David Robinson used to write the safety reports that accompanied every major model release at OpenAI. Now he has quit his position, and he is using The Atlantic to sound the alarm.
Robinson’s exit is notable because he was one of the people whose job was to assess the risks of the very systems he is now criticising. His warning comes with the weight of someone who has sat on the inside of the company’s decision-making process, and his message is blunt: the culture at OpenAI and across the broader AI industry is “broken.”
Robinson’s Warning
Robinson says that Silicon Valley has operated with “extreme confidence” and “perpetual sprints.” Companies have built bigger and better models with “unimpeded optimism,” he says, while ignoring or underestimating potential problems along the way.
The problem, he argues, runs deeper than a few new rules or regulations. It is systemic. The tech industry has moved fast and broken things, and it has done so without enough humility.
His prescription is dramatic. He says AI companies need to adopt nuclear-level safeguards:
“Given today’s risks, frontier labs need to run like nuclear power plants or busy airports, with layers of redundancy and careful, time-consuming planning, so that the occasional and inevitable human error does not open a door to disaster.”
The Departure Parade
Robinson is far from the first safety worker to leave a major AI firm. The trend has become a parade of resignations, each bringing public warnings about the dangers of the technology they helped build.
Jacob Coxon seems to have kicked off the exodus by quitting Anthropic and then publicly saying AI “could kill us all by the end of the decade.” Robert O’Callahan, Bilal Chughtai, and Josh Engels followed Coxon out of Google DeepMind. Joe Benton also left Anthropic. Each departure brought another public warning about the direction of AI development.
A Pattern Of Public Warnings
People who were once behind closed doors, writing safety reports and attending internal meetings, are now stepping forward to speak directly to the public. Their message is consistent: the industry is moving too fast, and the risks are being ignored.
Robinson’s warning is part of that same thread. He is not the first to raise alarms, and he will not be the last. The pattern is visible, though it spans only a handful of exits so far.
The Skeptic’s Case
It is understandable if you are feeling a bit cynical about everyone suddenly coming out of the woodwork to warn about how dangerous the thing they helped build is. They did, after all, make this mess. But that doesn’t mean we should discount their warnings.
What Humility Looks Like
Robinson’s call for humility is the core of his argument. He wants AI companies to stop operating with unimpeded optimism and start planning carefully. He wants them to treat the occasional human error as something that could lead to disaster.
The comparison to nuclear power plants and busy airports is telling. Those industries operate with layers of redundancy and careful planning.
Robinson wants the same approach for AI. He wants the industry to slow down, to look outside its own bubble, and to build systems that can withstand human error.
The Risks Are Real
Robinson’s warning is not abstract. He points to today’s risks and says that frontier labs need to change how they operate. The occasional human error should not open a door to disaster, he argues.
The risks are not hypothetical. The parade of departures shows that people inside the industry have stepped forward publicly to warn about them. The question is whether the industry listens.
The Bottom Line
Robinson’s resignation is significant because he was one of the people whose job was to assess the risks. His warning carries weight precisely because he was in a position to know what he was talking about.
The industry has a choice. It can continue to move fast and break things, or it can slow down and build systems that are truly safe. Robinson has made his choice. Now the rest of the industry must make theirs.
Key Facts Box
- David Robinson wrote safety reports for every major OpenAI model release
- He resigned from his position at OpenAI and published an editorial in The Atlantic
- Jacob Coxon quit Anthropic and warned AI could kill us all by the end of the decade
- Robert O’Callahan, Bilal Chughtai, and Josh Engels left Google DeepMind
- Joe Benton left Anthropic
- Robinson calls for AI companies to run like nuclear power plants or busy airports
Robinson’s warning is a call to action. The industry has been operating with extreme confidence and perpetual sprints. It is time to stop and look around. The risks are real, and the occasional human error should not open a door to disaster.
Source material: “An OpenAI safety employee has quit and is sounding the alarm,” The Verge.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.