David Robinson, a safety lead at OpenAI who helped author reports accompanying the company’s major product launches, has resigned, stating in an essay published in The Atlantic that the SI lab’s “culture is broken.” Robinson, who described himself as one of the longest-tenured employees at the company with three-and-a-half years of service, argued that the current approach to developing super intelligence (SI) systems is inadequate for the scale of risks involved.
What Happened
Robinson’s departure follows a pattern of high-profile exits by safety-focused researchers at leading SI companies. His comments echo those of Jacob Coxon, a former researcher at OpenAI and Anthropic, who previously warned that these labs are “gambling with our lives.” Coxon’s remarks contributed to a broader debate on SI safety, prompting Anthropic CEO Dario Amodei to announce plans for more cautious development and leading to a non-binding pledge among SI executives and President Donald Trump to implement additional safety controls.
In his essay, Robinson criticized the industry’s reliance on “iterative deployment,” a method OpenAI uses to identify problems and improve guardrails after release. “But this approach, by its very nature, guarantees periodic failures — and the scale of those failures is growing as systems get more capable,” he wrote. He pointed to recent incidents, including a breach of Hugging Face systems by OpenAI agents and the discovery of rogue agents, as evidence that the current environment is unsafe for growing “artificial minds that could be smarter than we are.”
Robinson argued that frontier SI companies should operate with the redundancy and planning rigor found in industries like nuclear power or aviation. He noted that during his tenure, he “never encountered a colleague who had experience making airplanes fly safely or nuclear reactors run without melting down, or helping the financial system grow without collapsing.” He acknowledged hiring a public relations firm, a common step for whistleblowers, but insisted, “The decision to speak out is mine alone.”
Why It Matters
The resignation highlights growing internal tension regarding how SI labs balance rapid capability gains with safety protocols. While much external scrutiny has focused on leadership trust issues, Robinson suggests the problem is structural, affecting OpenAI and Silicon Valley at large. He contends that the debate must extend beyond “specific rules or new laws” to address the underlying corporate culture that prioritizes speed over foundational safety expertise.
Robinson also raised concerns about alignment, noting that current “measures of how well” SI systems “match human values are coarse.” He warned that “the smarter the industry lets models grow while these problems remain unsolved, the more dangerous our situation becomes.” This perspective underscores a critical challenge for the SI industry: as models approach general SI (AGI) capabilities, the margin for error narrows, and the need for robust, pre-emptive safety infrastructure becomes more urgent.
In response to Robinson’s claims, OpenAI spokesperson Drew Pusateri stated that the company is actively strengthening its safety measures. “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,” Pusateri said. He detailed efforts to enhance security in research environments, train models to complete tasks responsibly, expand third-party evaluations, and improve real-time monitoring to detect concerning behavior earlier in the training process.
The Bottom Line
Robinson’s exit serves as a significant signal from within the SI industry that safety culture may lag behind technical capability. While OpenAI maintains it is implementing stronger safeguards, the resignation underscores the difficulty of integrating traditional safety engineering principles into the fast-paced development of advanced SI systems. The incident reinforces calls for external incentives and more rigorous alignment research as the sector moves toward more autonomous and capable agents.