Ilya Sutskever's deposition reveals OpenAI coup was 'rushed' and almost led to Anthropic merger
Nov 3, 2025
Key Points
- Ilya Sutskever's deposition reveals the November 2023 OpenAI board coup against Sam Altman involved a year-long plot with Mira Murati and a 52-page memo documenting allegations of lying, but Sutskever acknowledged the process was rushed due to board inexperience.
- The coup nearly triggered a merger between OpenAI and Anthropic, with Sutskever explicitly stating he did not want the combination, suggesting the internal chaos extended far beyond Altman's removal.
- When the board capitulated and reinstated Altman days later, Sutskever framed the reversal as pragmatic acceptance that Altman was most likely to survive and lead, a capitulation that signaled his own unfitness to lead what followed.
Summary
Ilya Sutskever's deposition, now public, shows the November 2023 board coup against Sam Altman was more ambitious and more chaotic than previously known.
Sutskever plotted with Mira Murati for over a year to remove Altman. Separately, Dario Amodei of Anthropic wanted Greg Brockman fired and sought to position himself in charge of all research, suggesting the coup had deeper fractures than a simple Altman ouster. Murati told Sutskever that Altman pitted her against Amodei, adding another layer of internal manipulation.
Sutskever prepared extensively, drafting a 52-page memo documenting allegations that Altman exhibited a consistent pattern of lying, along with a separate document targeting Brockman.
When pressed in deposition about whether the process was rushed, Sutskever acknowledged it was. He attributed it to board inexperience in board matters. The board lacked the institutional knowledge or discipline to execute a coordinated removal of a founder-CEO.
The coup nearly triggered an even larger restructuring. Sutskever states he was "very unhappy" about something and explicitly says "I really did not want OpenAI to merge with Anthropic," implying that during the chaos, a merger with Anthropic was actually on the table. That would have fundamentally rewritten AI's startup landscape.
When the board capitulated and reinstated Altman days later, Sutskever was asked to justify the reversal. His answer framed the capitulation as pragmatic resignation. Control of AGI will inevitably flow to someone adept with power, much like choosing between politicians. Altman, whatever his flaws, was the one most likely to survive and lead.
Commentary afterward frames Sutskever as a capable scientist but not a practical operator. The adage applies: if you go for the king, you better get his head. Sutskever didn't. The takeaway from the community is blunt. The inability to execute a coup is itself a signal of unfitness to lead what came next.