David Robinson, a leader on OpenAI’s security systems team who left the company this week, once had ambitions of bringing philosophy to journalism.
A Rhodes Scholar, Robinson graduated from Princeton in 2004 with a degree in philosophy. At Princeton, he ran the opinion section of the school newspaper.
In a 2003 interview with the Daily Princetonian, Robinson said he wanted to be the kind of journalist who “tells you the facts and then addresses the philosophical differences that explain the situation.”
Today, Robinson’s name is headline news – and a notable signing. He left OpenAI, the same week three other security researchers were fired, and published a scathing essay for The Atlantic about what he called OpenAI’s inadequate approach to security.
“I agree with other recently departed employees that the companies building this technology are not careful enough,” Robinson wrote in the essay.
Robinson and the public relations firm he said he hired responded to Business Insider’s requests for comment.
A former colleague at OpenAI, who was responsible for futurism at the company until July, said in an article on X that Robinson was not an ideologue.
“My impression of David, from our time at OpenAI, is of a sober, thoughtful person who did not approach the work from an ideological angle,” wrote Joshua Achiam.
“He’s not someone who came in with pre-formed opinions about existential risk,” he added. “He has always been productive, collaborative and well-reasoned in his approach to security.”
A review of Robinson’s published essays, interviews and books dating back to his undergraduate years paints a portrait of a man deeply concerned about the impact of technology on humanity.
In 2002, he wrote “Voices in the Code,” published by the Russell Sage Foundation, about creating an algorithm for kidney transplant donation.
His resume says he is writing a second book called “Reclaiming Who We Are” that will use “social psychology, Buddhist philosophy, feminist theory, history and sociology to argue for more judicious use of algorithmic labeling systems.”
“Throughout my life, I’ve been fascinated by the rules that surround technology and how we decide when and how we’re going to use it,” he said in a 2022 interview on the Princeton Alumni Weekly podcast.
From journalism to technology policy
Robinson grew up in Potomac, MD, an affluent suburb of Washington, D.C.
When he was a child in the 1990s with a mild case of cerebral palsy, Robinson said, his teachers created a new rule that allowed him to use a word processor instead of pen and paper.
“Since then, I have been interested in the social impacts of new digital technologies,” he said in an interview at UC Berkeley in 2021.
During his undergraduate summers at Princeton, he interned at Time Magazine and the Wall Street Journal, according to his LinkedIn page.
He told the Daily Princetonian in 2003 that after gaining experience as a journalist, he hoped to write for a literary magazine like The New Yorker or Harper’s.
After Princeton, Robinson earned a degree in philosophy, politics, and economics at Oxford before returning to Princeton to establish a research center on the relationship between technology and public policy.
“Part of the idea was that if you’re navigating the policy and value choices that present themselves around new technologies, it’s a big help to have real deep technical expertise,” he said in the Berkeley interview.
In 2011, Robinson and a Princeton colleague founded Upturn, an organization that aimed to use algorithmic technologies to make government data more accessible.
He joined OpenAI after lecturing on ethics to tech executives.
After research and teaching roles at several universities and a year-long stint as faculty director at Apple University, where Robinson gave seminars to Apple executives on ethical issues raised by new technologies, Robinson joined OpenAI in 2023.
Originally hired as head of policy planning at the creator of ChatGPT, Robinson wrote in a LinkedIn announcement of his new position: “I look forward to helping shape OpenAI’s engagement in the broader policy debate on AI at this critical time.”
During a 2023 appearance at the World Economic Forum’s AI Governance Summit, Robinson said OpenAI conducts rigorous security checks before releasing models and that the company is not motivated solely by profit.
“We believe in rolling out gradually and then learning as you go,” he said.
Recently, Robinson posted concerns about OpenAI’s ability to balance security with increasingly advanced models.
“Things @openai are indeed changing significantly from day to day, but I don’t know if we’re changing fast enough,” he said last month. “I think it’s still true, as @hilbertspaess says, that many of us ‘have not deeply internalized the civilizational issues’.”
Then, on Friday, Business Insider exclusively reported that Robinson was away from the company. The next day, he published a long essay in The Atlantic explaining why he left.
What’s next for Robinson?
Robinson’s departure follows that of Jacob Coxon, a former Anthropic engineer who resigned last month and warned that AI could cause mass extinction. In his Atlantic essay, Robinson addressed the argument that he could do more to improve security by staying at OpenAI than by leaving.
Robinson wrote that by working outside, he “can help more people understand the risks I’ve seen and strengthen the incentives OpenAI and other companies have to be safer.”
He said he chose to leave because “as the company moves from one launch to the next, it fails to achieve the level of care that I believe is necessary.”
In response to Robinson’s essay, an OpenAI spokesperson said in a statement that the company works to ensure that AI models do not exceed the security measures it has implemented.
“As our models become more capable, we continue to strengthen our safety and security practices to address the risks we see today, while developing safety cases to anticipate the risks of future advancements,” the spokesperson said. “We ensure that our models do not become more capable than we can safely handle and secure, and we pause training or hold back models when we need to slow down.”
Gn bussni
