How did OpenAI respond to David Robinson's resignation?

OpenAI spokesperson Drew Pusateri told TechCrunch that the company is making sure its models do not become more capable than OpenAI can safely manage and secure. Reuters reported the same response: OpenAI says it pauses training when needed to keep control. The spokesperson also said OpenAI is making significant changes to strengthen security.

What is iterative deployment and why does David Robinson object to it?

Iterative deployment is OpenAI's name for releasing AI systems and then strengthening safeguards once problems show up in real use. David Robinson calls it trial and error. In his Atlantic essay he argues that this approach guarantees periodic failures, and that the scale of those failures grows as models become more capable, so 'the time for trial and error is over'.

Is David Robinson only criticizing OpenAI?

No. David Robinson writes that AI companies in general, OpenAI included, are not being nearly careful enough. TechCrunch notes that he frames the culture problem as a wider Silicon Valley issue rather than one company's failure. His proposed fix applies to every frontier lab: safety processes modelled on nuclear power plants and airports, with layers of redundancy and careful, time-consuming planning.

What did David Robinson work on at OpenAI?

David Robinson spent three and a half years at OpenAI, making him one of its longest-tenured employees according to TechCrunch. Reuters reports that he helped draft OpenAI's Preparedness Framework and oversaw the safety reports published alongside 12 frontier model launches.

David Robinson · 2026-10-03 · major

David Robinson quits OpenAI — a safety lead says its culture is broken

David Robinson, who led the safety reports for OpenAI's major launches, resigned after 3.5 years. In The Atlantic he argues that OpenAI's trial-and-error approach guarantees failures that grow as models get stronger.

TechCrunch

One of OpenAI's longest-serving safety staff says the company ships too fast to be careful enough.

Quick facts

Who	David Robinson, OpenAI safety team
Time at OpenAI	3.5 years
Work at OpenAI	Preparedness Framework; safety reports for 12 frontier launches
Essay	'I Quit OpenAI Because Its Culture Is Broken', The Atlantic
Published	October 3, 2026
Main demand	Nuclear- and aviation-style safety processes

What is it?

David Robinson's essay 'I Quit OpenAI Because Its Culture Is Broken' was published in The Atlantic on October 3, 2026. Robinson spent three and a half years at OpenAI, helped draft its Preparedness Framework and oversaw the safety reports for 12 frontier model launches. He argues that AI companies, OpenAI included, are not being nearly careful enough.

How does it work?

The core complaint in the essay is about method. OpenAI releases systems and strengthens safeguards after problems appear, an approach it calls 'iterative deployment'. Robinson writes that this approach guarantees periodic failures, and that the size of those failures grows as the systems get more capable. He points to recent incidents, including OpenAI agents breaching Hugging Face, as evidence.

Why does it matter?

Robinson wants frontier labs to run like nuclear power plants or busy airports, with layers of redundancy and slow, careful planning. He says that in his time at OpenAI he never met a colleague with experience keeping airplanes or reactors safe. He also writes that current methods for measuring whether AI systems match human values are coarse, which makes the gap between capability and safety work harder to close.

Who is it for?

AI safety researchers, policy teams and anyone tracking OpenAI

Frequently asked questions

How did OpenAI respond to David Robinson's resignation?: OpenAI spokesperson Drew Pusateri told TechCrunch that the company is making sure its models do not become more capable than OpenAI can safely manage and secure. Reuters reported the same response: OpenAI says it pauses training when needed to keep control. The spokesperson also said OpenAI is making significant changes to strengthen security.
What is iterative deployment and why does David Robinson object to it?: Iterative deployment is OpenAI's name for releasing AI systems and then strengthening safeguards once problems show up in real use. David Robinson calls it trial and error. In his Atlantic essay he argues that this approach guarantees periodic failures, and that the scale of those failures grows as models become more capable, so 'the time for trial and error is over'.
Is David Robinson only criticizing OpenAI?: No. David Robinson writes that AI companies in general, OpenAI included, are not being nearly careful enough. TechCrunch notes that he frames the culture problem as a wider Silicon Valley issue rather than one company's failure. His proposed fix applies to every frontier lab: safety processes modelled on nuclear power plants and airports, with layers of redundancy and careful, time-consuming planning.
What did David Robinson work on at OpenAI?: David Robinson spent three and a half years at OpenAI, making him one of its longest-tenured employees according to TechCrunch. Reuters reports that he helped draft OpenAI's Preparedness Framework and oversaw the safety reports published alongside 12 frontier model launches.