David Robinson, one of the leaders of OpenAI’s safety team, has left the company over concerns the company and other AI firms aren’t “being nearly careful enough” in the development of their technology.

Robinson left OpenAI in late September and published his views in an essay in the magazine on 3 October, writing that he quit the firm because “its culture is broken”.

He wrote that a widely publicised hack of AI technology platform Hugging Face earlier this year, which spurred AI companies to uncover and disclose dozens of other security incidents, was “typical of the industry, given the speed and flexibility with which people operate”.

Security concerns

“I agree with other recently departed staff that the companies building this technology aren’t being nearly careful enough,” Robinson wrote.

“As the company sprints from one launch to the next, it is failing to achieve the level of care that I believe is needed.”

He said the company’s insufficient approach to safety arises from its culture of “unimpeded optimism” about being able to solve problems as they arise.

In response, OpenAI said in a statement that it is working to ensure that the capabilities of its models don’t exceed what it can “safely manage and secure”.

Robinson’s role at OpenAI involved transparency around safety issues, including developing and sharing reports that accompany the firm’s releases.

His resignation comes at a time when AI firms are facing criticism over the numerous security incidents that have come to light in recent weeks, and which have only been made public under outside pressure.

OpenAI last week acknowledged firing three staff members involved with safety and alignment for sharing company data, following a report that indicated one of the fired employees was the lead technical liaison with two outside firms investigating the Hugging Face breach.

The firings raise awkward questions around OpenAI’s approach to security and the disclosure of potential problems.