\n<\/p>\n
By his own admission, David Robinson is \u201csomething of a clich\u00e9\u201d: an employee at a leading AI company who issues a dire warning while resigning from their job.<\/p>\n
In an essay published in The Atlantic<\/a>, Robinson said he led the writing of safety reports that accompanied OpenAI\u2019s major product launches. He also said that with three-and-a-half years at OpenAI, he is \u201camong the longest-tenured employees at the company.\u201d Now he\u2019s quitting, because in his view, the company\u2019s \u201cculture is broken.\u201d<\/p>\n In some ways, Robinson\u2019s comments echo those of Jacob Coxon, who worked as a researcher at both OpenAI and Anthropic before quitting and declaring that these companies are \u201cgambling with our lives.\u201d<\/a> Coxon\u2019s comments led to a broader debate about AI safety, with Anthropic CEO Dario Amodei unveiling a plan for more cautious AI development<\/a>; AI executives met with President Donald Trump this week and signed what appeared to be hastily written, non-binding\u00a0 pledge to implement more safety controls<\/a>.<\/p>\n But in Robinson\u2019s view, the debate needs to go beyond \u201cspecific rules or new laws,\u201d addressing the overall culture at these companies. And while much of the reporting around OpenAI has focused on how the company\u2019s CEO Sam Altman lost the trust of former colleagues<\/a>, Robinson\u2019s essay suggests that OpenAI\u2019s culture issues are the same as those of Silicon Valley at large.<\/p>\n \u201cOpenAI has thrived by trial and error (which it calls \u2018iterative deployment\u2019), looking for problems and improving its guardrails in response,\u201d he wrote. \u201cBut this approach, by its very nature, guarantees periodic failures \u2014 and the scale of those failures is growing as systems get more capable.\u201d<\/p>\n Pointing to the recent breach of Hugging Face systems by OpenAI agents, as well as continuing revelations of OpenAI discovering more rogue agents<\/a>, Robinson argued, \u201cAn environment where things like this can happen is no place to grow artificial minds that could be smarter than we are and that might not do what we want them to.\u201d<\/p>\n Given the increased risk, Robinson argued that frontier AI companies need to start operating \u201clike nuclear-power plants or busy airports, with layers of redundancy and careful, time-consuming planning, so that the occasional and inevitable human error does not open a door to disaster.\u201d\u00a0<\/p>\n But Robinson said that in his time at OpenAI, he \u201cnever 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.\u201d<\/p>\n In response to Robinson\u2019s essay, OpenAI spokesperson Drew Pusateri said the company continues to improve its safety measures.<\/p>\n \u201cWe\u2019re making sure our models don\u2019t become more capable than we can safely manage and secure, and we pause training or hold back models when we need to slow down,\u201d Pusateri said in a statement. \u201cWe\u2019re making significant changes to strengthen security in our research and testing environments, train models to not just complete tasks but do so responsibly, expand our work with third-party evaluators, and improve real-time monitoring so we can detect and respond to concerning behavior earlier in the training process.\u201d<\/p>\n Beyond calling for changes in OpenAI\u2019s culture, Robinson also said it\u2019s time to ask bigger questions about alignment \u2014 something that he admitted could sound \u201ctouchy-feely,\u201d but he said it\u2019s critical as companies\u2019 current \u201cmeasures of how well\u201d AI systems \u201cmatch human values are coarse.\u201d<\/p>\n \u201cThe smarter the industry lets models grow while these problems remain unsolved, the more dangerous our situation becomes,\u201d he said.<\/p>\n Robinson\u2019s departure was first reported by Business Insider<\/a>. In his essay, he also acknowledged that he\u2019s following an apparently a common step in the AI whistleblower playbook: He\u2019s hired a PR firm<\/a>. But he insisted, \u201cThe decision to speak out is mine alone.\u201d\u00a0<\/p>\n \u201cPerhaps I should have stayed and fought for fundamental shifts in our staffing and culture, but in practice, my colleagues and I were so busy sprinting that we seldom had the chance to consider big changes, much less to actually make them,\u201d Robinson said. \u201cThat\u2019s why I concluded that stronger incentives for safety \u2014 coming from outside the company \u2014 are a big part of getting this right.\u201d<\/p>\n<\/div>\n When you purchase through links in our articles, we may earn a small commission<\/a>. This doesn\u2019t affect our editorial independence.<\/em><\/p>\n