OpenAI Safety Employee Resigns, Says Company Culture Is 'Broken'

David Robinson, a senior safety researcher who wrote the safety reports accompanying OpenAI's major product launches, has resigned and publicly stated the company's culture is "broken." His departure adds pressure on OpenAI at a moment when the company is pushing out new products faster than at any point in its history.

Robinson's exit follows that of Jacob Coxon, another researcher who quit both [OpenAI and](/government/pentagon-openai-google-replace-claude-classified) Anthropic earlier this year, calling AI labs collectively "reckless." Together, the departures form a visible pattern of safety-focused employees leaving the two most prominent AI labs in the world.

What Robinson Said

Robinson led the team responsible for writing system cards — the documents OpenAI publishes alongside major model launches that describe a model's known risks, safety evaluations, and mitigation measures. These are not internal memos; they are public-facing accountability documents read by regulators, researchers, and journalists worldwide.

His characterization of the company's culture as "broken" goes beyond a standard resignation disagreement. It targets the internal operating environment — how safety concerns are raised, heard, and acted on. The specifics of what he experienced have not been fully disclosed publicly, but the framing is deliberate and serious.

Context: A Pattern, Not an Anomaly

This is not the first high-profile safety departure from OpenAI. In 2024, several members of the company's superalignment team — including Ilya Sutskever and Jan Leike — left under similar circumstances. Leike wrote publicly that safety culture had taken a back seat to product development. OpenAI disputed that characterization at the time.

According to TechCrunch, Coxon's departure from Anthropic — a company founded explicitly on safety principles — suggests the criticism extends beyond OpenAI alone. That detail matters: it's harder to dismiss as company-specific dysfunction when the researcher quit both of the industry's leading labs.

OpenAI has not commented publicly on Robinson's resignation as of this writing.

Why It Matters

For OpenAI: Robinson was the author of the documents OpenAI uses to demonstrate responsible deployment. Losing the person who writes those documents — and having him publicly challenge the company's safety culture — undermines the credibility of future system cards. It raises a direct question: who is writing these reports now, and what culture are they operating inside?

For regulators: The EU AI Act requires high-risk AI systems to maintain documented risk management processes. U.S. regulators, including the FTC and NIST, have pointed to voluntary safety commitments from major AI labs as evidence that binding regulation can wait. Resignations like Robinson's are the kind of signal that accelerates those conversations.

For enterprises: Companies that have deployed OpenAI products based partly on published safety evaluations now have reason to ask harder questions about how those evaluations are produced and who is accountable for them.

What to Watch

OpenAI has not announced a replacement for Robinson's role. Watch for whether the company publishes system cards for upcoming model releases and whether those documents maintain the same level of detail Robinson was known for. Any change in the depth or format of future safety reports would be a meaningful signal.

The broader question — whether the pace of AI product deployment has outrun the internal capacity to evaluate risks — is now squarely back in public debate.