Ilya Sutskever officially departs OpenAI

Just one day after OpenAI stunned the world with GPT-4o, the other shoe finally dropped: co-founder and longtime Chief Scientist Ilya Sutskever is officially leaving the company.

Sam Altman announced the departure on X on Tuesday evening, praising Sutskever as “easily one of the greatest minds of our generation.” OpenAI’s Director of Research, Jakub Pachocki, has been named the new Chief Scientist.

Adding to the fallout, Jan Leike, who co-led OpenAI’s high-profile Superalignment team alongside Sutskever, has also resigned, according to The New York Times. Co-founder John Schulman will reportedly take over oversight of the group.

The November fallout reaches its conclusion

Sutskever’s departure officially caps a chaotic corporate drama that kicked off last November.

Ahead of Thanksgiving, Sutskever was a driving force behind the board’s sudden firing of Altman, reportedly frustrated that Altman was rushing commercial AI products out the door at the expense of safety research. After a massive employee revolt and investor backlash, Sutskever reversed his stance, and Altman was reinstated within days.

According to The Times, Sutskever never returned to work after that episode. In reality, Pachocki had effectively served as chief scientist since late last year.

Pachocki, who holds a PhD from Carnegie Mellon, joined OpenAI in 2017. He previously led the Dota 2 gaming AI project and spearheaded the development of GPT-4, establishing himself as a driving force behind OpenAI's deep learning scaling strategy.

In his exit note, Sutskever expressed confidence that OpenAI will build safe and beneficial AGI, adding that he is moving on to an undisclosed project that is “very personally meaningful.”

Why it matters

Ilya Sutskever was widely regarded as the scientific soul of OpenAI. His exit signals the definitive end of OpenAI’s era as an idealistic research collective, sealing its transformation into a relentless, product-driven tech behemoth.

Sutskever’s track record spans the original ImageNet breakthrough under Geoffrey Hinton, AlphaGo at DeepMind, and the architectural heavy lifting behind GPT. While Pachocki is universally respected for shipping GPT-4, losing Sutskever—along with safety lead Jan Leike—raises real alarm bells about internal oversight.

The Superalignment team was formed less than a year ago to rein in hypothetical superintelligence, backed by a promise of significant compute resources. With its top leadership wiped out and reports that the unit will be “deeply integrated” across general research, we think OpenAI’s safety guardrails are actively being subordinated to commercial release cycles.

For the broader industry, the biggest question now is what Sutskever builds next. If he follows in the footsteps of Anthropic’s founders and launches a rival safety-first lab, he will have zero trouble securing top researchers and massive capital overnight.