In fall 2023, OpenAI chief scientist Ilya Sutskever sent three board members secret memos that would upend the most valuable AI company in the world. Roughly 70 pages of Slack messages and HR documents, compiled with help from colleagues, made a direct case against CEO Sam Altman. One memo began with a list: "Sam exhibits a consistent pattern of..." The first item was "Lying." According to the Farrow investigation by Ronan Farrow and Andrew Marantz, Sutskever had grown convinced that Altman misrepresented facts to executives and deceived the board about safety protocols. He told another board member he didn't think Altman should "have his finger on the button" for technology that could alter civilization.
Altman learned his fate while at a Formula 1 race in Las Vegas. The board's public explanation was thin: he "was not consistently candid in his communications." Microsoft, which had sunk $13 billion into OpenAI, got word moments before. CEO Satya Nadella said he was "very stunned." Investor Reid Hoffman started calling around, looking for clear offenses like embezzlement or harassment. "I just found nothing," he told reporters. Altman flew to his $27 million San Francisco mansion and set up what he called a "government-in-exile." He was back in charge within a week after employees threatened mass resignation and investors pushed for his return.
The governance questions run deeper than one CEO's credibility. OpenAI was founded as a nonprofit whose board had a duty to prioritize human safety over company survival. That structure exists because the technology could be genuinely dangerous. The New Yorker report documents OpenAI's apparent shift away from safety, with staff reportedly claiming "existential safety isn't a thing" when someone tried to interview safety researchers. And this fits a longer pattern. In 2017, multiple outlets reported on allegations of toxic culture at Y Combinator during Altman's presidency. He denied the claims and offered a $2 million settlement. The traits that make someone good at building companies might be exactly what you don't want controlling AI that could reshape civilization.