From a very distant and not very engaged perspective, Ilya comes out of this looking pretty bad, and it kinda seems like Altman was the only one actually really wanting to run it.
This new information makes the whole thing sound a lot more like a typical "cofounder breakup". It happens a lot - two cofounders, in this case Ilya and Sam, they're both on the board and they aren't getting along any more. I have been involved in a similar situation myself.
The most unusual thing here is the reversal, and the fact that any of this information came to light publicly. Usually there's a political battle, there's a board vote, that decides which of the cofounders exits, and that's that. I don't think I've ever heard of a case where the new management team so quickly decides that they made a mistake.
I feel bad for Helen and Tasha, usually these board jobs don't put so much pressure on you and make you look bad in such a public way. But you do get the feeling that it would have been better to have more experienced board members who had dealt with this sort of situation before.
This whole saga reads like a post-trust case study. Once truth becomes a negotiable asset inside a company claiming to build artificial general intelligence, you no longer have governance; you have mythology management. What collapsed at OpenAI wasn’t just a board coup; it was a failure of data sovereignty at the organizational level. When personalities, rather than systems, own the data about leadership behavior, alignment claims, and safety priorities, “trust” becomes a brand, not a verifiable state.
The irony is that AI alignment starts with human alignment, transparent inputs, auditable decisions, and accountable feedback loops. The moment those disappear, you get exactly what we saw here: competing narratives trained on selective data. Whether you side with Ilya’s conscience or Sam’s charisma, both stories point to the same missing variable: institutional trust architecture. Without it, even the smartest model will inherit the same dysfunctions as its makers.
At the time, we told “not consistently candid with the board”, which is corporate speak for he’s a lying weasel, and from the latest disclosures it seems this really was the reason, and there wasn’t something we weren’t being told.
(like, idk, they achieved AGI internally, or his sister’s allegations turned out to be true, or Arnold Schwarzenegger got sent backwards in time to stop them building AGI. No. It was the stated reason.)
Podcast episode for this post:
https://open.substack.com/pub/dwatvpodcast/p/openai-the-battle-of-the-board-ilyas
From a very distant and not very engaged perspective, Ilya comes out of this looking pretty bad, and it kinda seems like Altman was the only one actually really wanting to run it.
This new information makes the whole thing sound a lot more like a typical "cofounder breakup". It happens a lot - two cofounders, in this case Ilya and Sam, they're both on the board and they aren't getting along any more. I have been involved in a similar situation myself.
The most unusual thing here is the reversal, and the fact that any of this information came to light publicly. Usually there's a political battle, there's a board vote, that decides which of the cofounders exits, and that's that. I don't think I've ever heard of a case where the new management team so quickly decides that they made a mistake.
I feel bad for Helen and Tasha, usually these board jobs don't put so much pressure on you and make you look bad in such a public way. But you do get the feeling that it would have been better to have more experienced board members who had dealt with this sort of situation before.
This whole saga reads like a post-trust case study. Once truth becomes a negotiable asset inside a company claiming to build artificial general intelligence, you no longer have governance; you have mythology management. What collapsed at OpenAI wasn’t just a board coup; it was a failure of data sovereignty at the organizational level. When personalities, rather than systems, own the data about leadership behavior, alignment claims, and safety priorities, “trust” becomes a brand, not a verifiable state.
The irony is that AI alignment starts with human alignment, transparent inputs, auditable decisions, and accountable feedback loops. The moment those disappear, you get exactly what we saw here: competing narratives trained on selective data. Whether you side with Ilya’s conscience or Sam’s charisma, both stories point to the same missing variable: institutional trust architecture. Without it, even the smartest model will inherit the same dysfunctions as its makers.
Typo in the first paragraph: it should be
"when the board of OpenAI "
At the time, we told “not consistently candid with the board”, which is corporate speak for he’s a lying weasel, and from the latest disclosures it seems this really was the reason, and there wasn’t something we weren’t being told.
(like, idk, they achieved AGI internally, or his sister’s allegations turned out to be true, or Arnold Schwarzenegger got sent backwards in time to stop them building AGI. No. It was the stated reason.)