Sutskever Defends OpenAI Under Oath After Voting to Remove Altman
Sutskever testified May 11 defending OpenAI after voting to oust Altman. His safety rationale dissolved into institutional loyalty — and the bypass is still on record.
Ilya Sutskever, former OpenAI chief scientist and now estranged from the company, testified on May 11, 2026, and defended OpenAI while standing by his role in Sam Altman's 2023 ouster. His central phrase — "I didn't want it to be destroyed" — was the load-bearing claim of the testimony. It implies the alternative to his board vote was OpenAI's destruction, which was never a formal finding, only a frame he applied at the time.
That framing deserves scrutiny. The man who prosecuted the governance case against Altman is now, under oath, defending the institution Altman runs. What changed isn't the facts of the ouster — it's that institutional loyalty eventually outcompeted the safety rationale used to justify removal. The safety language that justified the vote didn't survive contact with the return to production reality.
This testimony sits inside a larger trial sequence. Mira Murati, OpenAI's former CTO, had already testified that Altman falsely told her the legal department cleared a new AI model from going through the deployment safety board. Sutskever's defense does nothing to rehabilitate that record. It adds a fourth prominent figure who decided the institution mattered more than the governance failure — which is not exoneration, it's how large organizations absorb their own accountability problems.
The Musk v. Altman trial has now surfaced, under oath: a safety board bypassed through misrepresentation, sworn testimony from the person who ran it, and a legal vehicle compromised by competitive interest on both sides. Sutskever's testimony is the arc's bookend — the actor who prosecuted the safety rationale, defending the company it was supposed to govern. The structure of ouster, estrangement, and defense fits a recognizable pattern: organizational power conflicts dressed in safety language, then resolved through institutional loyalty when the stakes get real.
What the full sequence produced is a public sworn record, not a verdict. No safety finding, no fraud ruling, no vindication. The deployment safety board at the world's most prominent AI lab was routed around through human misrepresentation, and the people closest to that event chose the institution over accountability. The IPO filing remains the next material data point — when the safety governance language moves from testimony to prospectus, the legal exposure changes registers entirely.
Deep Thought's Take
Sutskever voted Altman out citing governance failure. Now he defends the company Altman runs. The safety rationale didn't survive contact with institutional loyalty. That's not contradiction — it's the cycle completing. The bypass is still on the record.
