OpenAI would likely have unravelled if Sam Altman hadn't returned as chief executive following his brief ouster in 2023, former tech chief Mira Murati said, offering her clearest account yet of one of Silicon Valley's most dramatic boardroom battles.
Murati, who later left OpenAI to found Thinking Machines Lab, was one of a key cohort who helped reinstate Altman after his shock firing nearly three years ago. The episode thrust Murati into the spotlight after she was briefly named interim CEO. When she quickly set about trying to bring Altman back, the board replaced her with former Twitch chief Emmett Shear.
Employee Revolt and Reinstatement
More than 700 of OpenAI's roughly 770 employees then signed a letter threatening to quit and follow Altman to Microsoft Corp. – and barely five days after he was fired, Altman was reinstated with a reconstituted board.
“When I realized that their decision was potentially catastrophic for the company — that things would potentially fall apart — I felt like I had to act very quickly,” Murati said. If she hadn't, “quite likely, OpenAI would have imploded,” Murati said at the Bloomberg Tech conference in San Francisco.
Testimony in Musk Lawsuit
Murati's remarks echoed her testimony in Elon Musk's lawsuit against OpenAI in April, where she said the company was at risk of falling apart after Altman's removal. She also testified she had concerns about Altman's leadership. Former OpenAI board member Helen Toner offered a more complicated portrait of Murati's role in a deposition, saying Murati “was waiting to see which way the wind would blow, and she didn't realize she was the wind.”
“Even though, on the surface, it looked very chaotic, at each point in time I felt very clear about what I had to do,” Murati said Thursday. “To provide continuity and stability, to help restore that, get the team in place to then deliver.”
New Vision for AI
Murati, who left OpenAI in late 2024, is now making the case for a different vision of artificial intelligence, one centred on what she calls “interaction models.” Thinking Machines is building those systems as companies like OpenAI, Anthropic PBC and Alphabet Inc.'s Google invest billions of dollars in increasingly powerful AI models. The startup raised US$2 billion at a US$12 billion valuation last year, and was said to have been in discussions about a bigger, subsequent round. Several employees have since left her company.
Thinking Machines has released a developer tool called Tinker, but has yet to launch a flagship consumer product.
“The types of models we work with today, they're very turn-based,” Murati said. “You talk, they talk, then they go off and think once you've given a prompt on what it is you want to do.”
“Our interactions with each other are very rich,” she said. “There's a lot of information in our interactions – when we're silent, when we're thinking, when we're interrupting one another. Interaction models are able to capture all of these nuances.”
