Key Safety Researchers Depart OpenAI

Summary

OpenAI co-founder and chief scientist Ilya Sutskever and superalignment team co-lead Jan Leike both left the company in May 2024, with Leike publicly criticizing OpenAI for deprioritizing safety in favor of "shiny products." The departures, following several other safety researchers leaving for Anthropic and other organizations, raised serious questions about OpenAI's commitment to its safety mission.

What Happened

On May 14, 2024, Ilya Sutskever — co-founder of OpenAI, its chief scientist, and a central figure in the November 2023 board crisis — announced his departure from the company. Sutskever, who had initially supported Altman's firing before switching sides, had been largely absent from OpenAI since the crisis. He would go on to found Safe Superintelligence Inc. (SSI), a new company focused exclusively on safety research.

The following day, Jan Leike, who co-led OpenAI's Superalignment team alongside Sutskever, also resigned. Leike was more publicly critical, posting on X that "over the past years, safety culture and processes have taken a backseat to shiny products" and that he had reached a breaking point over "the ability to do this safety-critical work at OpenAI." Leike subsequently joined Anthropic to lead its alignment science team.

The Superalignment team, which OpenAI had announced in July 2023 with a commitment to dedicate 20% of its compute resources to alignment research, was effectively dissolved. Other safety-focused researchers had departed in the preceding months, creating a pattern that was difficult to dismiss as coincidence.

Why It Matters

The safety departures struck at the core of OpenAI's self-narrative. The company had consistently justified its commercial activities — the Microsoft partnership, the capped-profit structure, the aggressive product launches — as necessary to fund its safety mission. If the most senior safety researchers at the company were leaving because that mission was being deprioritized, it undermined the entire rationale.

Leike's public criticism was particularly damaging because it was specific and came from an insider. Rather than vague concerns about corporate direction, he explicitly stated that safety culture had deteriorated and that the Superalignment team lacked the resources and attention it needed.

Sutskever's departure and the founding of SSI carried a different but equally significant message: the co-founder and chief scientist of OpenAI concluded that he could better pursue AI safety outside the company he had helped create. Whether this reflected an irreconcilable tension between commercial AI development and safety research, or simply organizational dysfunction, was debated — but neither interpretation reflected well on OpenAI.

The pattern of safety talent flowing from OpenAI to Anthropic (which was itself founded by former OpenAI safety researchers) continued a narrative that had been building since Anthropic's founding in 2021: that the organization most loudly proclaiming its commitment to safety was consistently losing the people most committed to doing safety work.