A newly released court deposition reveals the chaotic days surrounding Sam Altman’s 2023 ouster from OpenAI, and shows just how close the company came to merging with its top rival, Anthropic.
The sworn testimony is from OpenAI co-founder and former chief scientist Ilya Sutskever, who spoke under oath as part of Elon Musk’s lawsuit against OpenAI’s Sam Altman.
The deposition:
To dig into this on-the-record testimony, I talked with SmarterX and Marketing AI Institute founder and CEO Paul Roetzer on Episode 179 of The Artificial Intelligence Show.
Sutskever’s deposition confirms that in the immediate aftermath of Altman’s firing, talks began to merge OpenAI with Anthropic, a move that would have likely installed Anthropic’s CEO, Dario Amodei, at the helm.
Anthropic initially expressed “excitement” about the potential deal, according to Sutskever.
The talks quickly collapsed, however, due to what Sutskever described as “practical obstacles,” likely related to the complex investor agreements both companies held. Sutskever also described “deep mistrust” within OpenAI’s leadership.
He also testified that other board members were a lot more supportive and identified former board member Helen Toner as suggesting the merger.
Toner, however, quickly took to social media after the deposition was released to dispute this. She tweeted:
“For the record, for those dissecting Ilya's deposition: This part is false. I wasn't the one who made the board <> Anthropic call happen, and I disagree with his recollection that board members other than him were supportive of a merge.”
Roetzer notes that her public correction adds another layer of complexity to the competing narratives.
Beyond the merger drama, the deposition provided stunning, on-the-record confirmation of the 52-page memo Sutskever wrote detailing his case against Altman.
He sent the memo only to the board’s independent directors, testifying that he intentionally kept Altman in the dark. Why?
“Because I felt that, had he become aware of these discussions, he would just find a way to make them disappear,” Sutskever says.
The memo was damning. Its very first page, read into the court record, stated: “Sam exhibits a consistent pattern of lying, undermining his execs, and pitting his execs against one another.”
When asked what action he believed was appropriate based on that memo, Sutskever’s answer was: “Termination.”
The testimony also revealed that Sutskever’s memo was built, in part, on evidence supplied by others, including then-OpenAI Chief Technology Officer Mira Murati. Sutskever testified that many of the screenshots documenting Altman’s behavior came from Murati.
But perhaps the biggest surprise was the revelation that Altman wasn't the only target. Sutskever confirmed he also drafted a similar memo critical of OpenAI co-founder (and current President) Greg Brockman.
The entire saga, from secret memos and executive infighting to a near rival merger, paints a picture of internal conflict far more intense than previously known.
“If anybody watched The Social Network and the Facebook crazy story, this is it on steroids,” Roetzer says. “The OpenAI movie is going to be insane.”