DooDooLamb News
Ilya Sutskever Stands by His Role in Sam Altman’s OpenAI Ouster: ‘I Didn’t Want It to Be Destroyed’
Brief published May 13, 2026 · Original source published May 11, 2026
Original reporting by Paresh Dave, Maxwell Zeff at wired.com.
Automated brief. Verify important details at the original source.
Former OpenAI Chief Scientist Explains Why He Had to Destroy the Village to Save It
Ilya Sutskever took the witness stand on Monday to defend his role in temporarily firing Sam Altman, offering the kind of logic that makes perfect sense if you've spent your entire career thinking about artificial intelligence and absolutely none if you haven't. The former OpenAI chief scientist testified that his participation in November's boardroom coup wasn't about destroying the company, but rather about preventing its destruction, which apparently required destroying it first.
This courtroom drama unfolded as Elon Musk's lawsuit against OpenAI and Microsoft reached its final act, transforming what should have been a straightforward breach of contract case into something resembling a tech industry therapy session broadcast on Court TV. Sutskever, who helped orchestrate Altman's brief exile before immediately regretting it and begging for his return, found himself in the uncomfortable position of explaining why his attempt to save OpenAI looked suspiciously identical to setting it on fire.
The backstory reads like a Silicon Valley screenplay written by someone who gave up trying to make it believable. Last November, OpenAI's board (a collection of individuals whose combined business experience could fit in a startup's break room) decided that Altman was moving too fast in commercializing their AI technology. Their solution was to fire the CEO who had turned their research lab into a company worth $157 billion, apparently without considering that this might create some minor operational difficulties. Sutskever, as chief scientist, supported this decision with the kind of theoretical purity that sounds brilliant in academic papers and catastrophic in board meetings.
The firing lasted approximately four days, during which time OpenAI employees threatened mass exodus, Microsoft scrambled to figure out what they'd actually invested in, and the entire AI industry watched with the fascination typically reserved for particularly spectacular car crashes. Sutskever, witnessing the immediate implosion of everything he'd helped build, performed what might be the fastest philosophical U-turn in corporate history. He went from supporting Altman's removal to signing the employee letter demanding his reinstatement, presumably after someone explained to him how companies actually work.
Now, months later and no longer with OpenAI, Sutskever testified on behalf of his former employer in Musk's lawsuit. The case itself stems from Musk's claim that OpenAI violated its founding mission by becoming a for-profit entity focused on building products rather than ensuring artificial general intelligence benefits all of humanity. It's a lawsuit that essentially argues "you promised to save the world for free, but instead you're trying to save the world for money," which perfectly captures the kind of philosophical hair-splitting that happens when idealistic manifestos meet capitalist reality.
Sutskever's testimony revealed the central tension that's plagued OpenAI since its founding: the gap between building safe, beneficial AI and building a sustainable business. As chief scientist, he represented the faction that wanted to move carefully, ensuring their technology wouldn't accidentally end civilization. This is admirable in theory and somewhat impractical when you're competing against companies that view "move fast and break things" as a business strategy rather than a cautionary tale. The board's attempt to pump the brakes by firing their CEO was like trying to slow down a Formula 1 car by ejecting the driver mid-race.
The irony wasn't lost on observers that Sutskever was defending OpenAI in a courthouse while no longer working there. He left the company in May to start his own AI safety venture, presumably one where he won't need to choose between scientific caution and quarterly revenue targets. His departure marked the end of an era for OpenAI's research-first culture, though that era had arguably ended the moment they started selling API access to anyone with a credit card.
Microsoft CEO Satya Nadella also testified, bringing the perspective of someone who invested $13 billion in OpenAI and watched his partners temporarily lose their minds. Nadella's testimony likely focused on the more mundane aspects of their partnership, like how Microsoft integrates OpenAI's technology into products that actual humans use for actual work. This stands in stark contrast to the philosophical debates about AI safety that seem to consume much of OpenAI's internal bandwidth.
The whole spectacle underscores how the AI industry has evolved from a collection of research labs into something resembling a traditional tech sector, complete with corporate drama, strategic lawsuits, and former colleagues testifying against each other. What started as a movement to democratize artificial intelligence has become a very expensive argument about who gets to democratize it and how much they're allowed to charge for the privilege.
Sutskever's defense of his former company, despite their messy divorce, suggests either remarkable professional grace or the recognition that OpenAI's success validates his life's work regardless of who's currently running the show. His testimony essentially argued that the November board crisis, while spectacularly mishandled, came from a place of genuine concern about the technology they were unleashing on the world. It's the kind of noble motivation that's easier to appreciate in hindsight than it was during the four days when OpenAI looked like it might collapse entirely.
The real lesson here isn't about corporate governance or AI safety, but about the fundamental impossibility of maintaining research purity while building products that billions of people actually use. Sutskever learned this the hard way, discovering that sometimes the best way to destroy something is to try too hard to save it. His former colleagues at OpenAI are presumably taking notes.