Elon Musk had a head-spinning week in AI.
After agreeing with bitter rivals Dario Amodei and Sam Altman in their call for foundation model labs to slow the pace of development, Musk was contradicted by President Donald Trump and Nvidia CEO Jensen Huang, who said AI companies actually need to go faster.
Trump, whose stock portfolio includes shares of Nvidia and SpaceX according to his most recent disclosure, called fears of runaway AI a "hoax."
Speaking on a panel at the All-In Summit on Tuesday that included longtime friend and former White House AI czar David Sacks, Musk took a more middle-of-the-road approach. He urged companies to test each other's new models to identify and fix safety issues before they get released, an option he sees as preferable to the government taking a heavy-handed regulatory position in AI.
"You can always escalate the amount of regulatory oversight, but it is very difficult to reduce it," Musk said in a virtual appearance at the event. "It does tend to be very much a one-way ratchet."
Behind the scenes, the Wall Street Journal reported this week that Musk had joined Huang and Meta CEO Mark Zuckerberg in advising President Trump to oppose an industry-funded regulator.
Musk didn't immediately respond to a request for comment for this story.
The back and forth in just the past week is a microcosm of Musk's history in artificial intelligence, dating back to an early check he wrote to fund DeepMind, which Google acquired in 2014. The year after that deal, Musk co-founded OpenAI with Altman and others as a nonprofit research lab.
While Musk has long expressed controversial and often conflicting views in AI, torching friendships along the way with the likes of Google co-founder Larry Page and Altman, the stakes now are higher than ever. Model developers are rolling out technology that has industry insiders sounding the alarm about the potentially existential threats of AI, while internet giants spend hundreds of billions of dollars a year on the infrastructure to power it.
Musk's SpaceX is heavily invested in the technology's growth. The company merged with xAI, Musk's AI startup, in February, and completed a $60 billion acquisition of AI coding tools provider Cursor in August. In between those two events, SpaceX went public in the largest IPO on record, briefly turning Musk into a trillionaire.
Meanwhile, Musk's other public company, Tesla, continues work on so-called physical AI, including driverless cars and humanoid robots.
"Tesla is arguably the world's biggest robotics company," Musk said at Tesla's "AI Day" in 2021. "Because our cars are semi-sentient robots on wheels."
'Some scary outcomes'
Since Musk's early days in AI, he's simultaneously drummed up public fears of world-ending threats from AI "superintelligence" while building out his own AI businesses and fighting off regulation. He told CNBC's "Closing Bell" in 2014, "I like to just keep an eye on what's going on with artificial intelligence," adding, "There have been movies about this, you know, like '[The] Terminator.' There are some scary outcomes."
His fear-plus-optimism campaign is part of a strategy documented by AI ethics and computer science researchers Timnit Gebru and Emile Torres in a March 2024 paper.
The paper mentions Musk more than 30 times and concludes that he and a small group of elite players in AI have claimed they are, "safeguarding humanity's future, while simultaneously creating unsafe products, centralizing power, and evading accountability," and disproportionately harming marginalized groups.
In late 2014, Musk told students at the Massachusetts Institute of Technology that AI development was like "summoning the demon." Around that time, he donated $10 million through his foundation to seed the Future of Life Institute (FLI), a nonprofit led by MIT researcher Max Tegmark, who aimed to raise awareness of AI's terrifying potential and to steer the technology in a manner that would avoid driving humans to extinction.
Speaking at a keynote at the 2015 Effective Altruism conference, Musk said the threat of AI advancements spinning out of human control was the biggest existential risk that he could envision. He soon signed an open letter from FLI calling to "pause AI."
That same year, Musk co-founded OpenAI with a mission to "advance digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return."
His rift with Altman came in 2018, when Musk privately pressured his OpenAI co-founders to give him control of the lab and to make OpenAI a for-profit subsidiary of Tesla. Musk poached key talent from OpenAI for Tesla, and resigned from the organization's board that year.
It was another four years before the magnitude of that breakup would start to become apparent. In November 2022, OpenAI released ChatGPT, kicking off the generative AI boom, with Musk largely on the sidelines.
A few months later, in March 2023, Musk signed another "pause AI letter" from FLI.
"AI systems with human-competitive intelligence can pose profound risks to society and humanity, as shown by extensive research and acknowledged by top AI labs," the letter began.
At roughly the same time, Musk was incorporating his new AI venture, xAI, in Nevada, but he kept it quiet from Tesla investors and the public until July.
Musk v. Altman
Then came the legal battle with Altman.
In August 2024, Musk filed a suit against OpenAI and its co-founders, claiming they had violated a promise to keep the lab a nonprofit. Musk lost the high-profile California trial in May, when a jury and federal judge determined his claims were filed outside of the statute of limitations. Musk said he would appeal but has yet to do so.
Correspondence brought to light in the case showed that Musk's obsession with creating OpenAI was built on fears that Google would control the technology after acquiring DeepMind, though in public he and Altman touted more altruistic aspects of their mission.
Now, Musk is back to calling for safety. Following Amodei's essay last weekend imploring the industry to "pace the frontier," Musk posted on X, "Dario is right."
That's quite the contrast from February, when Musk called Anthropic "misanthropic and evil" and said it "hates Western civilization." Things changed dramatically in May, when Anthropic agreed to pay SpaceX up to $1.25 billion a month to rent computing infrastructure in Memphis, giving Musk's company a hefty dose of sorely needed AI revenue before SpaceX's IPO.
"I was clearly wrong about Anthropic," Musk wrote on X in July. "They are obviously currently the leader in AI."
Musk's recent track record on AI safety has been rocky.
He's built out data centers in and around Memphis, Tennessee, while fighting off litigation from residents nearby suffering from air and noise pollution from the operations.
At SpaceX, Musk sold a version of xAI's Grok models to the U.S. military which "enabled U.S. forces to deploy over 2,000 munitions to 2,000 distinct targets within 96 hours," according to DOJ legal filings.
And at xAI — now known as SpaceXAI — Musk sued to stop state regulations that would ban Grok's "nudify" features in Minnesota, and that would require his company to be transparent about the data used to train its models in California.
As Musk and his peers and rivals debate the right approach to AI safety, and whether the government should have a big role or no role at all, lawmakers and third-party researchers are growing increasingly skeptical of the industry's ability to police itself.
Tyler Whitmer, CEO of the nonprofit Legal Advocates for Safe Science and Technology (LASST), told CNBC that self-regulation won't be enough to protect the public from existing and future harms enabled by AI systems.
"We need real regulation that creates a level playing field for companies but also protects the public, and the companies just can't do that themselves," Whitmer said. "There needs to be a government-enforceable requirement that third-party evaluators be involved, not just the companies themselves."
