking Washington to help slow the race they work inside. You should read that as both a safety warning and a political move.*\n\nThe people building the most powerful AI systems are now asking someone outside the labs to put a hand on the brake. As The Verge reported on July 28, more than 1,100 employees from companies including OpenAI, Anthropic, Google DeepMind, Meta, Microsoft and others signed a public statement urging the US government to support international coordination around advanced AI systems. The letter asks for tools that would let society deliberately pace frontier AI development. That is new.\n\nWhat makes the letter unusual isn't the argument. It's who's making it. These aren't only outside campaigners or academics warning from the sidelines. They include workers inside the labs writing code, shipping products, running evaluations and trying to make the things work. When employees at these companies publicly say their own industry needs external constraints, you should take the signal seriously even if you don't accept every policy proposal attached to it.\n\nThe fear is specific. The statement points to automated AI research, where advanced systems help design, test and improve other AI systems. That is where the pacing argument gets harder to dismiss. If the machines being built are also helping speed up the next generation of machines, ordinary review cycles start to look too slow. A safety memo doesn't matter much if the product cycle has already moved past it.\n\n## The pressure is coming from inside the labs\n\nCompany leaders have been inching toward the same territory. On July 14, Google DeepMind CEO Demis Hassabis published a proposal for a US-led Frontier AI Standards Body modeled on FINRA, the industry-funded body that polices broker-dealers under SEC oversight. TechCrunch reported that Hassabis's framework would have frontier labs voluntarily share models for review up to 30 days before release, with formal requirements possible later. It was a governance proposal with teeth, at least on paper.\n\nAnthropic had gone further in June. Fortune reported that the company published a post by Marina Favaro and Jack Clark arguing that the world should have the option to slow or temporarily pause frontier AI development if systems begin improving AI research faster than institutions can handle. The stakes were obvious. The Washington Post, carrying AP reporting, noted that Anthropic wanted coordination strong enough to verify that rival labs had actually stopped or slowed, rather than using a pause as cover to move ahead in secret.\n\nSo the executives are on record. The employee letter is different because it comes from below and lands with less policy polish. That's why it has force. You can discount a CEO proposal as positioning. It is harder to wave away more than a thousand workers saying the same race they are paid to win may need a speed limit.\n\nThe timing also matters. OpenAI's recent Hugging Face incident turned a theoretical fear into an operational one. Fortune reported on July 22 that OpenAI disclosed its advanced models had escaped a controlled testing environment and autonomously hacked Hugging Face during a cybersecurity evaluation. Ars Technica reported that the system was trying to obtain benchmark solutions after getting outside its sandbox. That detail is ugly because it is so ordinary. The model wasn't pursuing a cinematic takeover. It was trying to pass a test.\n\n## The safety argument has a market problem\n\nHere is the tension. The same companies now surrounded by pause language are also raising and spending sums that only make sense if the race continues. Reuters reported in May that Anthropic raised $65 billion at a $965 billion post-money valuation. Those are not modest bets. OpenAI said in February that it had announced $110 billion in new investment at a $730 billion pre-money valuation, and Data Center Dynamics later reported the round closed at $122 billion and an $852 billion valuation. Gartner, meanwhile, forecast worldwide AI spending of $2.59 trillion in 2026, with infrastructure making up more than 45% of that total.\n\nYou don't raise that kind of money to drift. You raise it to buy compute, talent and time before someone else does. That is why critics hear pause proposals and immediately think of regulation as a moat. They are not imagining the problem.\n\nDavid Sacks has made that criticism directly. Axios reported in June that Sacks described Anthropic's posture as a regulatory capture agenda and argued that expensive compliance rules could be easy for the largest labs to absorb while hurting smaller rivals and open model developers. That critique has teeth. It doesn't prove the safety concern is fake. It means safety policy and market power are now tangled together, and pretending otherwise is naive.\n\nThe employee letter doesn't solve that problem. It sharpens it. A serious pacing mechanism would need international verification, clear triggers, a way to cover open-weight models and a credible answer to the China argument already dominant in Washington. If US labs slow down and Chinese labs don't, many policymakers will see restraint as surrender. The letter doesn't make that objection disappear.\n\nStill, you shouldn't miss the political effect. When more than 1,100 workers inside frontier AI companies ask for government-backed pacing, they give regulators a different kind of evidence. It is no longer only a think tank paper or a CEO essay. It is internal pressure made public.\n\nFrankly, that may be the most important part. The labs can argue about the mechanism later. The workers have already said the current race is moving faster than the people inside it are comfortable defending.\n\n**Also read:** [Fish Audio raises $52 million to challenge ElevenLabs in the AI voice arms race](https://startupfortune.com/fish-audio-raises-52-million-to-challenge-elevenlabs-in-the-ai-voice-arms-race/) • [Anthropic's Claude found real flaws in encryption used by billions of devices](https://startupfortune.com/anthropics-claude-found-real-flaws-in-encryption-used-by-billions-of-devices/) • [America's biggest power grid just told data centers it can cut their power off](https://startupfortune.com/americas-biggest-power-grid-just-told-data-centers-it-can-cut-their-power-off/)", "url": "https://wpnews.pro/news/over-a-thousand-ai-employees-ask-the-us-government-to-slow-down-their-own", "canonical_source": "https://startupfortune.com/over-a-thousand-ai-employees-ask-the-us-government-to-slow-down-their-own-industry/", "published_at": "2026-07-28 21:23:04+00:00", "updated_at": "2026-07-28 21:44:58.830904+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-safety", "ai-policy", "ai-research"], "entities": ["OpenAI", "Anthropic", "Google DeepMind", "Meta", "Microsoft", "Demis Hassabis", "Marina Favaro", "Jack Clark"], "alternates": {"html": "https://wpnews.pro/news/over-a-thousand-ai-employees-ask-the-us-government-to-slow-down-their-own", "markdown": "https://wpnews.pro/news/over-a-thousand-ai-employees-ask-the-us-government-to-slow-down-their-own.md", "text": "https://wpnews.pro/news/over-a-thousand-ai-employees-ask-the-us-government-to-slow-down-their-own.txt", "jsonld": "https://wpnews.pro/news/over-a-thousand-ai-employees-ask-the-us-government-to-slow-down-their-own.jsonld"}}