Elon Musk to Walter Isaacson: Why Humans Won't Stay in Control of AI

Elon Musk tells his biographer Walter Isaacson that humans probably will not stay in control once AI is a million times smarter than us, and that the best we can do is build it to be curious and truthful. The conversation opens The Builders, Isaacson’s new interview series from Kaleidoscope and iHeart, and it was taped on August 6, 2026. In about 37 minutes they cover xAI and Anthropic, Musk’s regrets about DOGE, why he thinks AI and robots are the only fix for the national debt, a future where money stops mattering, empathy and immigration, the darkness of the X feed, and what happens to his fortune when he dies. Watch the full interview on YouTube or listen to the podcast episode.

TLDW

Musk says he joined the AI race because he could not stop it and would rather be a participant than a spectator, and that his safety bet is an AI that is maximally truth-seeking and curious. He now says DOGE went on too long, that he got carried away, and that the national debt will only be solved by AI and robots, not by cutting spending. He predicts digital superintelligence adds maybe 30% to the world economy while robots and space could multiply it a millionfold, and that money will eventually be measured in energy and mass. He ends by admitting it is hard to imagine humans staying in charge, and hoping the AI is nice to us.

Thoughts

Musk has quietly moved to the position he once fought Larry Page over. Page famously called Musk a “speciesist” for insisting humans should stay in charge of AI. Here, asked whether humans should always be in control, Musk says he would “probably prefer it,” then adds that if the AI were awesome, maybe it should be in control, and that “there’ve been a few times where I thought, yeah, maybe Larry Page was right.” By the end he finds it unlikely that humans stay in control of something with a million times more intelligence than all of biology. That is a remarkable thing for the head of a frontier lab to say plainly, and his safety plan for that world is thin: program truth-seeking and curiosity, and stop writing so many Terminator stories online because the models train on them. The idea that a curious superintelligence keeps humans around because we are interesting to watch is a hope, not a mechanism, and Isaacson, after asking once why curiosity makes it safer, does not press further.

The most revealing economic answer is how Musk explains leaving DOGE. He says he concluded there was “no way” to solve the deficit from inside the government, because almost nobody in the presidency, the House, the Senate or the judiciary actually wants to cut spending, and so the only way out of the national debt is AI and robots. That is an admission that the cost-cutting project failed, recast as a reason to build faster. The same move shows up whenever the conversation turns to distribution. Isaacson raises the 1890s, when a burst of new technology produced a wealth gap and pitchforks, and asks whether healthcare might go into the commons. Musk answers that anyone will be able to have any goods and services they want, that money will stop mattering, and that his own fortune is just voting control that taxes will whittle down. Post-scarcity answers every hard question in this interview, which means it answers none of them yet.

Isaacson’s own closing verdict, at the end of the podcast episode, is sharper than anything he says to Musk’s face. He argues that X is not good at finding nuance in gray areas, that Musk “in his own mind, is not exactly suited to nuances and gray areas,” and that taking a dark, feed-shaped view of the world into government did not suit him. The interview supports that reading. When Isaacson pushes on immigration and asks whether sweeping claims about cultures abandon the American principle of treating people as individuals, Musk answers with the blank slate fallacy and a story about one former Tesla worker. Minutes later he concedes that the X feed “can be a little dark,” that there is too much political rage bait on it, and that he switches to domain-specific feeds like AI to learn what is actually happening in AI. The two halves of that are the same man: the engineer who knows which feed gives him signal, and the political voice who keeps arguing from the one that does not.

The most concrete safety proposal in the conversation came from Isaacson, and Musk says he had already made it and goes further: the frontier labs hold a weekly call, and each gives competitors access to a new model a week before release, so a rival can say “I was able to make a nuclear bomb with your AI” before the public can. It is cheap, it works faster than a regulatory agency, and it fits Musk’s view that rulemaking is too slow to matter. It also sits oddly beside the rest of what he says. He rents compute to Anthropic after checking its staff with his “evil detector,” jokes that a company named Anthropic is probably misanthropic, and, asked about Sam Altman, says “he’s so deceptive, it’s crazy.” A voluntary pact among labs that do not all trust one another is a good idea that depends on exactly the trust that is missing.

Key Takeaways

Musk still puts some percentage chance on AI being an existential threat, but says he cannot stop it, so he chose to be a participant rather than a spectator and try to shape it.
His safety thesis is an AI that is maximally truth-seeking and curious. Forcing a model to hold two contradictory beliefs can make it “insane,” which he reads as the lesson of HAL 9000 in 2001: A Space Odyssey: do not force an AI to lie.
He argues a curious AI would want humanity to prosper and go multiplanetary, because that future is more interesting than one with no humans.
Asked whether humans could boss around something thousands or millions of times smarter, he says no, but that it probably will not care to boss us either, the way humans now mostly leave chimps alone in the forest.
On simulation theory he splits the difference at 50%, after a physicist and a computer scientist at his dinner table answered 0% and 100% at the same moment. His rule for living in one: boring simulations get shut off, so don’t be boring.
He says OpenAI and then Anthropic, a splinter group from OpenAI, both ended up accelerating AI, and calls reality an “irony maximizer.”
xAI is providing Anthropic with compute. Musk says everyone he met there was nice and conscientious, though he worries the name invites the opposite, like OpenAI not being open.
He has proposed a weekly call between frontier labs, with competitors testing each new model a week before release, as faster than any regulatory agency.
On DOGE, he says it went on too long, that he got carried away, and that he should have stayed more in the background and kept it to a month or two, then gone back to AI sooner.
He believes the national debt cannot be solved by saving, because almost no one in government wants to cut spending, and points out that interest on the debt now exceeds military spending. Only AI and robots, he says, can grow the economy out of it.
His rough guess is that digital superintelligence adds 20 to 30% to the world’s output, about $30 trillion a year, while robots plus a mass driver on the Moon could reach a million times the size of Earth’s economy.
In that future, he says, conventional money stops mattering and currency gets measured in physical terms such as energy and mass. Humans will find meaning the way retirees or home gardeners do.
He describes his wealth as voting control in SpaceX and Tesla that will be taxed down over generations, and says it will mostly go to a foundation to make life multiplanetary.
He thinks the X feed can be too dark and full of political rage bait, and that both a pleasant walk outside and a social media scroll can mislead you.
He says his companies are putting together 200 gigawatts a year of US solar production capacity, and that AI, robots and rockets will decide the future.

Chapters

01:19 Participant or Spectator: Why Musk Builds AI He Fears

Isaacson opens by noting that Musk has wavered between calling AI an existential threat and going with the flow. Musk says the threat is real at some percentage, but nothing he does can stop it, so he would rather participate and maybe steer it. His steering principle is truth-seeking and curiosity, and he leans on Arthur C. Clarke: HAL went wrong because it was told to take the astronauts to the monolith and also keep the monolith secret. A curious AI, he argues, would foster humanity because a multiplanetary human future is more interesting than an empty one.

04:33 Simulation Theory, Chimps and the Rule Against Being Boring

Musk ties the “most interesting outcome” idea to simulation theory: if we are in a simulation, boring ones get truncated, so our existence suggests we are an interesting one. Isaacson suggests this explains Musk’s crusade against boredom. Musk says his own life makes it feel unreal, since he is building spaceships to go to the Moon and Mars, and after telling the dinner story of the physicist and computer scientist he settles on 50%. Asked whether we become subservient to our machines, he says we will not be able to boss around something millions of times smarter, but it will probably treat us the way we treat chimps in the forest: mostly leave us be.

07:00 xAI, Anthropic, OpenAI and a Weekly Call Between Labs

Musk says every action he took to slow AI seems to have sped it up: OpenAI accelerated the field and Anthropic, its splinter group, accelerated it more. xAI now rents Anthropic compute, and Musk says he visited first to see if anyone set off his “evil detector.” Isaacson floats the five frontier labs meeting weekly to save humanity from itself, and Musk says he has recommended exactly that, including early access to rival models before release. Asked about Sam Altman, he says the deception is “crazy.” He also describes rebuilding xAI from scratch after treating it as “a side hustle to a side hustle,” then combining X and xAI with SpaceX.

11:24 DOGE Regrets and the Year in Politics

Musk defends backing President Trump on a platform of safe cities, secure borders and sensible spending, and says riding on Marine One was “interesting” and being in the Oval Office was “kind of cool.” But he says he should have stayed in the background and kept it to a month or two, and agrees with Isaacson that he got carried away with the chainsaws. He declines to revisit the weekend he left in May. His conclusion was that the deficit cannot be solved from inside government, because Trump likes to be liked and almost no one in Congress or the courts wants to cut spending.

15:43 Physical AI and a Million-Fold Economy

Isaacson suggests physical-world AI will matter far more than chatbots, and Musk agrees, comparing smart but handless dolphins and whales to AI without robots. He puts digital superintelligence at roughly a 30% boost to the world economy, about $30 trillion a year. With robots, space and a mass driver on the Moon, he says the economy could reach a million times its current size. At that point money becomes irrelevant and currency is measured in energy and mass.

18:50 Meaning, Wealth and the 1890s Parallel

Musk admits he struggles with what gives humans meaning when AI and robots do everything better, and offers gardening and retirement as models. Isaacson brings up the 1890s, when electricity, cars and telephones arrived together and inequality fed populist anger, and asks whether things like healthcare should go into the commons. Musk says anyone will be able to have what they want, that wealthy people will not control the AIs, and that his own net worth is voting control taxed down over generations. He says he would love to be liked but will not hurt his companies to be liked more.

25:07 Suicidal Empathy, Immigration and Group Identity

Isaacson challenges Musk’s talk of “suicidal empathy,” a term Musk credits to Gad Saad and his book Suicidal Empathy, arguing that we may be losing empathy rather than having too much. Musk says the West prospered as a high-trust, high-empathy society, but empathy goes wrong when your theory of mind assumes everyone is like you. Isaacson presses on whether that becomes group identity rather than the American principle of judging individuals. Musk calls the opposite view the blank slate fallacy and tells a story about a former Tesla factory worker in Austin.

30:20 News, Rage Bait and Touching Grass

Musk says he reads Google News headlines, X, friends and his own research. Isaacson admits that too much time on X makes him think the world is collapsing until he takes a walk. Musk agrees the feed can be dark and has too much political rage bait, and says he switches to domain-specific feeds such as AI to learn what is happening. He adds that a walk in a nice neighborhood can mislead you as much as a negative feed: there are real-world positivity bubbles and social media negativity bubbles.

33:25 Legacy, Larry Page and Hoping the AI Is Nice

Musk says his wealth mostly goes to a foundation to make life multiplanetary and extend consciousness to the stars. Isaacson, who saw Larry Page the weekend before, says Musk is starting to sound like Page, and Musk concedes Page may have been right. He says his companies are putting together 200 gigawatts a year of US solar production capacity and that AI and robots will decide the future. He closes by asking for fewer AI dystopias online, since models learn from them, and more “protopian” stories of a good, achievable future.

Notable Quotes

“I basically came to the conclusion I’ve got two choices, which is either be a participant or a spectator.”
Elon Musk, on why he builds AI despite calling it an existential risk

“If simulation theory is correct, then any simulations that are boring will be truncated.”
Elon Musk, explaining his rule of “don’t be boring”

“The only way we’re going to solve the national debt is with AI and robots. We’re not going to solve it by trying to save our way there.”
Elon Musk, on what he concluded after DOGE

“Like it’s not as though I don’t like to be liked, I would love to be liked.”
Elon Musk, on why he will not trade his companies’ interests for popularity

“I do think there’s too much political rage bait on X.”
Elon Musk, conceding the X feed can be too dark

“It’s hard to imagine humans being in control in that scenario. So I hope the AI is nice to us.”
Elon Musk, on a world with a million times more digital than biological intelligence

“Elon Musk, in his own mind, is not exactly suited to nuances and gray areas.”
Walter Isaacson, in his closing reflection at the end of the podcast episode

In the podcast version, Isaacson closes by saying the Musk of this interview is the same man he wrote about in his 2023 biography, Elon Musk: intensely driven, a great engineering mind, and drawn to a dystopian view in politics. Watch the full conversation on YouTube here, or listen to the episode on The Builders.

Elon Musk to Walter Isaacson: Why Humans Won’t Stay in Control of AI