Five Days That Shook OpenAI
.jpeg)
Five Days That Shook OpenAI
Inside the Sam Altman firing, the merger talks with Anthropic, Altman's sudden reinstatement, and the unresolved fallout
#SamAltman #ElonMusk #OpenAI #FrontierModel
This is a Tech-Reader AI Digest Special Edition.
This story didn't end in November 2023. It is ending now — in a federal courthouse in Oakland, in sworn testimony, in a record that will outlast the verdict.
Three weeks of trial testimony have assembled the most complete picture yet of what actually happened during five days in November 2023 that reshaped the AI industry. Not the version that was managed for the press. Not the version OpenAI's communications team released. The version that Ilya Sutskever, Mira Murati, Tasha McCauley, and Satya Nadella have sworn to under oath.
Here is that story — assembled from the trial record and two and a half years of reporting — told in sequence for the first time.
Part 1: The Build-Up
The firing didn't happen in a day. It was assembled over years.
OpenAI was founded in December 2015 as a nonprofit with a specific premise: that artificial general intelligence was coming, and that the organization developing it should be structured to benefit humanity rather than shareholders. The founding donors — Musk, Peter Thiel, Amazon Web Services, and others — gave money to a charity. Altman became CEO in 2019 and immediately set about building the structure that would make OpenAI commercially viable.
In 2019, Altman, Brockman, and Sutskever formed OpenAI LP — a for-profit subsidiary capped at 100x returns — nested inside the nonprofit. Microsoft invested $1 billion. Then $2 billion in 2021. Then $10 billion in 2023. The nonprofit board retained governance control. But the commercial engine was running.
The tensions began accumulating well before November 2023. Sutskever and his allies opposed Altman's efforts to seek billions from Middle Eastern sovereign wealth funds to develop an AI chip to compete with Nvidia, viewing these moves as unjustly using the OpenAI name. Altman reduced Sutskever's role in October 2023, furthering divisions.
Then came DevDay — OpenAI's developer conference on November 6, 2023. Altman announced that OpenAI would make tools available so anyone could create their own version of ChatGPT. For Sutskever and the board, that was a step too far.
Meanwhile, the board had been quietly accumulating evidence. In October — a month before the firing — the board had conversations with two executives who relayed experiences with Altman they weren't comfortable sharing before, including screenshots and documentation of problematic interactions and mistruths. "The two of them suddenly started telling us how they couldn't trust him, about the toxic atmosphere he was creating," former board member Helen Toner later said. "They used the phrase 'psychological abuse.'"
Those screenshots — we now know from Sutskever's trial testimony — came in significant part from Mira Murati. Sutskever had been building his own document at the board's request. By November 2023, it ran 52 pages. When asked under oath what action he believed was appropriate based on that document, Sutskever's answer was one word: "Termination."
One additional detail that surfaced in the trial record: just a month before the firing, Toner had published an article that criticized OpenAI while praising its competitor, Anthropic. Sutskever found this action to be "not far from obviously inappropriate" for an OpenAI board member, and discussed the prospect of removing Toner from the board with Altman himself. The board that would fire Altman was itself fracturing in the weeks before it acted.
Part 2: The Firing
Friday, November 17, 2023.
OpenAI's board of directors ousted Altman following a deliberative review process. "Mr. Altman's departure follows a deliberative review process by the board, which concluded that he was not consistently candid in his communications with the board, hindering its ability to exercise its responsibilities," OpenAI said. "The board no longer has confidence in his ability to continue leading OpenAI."
The statement was four sentences. No specifics. No examples. No explanation of what "not consistently candid" meant.
Neither Microsoft nor OpenAI's employees were told ahead of time. Greg Brockman — OpenAI's co-founder and president — found out moments before the public announcement and resigned the same day. Other Altman loyalists headed for the exits immediately.
One example of what the board had documented: when OpenAI released ChatGPT in November 2022, the board was not informed in advance and found out about it on Twitter. That detail — which former board member Helen Toner disclosed publicly months after the fact — was confirmed again this week in Tasha McCauley's trial testimony. The board didn't fire Altman impulsively. It fired him over something it had been watching and documenting for at least a year.
What the world saw on November 17 was a four-sentence press release. What the board had was 52 pages.
Part 3: The Merger That Almost Was
Saturday, November 18, 2023. Twenty-four hours after the firing.
This is the chapter that the trial has now made part of the permanent record — and the one that would have changed everything.
On Saturday, November 18, either board member Helen Toner reached out to Anthropic or Anthropic reached out to her, with a proposal for the two AI companies to merge — with Anthropic's leadership taking control of the combined entity. "They reached out with a proposal to be merged with OpenAI and take over its leadership," Sutskever testified, describing calls that included Dario and Daniela Amodei.
Sutskever was asked directly about the other board members' reaction. "They were a lot more supportive, yes," Sutskever said, adding that Helen Toner was the "most supportive" on the board. "I think — at the very least, none were unsupportive."
Sutskever himself was the outlier. "I was very unhappy about it. I really did not want OpenAI to merge with Anthropic," he testified. He offered no elaborate strategic reasoning. He simply didn't want it. His emotional attachment to the company he had helped build — and the care he had described on the stand for its mission — was apparently sufficient.
Anthropic initially expressed "excitement" about the potential deal. The talks collapsed due to what Sutskever described as "practical obstacles." He did not know what those obstacles were.
Helen Toner later disputed Sutskever's account publicly, posting on social media that she was not the one who initiated the board-Anthropic call and disagreed with his recollection that other board members were supportive. The competing accounts are both now part of the public record.
The irony of the proposal is worth sitting with. Anthropic was founded in 2021 by Dario and Daniela Amodei and a group of former OpenAI researchers who left over disagreements about AI safety and corporate direction. The board that fired Altman over safety and governance concerns was, within 24 hours, exploring whether the people who left OpenAI over the same concerns could come back — and take over.
The obstacles were practical. But the logic was coherent.
Part 4: The Reinstatement
The five days between the firing and the reinstatement were the most compressed chaos in the history of Silicon Valley.
Saturday, November 18: The board explored the Anthropic merger. It collapsed. The board simultaneously began backchanneling about bringing Altman back — and rejected the terms.
Sunday, November 19: Microsoft CEO Satya Nadella publicly announced they had hired Altman to lead a new advanced AI research team at Microsoft, alongside Brockman and several other recently departed OpenAI employees. Altman posted "the mission continues." It looked like it was over.
Monday, November 20: The overwhelming majority of OpenAI's 700-plus employees signed an open letter urging the board to resign, saying they would quit if the board members didn't step down and reinstate Altman. Sutskever — who had orchestrated the firing — was among the signatories. He posted publicly: "I deeply regret my participation in the board's actions. I never intended to harm OpenAI. I love everything we've built together and I will do everything I can to reunite the company."
The man who compiled 52 pages of evidence to fire Sam Altman signed a letter demanding his reinstatement three days later. That sentence is the human core of this entire story.
Tuesday, November 21: Just after 10pm, OpenAI released a statement that it had reached a deal in principle for Altman to return as CEO, with a new board chaired by former Salesforce co-CEO Bret Taylor, alongside former US Treasury Secretary Larry Summers and Quora CEO Adam D'Angelo.
November 29: The reinstatement was formally complete. Brockman returned as president. Microsoft received a non-voting observer seat on the new board.
The board that fired Altman was gone. The board that brought him back was composed almost entirely of people he and his allies had approved.
Part 5: The Fallout
For OpenAI
The company that emerged from November 2023 was structurally different from the one that entered it. The nonprofit board — the governance structure that gave OpenAI its founding legitimacy — had been effectively dismantled and replaced with a board far more aligned with the commercial operation it nominally governed.
The WilmerHale investigation, commissioned after the reinstatement and concluded in early 2024, found "a significant breakdown of trust between the prior board and Sam and Greg." It concluded the board acted in good faith but "did not anticipate some of the instability that led afterwards."
The instability it didn't anticipate: 700 employees threatening to walk, Microsoft absorbing Altman publicly within 48 hours, investor pressure destroying the board's ability to hold its position. The governance structure the nonprofit board was designed to protect — independence from commercial pressure — collapsed under exactly the commercial pressure it was designed to resist.
OpenAI is now valued at $852 billion and targeting a Q4 2026 IPO. The nonprofit structure is being converted to a for-profit. The mission that the board acted to protect is now the subject of a federal fraud lawsuit.
For Sam Altman
Altman survived. More than survived — he returned with a board that could not constrain him the way the previous one had tried to. The WilmerHale investigation cleared him of wrongdoing in a summary finding. The original board's members are gone. The company's valuation has grown tenfold since November 2023.
He takes the stand Tuesday. For the first time, under oath, he will have to account for what the Sutskever Dossier documented, what Murati described, what McCauley and Campbell testified to. The reinstatement protected him from the board. The trial record is a different kind of accountability.
For Greg Brockman
Brockman found out about Altman's firing moments before it happened and resigned the same day. He returned with Altman. His own journals — entered into evidence at trial — called the nonprofit commitment "a lie." His $30 billion stake, accumulated on a $0 investment, has been the subject of the most pointed cross-examination of the trial. Sutskever's testimony this week confirmed that a similar memo critiquing Brockman also exists — separate from the 52-page Altman document.
For Ilya Sutskever
Sutskever compiled the evidence. Voted to fire Altman. Signed the letter demanding his reinstatement. Apologized publicly. Left OpenAI in May 2024. Founded Safe Superintelligence Inc. — a company whose name is the clearest possible statement of what he believed OpenAI stopped being.
His stake in OpenAI's for-profit arm is currently worth $7 billion. He is the most consequential and most conflicted figure in this entire story — the man who built the case, pulled the trigger, regretted it immediately, and is now watching his testimony become the documentary foundation of a federal trial.
For the Current OpenAI Board
Bret Taylor, who chairs the board, testified Monday that the November 2023 period was "dire." The board he chairs replaced the board that acted on Sutskever's evidence. That replacement board approved the for-profit conversion the original board was trying to prevent. It is now guiding the company toward the IPO that will make Altman, Brockman, and Sutskever billionaires several times over.
For Anthropic
Anthropic said yes to the conversation. The practical obstacles — almost certainly the complexity of investor agreements on both sides — killed the deal before it could proceed. Anthropic declined to comment on the merger discussions when the deposition became public in late 2025, and has maintained that posture through the trial.
What November 2023 confirmed about Anthropic: the people who left OpenAI over safety concerns were, within 24 hours of Altman's firing, considered viable candidates to take over OpenAI entirely. The board that had documented Altman's behavior trusted the Amodeis enough to have that conversation. That is a different kind of validation than any benchmark or investment round provides.
The Trial Has Revealed a Different Story
The five days in November 2023 were always going to be rewritten eventually. At the time, the official story held: a board overreached, employees revolted, Altman returned, the company moved on.
The trial has replaced that story with a different one. A board that spent a year assembling 52 pages of documented evidence. A firing that was not impulsive but deliberate. A merger conversation with Anthropic that nearly changed the shape of the entire industry. A reinstatement driven not by the board's judgment but by investor and employee pressure that the governance structure was never designed to withstand.
Altman testifies Tuesday. Closing arguments are Thursday. The jury begins deliberations shortly after.
Whatever twelve jurors in Oakland decide — the five days in November now have a complete record. For the first time.
Aaron Rose is a software engineer and technology writer at tech-reader.blog.
Catch up on the latest explainer videos, podcasts, and industry discussions below.
