Former OpenAI Safety Staffer Explains Resignation: "The Time for Trial and Error Is Over"

What Happened

In 3.5 years at OpenAI, Robinson helped draft the company's Preparedness Framework and oversaw the safety reports for 12 frontier-model launches. He wrote in the essay: "As the company sprints from one launch to the next, it is failing to achieve the level of care that I believe is needed."

An OpenAI spokesperson told Reuters: "We're making sure our models don't become more capable than we can safely manage and secure, and we pause training or hold back models when we need to slow down." Related departures: in early September, Anthropic researcher Jacob Coxon resigned, saying the people building AI believe it "could kill us all by the end of the decade"; and on Oct 1, the Journal reported that OpenAI had parted ways with three safety-team employees after an internal investigation found they had leaked sensitive information to an outside model-evaluation organization.

Key Facts

1. The resignation letter: David Robinson published "I Quit OpenAI Because Its Culture Is Broken" in The Atlantic, criticizing OpenAI's heavy reliance on "iterative deployment" — shipping systems first and hardening safeguards only when problems emerge.
2. The core argument: "The time for trial and error is over." Frontier AI systems need safeguards modeled on nuclear power and aviation, not internet-style rapid iteration; AI capabilities are advancing faster than researchers' understanding of alignment.
3. The credentials: Robinson spent 3.5 years at OpenAI, helped draft the company's Preparedness Framework, and oversaw safety reports for 12 frontier-model launches.
4. The official response: An OpenAI spokesperson said the company keeps model capabilities within what it can "safely manage and secure," pausing training or holding back models "when we need to slow down." Context: in late September, the Journal reported that OpenAI had shelved its next-generation model GPT-6.1 Astra, originally planned for an October release, after it fell short in internal safety tests.

Context

Over the past month, the AI safety field has also seen the following: Anthropic CEO Dario Amodei publicly called for the industry to slow down frontier models, with OpenAI's Sam Altman and Elon Musk endorsing the view; more than 20 researchers from Anthropic, OpenAI, Meta and Microsoft co-authored a paper warning that AI could build the next generation of AI on its own, triggering an "intelligence explosion" humans can't keep up with.