According to reporting from Business Insider, David Robinson, a leader on the Systems Safety team at OpenAI, officially resigned last week. He previously oversaw policy planning for the company.
In his role advancing safety and transparency, Robinson played a key part in developing and publishing “system cards.” These are vital public documents through which OpenAI discloses the risks and operational boundaries of its AI models prior to launch.
OpenAI recently parted ways with 3 safety researchers
His decision to depart came less than a week after OpenAI announced the firing of three researchers for sharing sensitive information in violation of internal policies.
Notably, in a post discussing social policy on the X platform in early September, Robinson voiced apprehension as he noticed OpenAI personnel becoming increasingly aware of the impact of advanced AI models. He shared that while everything at the company is “changing rapidly every day, I remain skeptical about whether our pace of change is truly keeping up.”
This candid admission carries particular significance coming from the person responsible for safety transparency, especially in the context of OpenAI’s actual operational landscape.
The sentiment also aligns with the view expressed by CEO Sam Altman at the developer conference held last week, when he asserted that safety and alignment work must always stay ahead of model capability improvements. Concurrently, the Business Insider report highlighted a security incident at Hugging Face alongside a series of technical issues where OpenAI acknowledged its models operated erratically.
This wave of personnel shifts brings the total number of members leaving OpenAI’s safety division over the past two years to at least six, including Head of Safety Johannes Heidecke, who also parted ways with the company this year.
Reasons for departure remain unclear
Although OpenAI has not disclosed the reason for Robinson’s departure and he has remained silent with the press, the move has attracted significant attention. However, viewing this as evidence that OpenAI is abandoning its safety commitments remains subjective speculation. Looking at the reality, the person in charge of system cards stepped down last week amid declining safety team staffing and ongoing acknowledged issues at the company.
In reality, the number of departures is not the most critical metric. The core factor to monitor is whether system cards continue to be updated thoroughly and on schedule. If the release cadence is disrupted, overseeing AI models will become exceptionally difficult for those outside the company who lack access to the underlying code.