In a deposition revealed recently, Ilya Sutskever, co-founder and former chief scientist of OpenAI, testified that he spent more than a year gathering evidence against then-CEO Sam Altman — eventually compiling a 52-page memo alleging that Altman systematically lied to the board, pitted executives against each other and concealed information from governance processes. This internal conflict helped spark Altman’s ouster in November 2023.
Sutskever claimed that Altman repeatedly provided conflicting information about company strategy — for example telling different executives divergent things about OpenAI’s research direction and board-reporting norms — which bred distrust among senior leadership. Sutskever described the memo’s creation as a board-sanctioned effort: he was asked by independent directors to compile concerns, and sent it via disappearing email, fearing that Altman might “make it disappear” if alerted.
Even more dramatic, the deposition revealed that during the week after Altman’s removal, OpenAI’s board considered a merger with rival AI company Anthropic — a move Sutskever says he opposed. Within days, however, the employee revolt overwhelmed the board, and Altman was reinstated as CEO just four days later, forcing the ousted leadership group (including Sutskever) to retreat or leave.
The broader significance of this testimony extends beyond Silicon Valley intrigue: it shines a light on the governance risks present in organisations developing frontier AI. When a high-stakes company such as OpenAI experiences internal breaches of trust, opaque decision-making and leadership conflicts at the top, questions arise about accountability, alignment with mission, and the stability of institutions handling highly consequential technologies.
