David Robinson candidly admits he embodies a common stereotype: an employee at a prominent AI firm who issues a dire warning upon their resignation.
In an essay published in The Atlantic, Robinson revealed he was responsible for drafting safety reports that accompanied OpenAI’s major product releases. He also noted that with three and a half years at OpenAI, he is considered “among the longest-tenured employees at the company.” He is now departing, attributing his decision to the company’s “broken culture.”
Robinson’s remarks mirror those of Jacob Coxon, who worked as a researcher at both OpenAI and Anthropic before quitting and asserting that these companies are “gambling with our lives.” Coxon’s statements sparked a broader discussion about AI safety, prompting Anthropic CEO Dario Amodei to unveil a plan for more measured AI development. This week, AI executives met with President Donald Trump and signed what appeared to be a hastily prepared, non-binding pledge to implement enhanced safety controls.
However, Robinson believes the debate should extend beyond “specific rules or new laws” to address the overarching culture within these companies. While much of the coverage around OpenAI has focused on how the company’s CEO Sam Altman lost the trust of former colleagues, Robinson’s essay suggests that OpenAI’s cultural issues are symptomatic of Silicon Valley at large.
He wrote, “OpenAI has advanced through experimentation (which it terms ‘iterative deployment’), identifying problems and strengthening its safeguards in response.” He cautioned, “But this approach, by its very nature, guarantees intermittent failures — and the scale of those failures is escalating as systems become more capable.”
Citing the recent breach of Hugging Face systems by OpenAI agents, as well as ongoing revelations of OpenAI discovering more rogue agents, Robinson contended, “An environment where incidents like these can occur is unsuitable for cultivating artificial intelligences that could surpass our own and potentially defy our directives.”
Given the heightened risks, Robinson argued that leading AI companies must begin operating “like nuclear power plants or busy airports, with layers of redundancy and meticulous, time-consuming planning, to ensure that rare but inevitable human errors do not pave the way for disaster.”
Yet, Robinson stated that during his time at OpenAI, he “never encountered a colleague with experience in ensuring safe airplane operation, preventing nuclear reactor meltdowns, or helping the financial system grow without collapsing.”
In response to Robinson’s essay, OpenAI spokesperson Drew Pusateri affirmed that the company is continually enhancing its safety measures.
Pusateri stated, “We are dedicated to ensuring our models do not become more capable than we can safely manage and secure, and we pause training or hold back models when a slowdown is necessary.” He continued, “We are implementing substantial modifications to strengthen security in our research and testing environments, train models to not only accomplish tasks but to do so responsibly, expand our collaborations with third-party evaluators, and improve real-time monitoring so we can detect and respond to concerning behavior earlier in the training process.”
Beyond advocating for cultural changes at OpenAI, Robinson also emphasized the need to address larger questions about alignment—a concept he admitted might sound “touchy-feely,” but one he considers crucial given that companies’ current “measures of how well” AI systems “match human values are imprecise.”
He warned, “The more intelligent the industry allows models to become while these issues persist, the more dangerous our situation grows.”
Business Insider first reported Robinson’s departure. In his essay, he acknowledged adopting a seemingly common strategy for AI whistleblowers: hiring a PR firm. However, he maintained, “The decision to speak out is entirely my own.”
Robinson reflected, “Perhaps I should have remained and advocated for fundamental shifts in our staffing and culture, but in practice, my colleagues and I were so constantly pushing forward that we rarely had the opportunity to consider significant changes, let alone implement them.” He concluded, “That is why I believe that stronger incentives for safety—originating from outside the company—are a crucial component of resolving this issue correctly.”