Over 1,300 AI researchers say the industry's race to build more powerful systems is putting humanity at risk
A letter signed by scientists and engineers at OpenAI, Anthropic and Google DeepMind pushes back hard against the idea that AI insiders aren't worried. They are.
Key points
- A letter signed by 1,367 AI researchers and engineers, mostly from OpenAI, Anthropic and Google DeepMind, warns that the current pace of AI development carries serious risks to humanity.
- Signatories work on frontier AI (the most powerful and advanced systems being built today), making this one of the most direct insider warnings yet.
- The letter contradicts the common claim that only outsiders and critics worry about AI catastrophe.
- No specific demand or policy proposal was attached, but its existence signals growing unease inside the labs themselves.
For years, a reassuring line has circulated in tech circles: the people actually building AI aren't scared. Anyone raising alarms either doesn't understand the technology or is pushing a fringe view rooted in science fiction.
A newly published open letter blows a large hole in that argument.
Who signed it, and why does that matter?
1,367 researchers and engineers at frontier AI labs put their names to the letter. The bulk come from two organisations you will recognise immediately, OpenAI and Anthropic, with a substantial contingent from Google DeepMind. Stuart Russell, a prominent AI researcher, wrote about the letter's implications in The Guardian.
These are not outside critics. Their professional lives are spent making AI more capable. They wrote the code. They ran the experiments. And they are worried enough to say so publicly, which researchers at major labs almost never do.
What are they actually concerned about?
Catastrophic risk: the possibility that advanced AI systems could cause large-scale, hard-to-reverse harm. The signatories describe the current moment as dangerous.
This is a direct challenge to a pattern common in public debate. Critics of AI safety concerns are regularly dismissed as "doomers" or "Luddites" pushing a "fringe" view. The letter says, plainly: that characterisation is wrong. AI2Day has tracked this tension across 16 stories tagged ai-risk since 14 July 2026, and the insider dissent is getting louder, not quieter.
What does this mean for ordinary people?
When more than a thousand of the people building these systems say the risks are real, that is worth paying attention to. It does not guarantee catastrophe. What it does puncture is the convenient line that experts aren't worried: they are, and now they have signed their names to it.
My read: the letter matters less for what it demands, which is nothing specific, and more for what it ends. The "don't worry, the builders aren't scared" talking point is no longer available.
What happens next?
Without a list of demands or a proposed policy fix, the letter's immediate policy impact is limited. It functions as a signal, a visible crack in the confident public face the big labs have maintained.
Pressure for government oversight has been building on both sides of the Atlantic. This letter adds weight to those calls. Compare it with the July 2026 push by over 1,100 AI employees asking the US government to slow the race before it outruns safety: the number of signatories has grown, and the language has hardened. Regulators should notice that trend.
