How the bromance between Elon Musk and Sam Altman turned toxic

Altman, now chief executive of artificial-intelligence pioneer OpenAI, was then in his 20s and had recently sold his first startup for a disappointing sum. Musk was 14 years his senior and planning to send rockets to Mars. “I left thinking, ‘huh, so that’s the benchmark for what conviction looks like,’" Altman later wrote.

Recently, Altman got a far more personal taste of his former mentor’s conviction: Musk declared war on Altman himself. In a lawsuit filed last week, Musk accused his former protégé of abandoning OpenAI’s original mission in pursuit of profit, which OpenAI denies. Their intensifying feud exposes the shifting balance of power in the artificial-intelligence arms race.

The two men co-founded OpenAI as a nonprofit research lab in 2015, with tens of millions of dollars of early funding coming from Musk himself. Now, Musk is gone and OpenAI has gone commercial, with billions of dollars of funding from Microsoft. OpenAI is leading the AI boom, while Musk’s own artificial-intelligence startup, xAI, has been left playing catch-up with the organization his money helped get off the ground.

“On a personal level it’s sad," Altman said in a memo to employees the day the lawsuit was filed. He called Musk a personal hero. “I like to think of Elon as a builder and someone who competes by attempting to build better technology, and someone who I’d hope to be on our side," he wrote.

In a blog post Tuesday evening, OpenAI said it intended to move to dismiss all of Musk’s legal claims.

For years, Altman relied on Musk’s fame and fortune to get OpenAI off the ground. The two built an alliance to try to prevent Google, which had taken the lead in the AI race, from dominating the industry. People close to the two men offered conflicting accounts of how their partnership soured.

Those close to Altman say Musk is jealous of how his former mentee upstaged him in the AI race, and that Musk cares more about beating OpenAI than AI safety. Those close to Musk say his concerns about safety are genuine, and that he sees xAI as crucial to developing a better alternative to OpenAI.

This account is based on conversations with dozens of people familiar with Altman, Musk and OpenAI, and on court filings and internal documents.

Altman and Musk were introduced by Geoff Ralston, a partner at Y Combinator, the powerful startup accelerator that had helped fund Altman’s first company, a location-based social-networking app called Loopt.

When Altman first met Musk, he had just sold Loopt for only about what earlier investors had poured into it. He was somewhat adrift, spending time at an ashram in India and mulling whether to start a new company or devote himself fully to investing.

Musk, for his part, had developed a commercial spacecraft capable of picking up cargo from a space station and returning to Earth—which he viewed as an essential first step in making humans an interplanetary species.

Musk was growing worried about rapid advances in artificial intelligence. According to Musk’s lawsuit, his fears grew out of a 2012 meeting with Demis Hassabis, co-founder of AI company DeepMind, who “emphasized the potential dangers that the advancement of AI presents to society." Unmentioned in the lawsuit was that Musk invested in DeepMind after that meeting, to keep better tabs on the technology.

Later, Musk was alarmed to hear that Google planned to buy DeepMind. Together with Luke Nosek—like Musk, a PayPal co-founder—he made a counteroffer, according to the complaint. Their bid failed, and Google bought DeepMind.

Altman had been obsessed with the idea of AI since childhood, and at age 18 had placed it at the top of a list of problems he wanted to explore.

As Altman’s profile in Silicon Valley rose, he tried to focus tech-industry attention on AI’s potential. In 2014, on his personal blog, he referred to AI as potentially “the biggest development in technology ever." Days later, he was appointed to lead Y Combinator, the startup accelerator that had funded Dropbox and Airbnb—and later, OpenAI itself.

Like Musk, Altman also worried about the technology’s dangers. In February 2015, he wrote that AI was “probably the greatest threat to the continued existence of humanity."

Musk and Altman had kept in touch about such concerns. That March, Altman reached out to Musk to gauge his interest in drafting an open letter to the U.S. government about AI. In May, he emailed Musk, proposing that Y Combinator start a “Manhattan Project" for artificial intelligence. Musk responded: “probably worth a conversation."

The two men began working on a new AI lab, which Musk would name OpenAI. Altman proposed in an email that June that the two of them sit on a five-member board that would govern the nonprofit. He suggested waiting to send the open letter calling for AI regulation until after the lab was formally launched. Musk replied: “Agree on all."

Altman brought in Greg Brockman, a computer scientist who served as chief technology officer of payment-processing company Stripe. Musk helped recruit Ilya Sutskever, a leading AI scientist who was working at Google, according to the complaint. To make compensation at the nonprofit more attractive, OpenAI planned to offer recruits equity in Tesla and SpaceX, and a chance to benefit from Y Combinator investments.

Musk and Altman became OpenAI’s first co-chairs. Behind the scenes, it was Musk who was seen to hold more influence and control over the organization, according to former employees.

At first, Musk was a regular presence in the office, pitching far-fetched ideas and conducting polls on when employees thought artificial general intelligence could be achieved. OpenAI shared office space with Neuralink, Musk’s brain implant startup.

Musk also was OpenAI’s financial linchpin. He pushed the nonprofit to announce it had secured $1 billion in funding commitments to “avoid sounding hopeless relative to what Google or Facebook are spending" and promised to cover any funding shortfalls, according to an email from Musk included in OpenAI’s Tuesday blog post. Altman and Brockman had earlier planned to raise $100 million, the post said.

Musk ended up donating $44 million, according to the complaint. He gave $15 million in 2016 and $20 million in 2017, making him the largest funder for both years. He also paid OpenAI’s rent for a number of years, according to the complaint.

Musk pushed OpenAI researchers to come up with projects to give it an edge over DeepMind. One research team set out to beat the world’s best players at a complex videogame called Dota 2.

By 2017, though, OpenAI hadn’t achieved any major research breakthroughs. Musk appeared to be growing restless, former employees said. He ratcheted up the pressure on employees, sometimes threatening to walk away from the project.

That same year, Google published a paper about a new AI model called the Transformer, which would pave the way for the kind of large language models that power tools such as humanlike chatbots. The paper showed that the way to achieve that was by crunching huge data troves, which required enormous computing power.

To pay for such computing power, Brockman and others suggested changing OpenAI’s structure to a for-profit, according to the complaint. That would enable it to raise capital from investors such as Microsoft.

Musk resisted the idea, writing to Brockman, Sutskever and Altman that either “go do something on your own or continue with OpenAI as a nonprofit. I will no longer fund OpenAI until you have made a firm commitment to stay or I’m just being a fool who is essentially providing free funding to a startup. Discussions are over."

Altman replied that he remained “enthusiastic about the non-profit structure!"

In the Tuesday blog post, OpenAI said Musk recognized the need for a for-profit entity, but wanted majority equity, initial board control and to be CEO. In the middle of the discussions, the post said, he withheld funding. OpenAI relied on another investor, LinkedIn co-founder Reid Hoffman, to pay its bills, the post said. Hoffman didn’t respond to a request for comment.

Musk pushed for more control over OpenAI, at one point recommending that OpenAI become part of his car company, Tesla. In an email attached to the blog post, Musk wrote: “Tesla is the only path that could even hope to hold a candle to Google. Even then, the probability of being a counterweight to Google is small. It just isn’t zero."

Musk began trying to coax OpenAI researchers to jump ship and join Tesla, irking his colleagues. By February 2018, OpenAI executives had rejected his bid to take control, and Musk decided to step down as co-chair. Altman became chief executive.

Altman called an all-hands meeting, attended by Musk. He thanked Musk for his time at the organization. Musk told employees he intended to pursue his own AI research at Tesla. After a young researcher challenged Musk’s decision, suggesting it would exacerbate the AI arms race, he called the researcher a “jackass" and stormed out of the building.

In late 2018, Musk sent another email to Altman, Brockman and Sutskever predicting OpenAI’s eventual failure. “My probability assessment of OpenAI being relevant to DeepMind/Google without a dramatic change in execution and resources is 0%," Musk wrote, according to the blog post.

Altman and Musk continued to discuss AI, and while Musk had stopped making cash donations to OpenAI, he kept paying its rent. Altman continued to support Musk publicly after launching OpenAI’s for-profit subsidiary in March 2019.

When investors began shorting Tesla’s stock, he rushed to Musk’s defense. “It’s gross seeing so many root against Tesla," he tweeted in May. He added that “betting against Elon is historically a mistake."

Their relationship took a hit in November 2022 when OpenAI released ChatGPT, a chatbot that could write poetry and generate computer code.

To Altman’s irritation, Musk publicly aired concern that the chatbot had accelerated a dangerous race to develop powerful AI. Musk also began questioning via tweets how the nonprofit he co-founded had launched a for-profit entity that raised billions of dollars from Microsoft. Those complaints became the basis of his lawsuit.

Shortly after OpenAI released ChatGPT, Musk announced he had severed OpenAI’s access to a pipeline of data from Twitter, the social-media platform he had recently acquired. OpenAI had been considering using Twitter’s data to train its models.

Altman invited Musk to OpenAI’s headquarters. The two men held a long, closed-door meeting about the Twitter decision and ChatGPT.

Around that time, Musk told Altman about his plans to start a competing artificial general intelligence company, xAI. Altman questioned whether adding another entrant into the AI race would ease Musk’s concerns. In the months after ChatGPT’s release, Musk began trying to poach OpenAI employees for xAI. He also threatened to sue Altman and OpenAI.

In November, Musk announced his own chatbot, Grok. He described it as a less “woke" competitor to ChatGPT. Musk gave the new chatbot access to Twitter’s trove of online data, as well as space in its San Francisco office. In recent weeks, xAI has begun preparing for a new fundraising round, a move likely to intensify its race against OpenAI.

After Musk filed his lawsuit, Altman wrote a memo to his staff: “The implication that benefiting humanity is somehow at odds with building a business is confusing," he said. “I miss the old Elon."

Write to Berber Jin at berber.jin@wsj.com, Keach Hagey at Keach.Hagey@wsj.com and Deepa Seetharaman at deepa.seetharaman@wsj.com