How a dramatic coup against ‘the Steve Jobs of AI’ unravelled in just 36 hours

The knife was wielded in true Silicon Valley fashion: over video call.

On Thursday night, OpenAI’s chief scientist Ilya Sutskever texted Sam Altman with a link to a Google Meet call the following day.

The request was not necessarily unusual. OpenAI’s trendy San Francisco offices are largely empty on Thursday and Friday as staff work from home, so video calls are the norm.

But what happened on the call was out of the ordinary. When Altman, the company’s 38-year-old co-founder and chief executive, joined the meeting at noon, OpenAI’s board swiftly informed him that he was being fired.

Immediately after, Greg Brockman, an Altman ally and the company’s chairman, was separately informed that he had lost his board seat. Within minutes, OpenAI made the news public.

The coup d’etat at America’s hottest company had been carried out in half an hour by a handful of largely unknown individuals. One of the tech industry’s most public figures – the man who made ChatGPT a global phenomenon – had been ousted from the company he helped found and had built into one of the world’s most valuable ventures.

iPhones across Silicon Valley exploded with gossip, filling a vacuum created by OpenAI’s lack of explanation. The company’s Friday statement merely said that Altman had not been “consistently candid in his communications with the board”.

Explanations ranged from rumours that OpenAI had achieved superhuman intelligence – the term for artificial intelligence (AI) that can rival the human mind – to speculation that Altman had been rumbled for coordinating an elaborate scheme to eliminate the board.

Whatever the true cause, within 36 hours the sacking appeared to backfire. By Sunday, Altman looked close to returning, with chief strategy officer Jason Kwon saying the company was “optimistic” about him coming back.

That was until OpenAI announced the hiring of ex-Twitch Emmett Shear as its new chief executive late on Sunday. 

Regardless, pressure will remain on the board, which appears to have underestimated the strength of support Altman had both internally and with investors, and overestimated its own importance to the company.

Altman typically spends his weekends with his partner Oliver Mulherin at a ranch in Napa Valley, where staff tend vineyards and cows. He is prone to long walks, enjoying the digital escape provided by the lack of mobile phone signal on the grounds.

No such respite was available this weekend. OpenAI’s defenestrated chief was barricaded in his San Francisco mansion, liaising with Brockman, shoring up the support of investors and hatching plans to respond, including immediately setting up a rival company before ending up close to victory in a power struggle.

“No one in the world is better than Sam at dealing with this kind of situation,” Paul Graham, the legendary tech investor and mentor to Altman, tweeted.

The saga – rivalling the collapse of Silicon Valley Bank for the industry’s most dramatic weekend of the year – has exposed deep splits at the centre of the company behind ChatGPT, a product that has turned the tech world on its head with its ability to write emails, essays and computer code.

OpenAI was set up in 2015 by Altman, then the boss of the gilded start-up incubator Y Combinator, and Elon Musk, who had fallen out with Google boss Larry Page over the latter’s laissez-faire approach to AI’s risks. (Altman had considered running for California governor, but decided AI was more important).

Musk provided the start-up with $100m (£80m) alongside investors including LinkedIn founder Reid Hoffman and renowned Silicon Valley tech investor Peter Thiel. Altman set about creating a crack team including Brockman and Sutskever, a Russian-born genius who had studied under the British “godfather of AI” Geoff Hinton.

OpenAI was set up as a non-profit, meaning staff could not expect riches. However, they were enticed by the overarching mission of developing “artificial general intelligence”, the point at which machines surpass human-level reasoning. The breakthrough came a couple of years later, when researchers at Google developed a technique called the “transformer” that allowed AI to process and recreate data at greater scale.

While Google failed to exploit its invention, OpenAI did. It developed the technology that several years later would become ChatGPT.

From the start, OpenAI was rocked by infighting. Musk and Altman fell out, leading the Tesla billionaire to pull his funding.

OpenAI suddenly needed cash to operate its increasingly resource-intensive AI systems. However, Altman was reluctant to give up its non-profit aims and raise fresh cash by selling shares. The company came up with a unique solution. It created a “for-profit” subsidiary that would be ultimately governed by the non-profit parent’s board, but could give employees shares and raise money.

Raise money it did. Microsoft invested $1bn in the company in 2019, the same year Altman became chief executive. It has since upped that to $13bn, giving it a rumoured 49pc stake of the for-profit subsidiary. Most of that is believed to be in computing resources, rather than cash. This computing power has been crucial in training ChatGPT to become one of the most advanced AI models in the world.

The wild success of ChatGPT led OpenAI to be valued at $29bn earlier this year and the company has been finalising a new investment that would value it at $86bn – three times that.

OpenAI’s increasingly commercial bent has heightened tensions internally, however. In 2021, a group of employees quit in protest at the new direction, forming AI rival Anthropic.

Even before this weekend’s drama, OpenAI had lost three directors this year: Hoffman, who left to found AI company Inflection; Shivon Zilis, a director at Musk’s brain implant company Neuralink; and Republican former congressman Will Hurd, who dropped out to launch a brief presidential run.

The shake-ups have had two effects: First, Altman has increasingly become the face of OpenAI, appearing in front of Congress and trotting the globe, including playing a central role at Rishi Sunak’s AI Safety Summit at Bletchley Park earlier this month. Altman, who describes Steve Jobs as his idol, has become the closest thing to the AI industry’s equivalent.

Secondly, the thinning of the board’s ranks has left a team of relatively unknown individuals in power. Besides Altman, Brockman and Sutskever, OpenAI’s other directors comprise: Adam D’Angelo, who was an early Facebook executive and the boss of Q&A website Quora; Tasha McCauley, a technology entrepreneur who is married to the actor Joseph Gordon-Levitt, and Helen Toner, an academic at Georgetown University.

“It was a pretty unique setup, the board has been massively outpaced by the scale at which the company got to,” says James Wise, a partner at venture capital firm Balderton.

According to the company’s website, OpenAI’s board is duty-bound to humanity, not investors. What that means in practice is up for debate.

Reports suggest Altman’s growing ambitions were at odds with Sutskever, the chief scientist who had become increasingly focused on risks arising from a commercial influence over AI.

In comparison, Altman, an AI optimist, has focused on growing the technology at pace. He has sought billions from investors including SoftBank for projects such as a hardware device and a plan to develop AI microchips.

On Saturday, Altman tweeted: “i love you all”. The words’ first letters – Ilya – were interpreted by some as outing Sutskever as the mastermind behind his sacking.

Board power is often illusory, however. Microsoft chief executive Satya Nadella, who found out about the sacking only minutes before it was made public, was reportedly livid.

The confusion now throws the ongoing investment round into doubt, with at least one investor said to be delaying wiring money.

Employees, who were set to make fortunes from the now-jeopardised share sale that would value OpenAI at $86bn, led a co-ordinated show of support for Altman, tweeting heart emojis in every colour in response to the ousted founder’s post. Those publicly backing him even included Mira Murati, the company’s chief technology officer who was drafted in to replace Altman.

By Saturday night on the US west coast, OpenAI was in talks over bringing Altman back, a move that may lead to the board itself being replaced.

Microsoft is expected to demand a board seat that it has always been denied and OpenAI’s non-profit ethos looks more threatened than ever.

Wise, of Balderton, says regulators may also take notice. “AI is in the regulatory spotlight and people are rightly concerned about the speed of development. Seeing what are frankly playground shenanigans on the board isn’t going to help those of us who say we don’t want to regulate now.”

It may have been one of the shortest coups in corporate history, but even if OpenAI’s founder is back to work later on Monday, his brief sacking is likely to have lasting effects.