David Robinson, who worked on safety research at OpenAI, announced his departure in early October 2026. In an internal essay he described the company’s internal environment as fundamentally flawed, arguing that the problem extends beyond technical safeguards to the organization’s overall ethos. His resignation adds to a growing list of safety-focused staff leaving top AI firms.
Robinson’s concerns echo those voiced previously by Jacob Coxon, a researcher who left both OpenAI and Anthropic after warning that the firms were effectively gambling with public safety. In response, Anthropic’s chief executive Dario Amodei announced a new initiative aimed at slowing development to prioritize caution, signaling a rare public acknowledgment of the risks.
The debate over safety escalated this week when senior AI leaders convened with President Donald Trump and signed a loosely drafted, non-binding pledge to adopt additional safety controls. Observers noted the rapid timing of the agreement and questioned whether the pledge would translate into enforceable standards, highlighting the tension between voluntary industry commitments and formal regulation.
Robinson argues that focusing solely on new regulations or isolated rules will not address the deeper cultural issues he identifies at OpenAI and across Silicon Valley. He points to recent reports that former CEO Sam Altman has lost the confidence of several colleagues, suggesting that leadership trust deficits may be symptomatic of a broader systemic problem.
OpenAI’s spokesperson Drew Pusateri responded by saying the organization is continuously enhancing its safety protocols. He warned that allowing increasingly powerful models to proliferate without resolving existing safety gaps makes the overall situation more perilous, underscoring the company’s claim that it remains committed to mitigating risk.
The resignation and accompanying commentary were reported by TechCrunch’s weekend editor Anthony Ha, whose byline notes a background in technology journalism and venture-capital media. The article includes a standard disclosure that affiliate links generate a small commission, though the outlet asserts this does not affect editorial independence.