A former senior employee at OpenAI has voiced concerns about the company’s focus on “shiny products” at the expense of safety, revealing that his departure was prompted by a disagreement over priorities that had reached a critical point.
Jan Leike, previously a key safety researcher and co-head of superalignment at OpenAI, raised these issues ahead of a global artificial intelligence summit in Seoul. His departure from the San Francisco-based company follows closely on the heels of Ilya Sutskever, OpenAI’s co-founder and another co-head of superalignment.
Leike explained his reasons for leaving in a post on X, highlighting a shift in safety culture as a central issue. He expressed concerns that safety processes had been sidelined in favor of developing new AI models, ultimately leading to his decision to resign.
“Over the past years, safety culture and processes have taken a backseat to shiny products,” Leike wrote.
OpenAI was established with the mission of ensuring that artificial general intelligence benefits humanity as a whole. However, Leike stated that he had been at odds with OpenAI’s leadership regarding the company’s priorities, culminating in his departure.
He emphasized the importance of investing more resources in areas such as safety, social impact, confidentiality, and security for the next generation of AI models. Leike stressed the challenges of building machines smarter than humans and called for OpenAI to prioritize safety as it moves forward.
In response, Sam Altman, OpenAI’s chief executive, acknowledged Leike’s concerns and expressed the company’s commitment to addressing them.
“He’s right we have a lot more to do; we are committed to doing it,” Altman wrote.
Meanwhile, Sutskever, in his own post announcing his departure, expressed confidence in OpenAI’s ability to develop safe and beneficial AGI under its current leadership.
Leike’s departure coincided with the release of a report by international AI experts, highlighting concerns about the potential for powerful AI systems to evade human control. The report underscored the need for regulators to keep pace with rapid technological advancements to ensure responsible AI development.
