"If you don't win, your morality means nothing." Eric Weinstein said that is what he told Sam Altman, OpenAI's chief executive, when Altman "was trying to figure out how to be good and how to be decent to the mathematicians and the scientists." Weinstein added that he would say the same to Dario Amodei, who runs the rival lab Anthropic.
Weinstein is a mathematician, physicist and economist with a Harvard PhD and a former managing director of Thiel Capital. He made the argument in a long conversation with Steven Bartlett on The Diary of a CEO, published October 12, 2026. The episode opens with a teaser of the exchange. Bartlett reminds him that AI could lead to human extinction. Weinstein replies that the lab leaders "have to win in a ruthless competition," and says he is "going to regret saying this." He ends the clip with: "Let us not sit around like our hands are clean."
The position sounds contradictory. Weinstein treats advanced AI as an enormous danger, yet he urges the people building it to win the race to superintelligence, meaning AI that outperforms humans across the board. He spent much of the conversation explaining how he holds both views and where he places himself in the story. Everything he said about his private conversations with Altman and Amodei is his own account. He said he would generally repeat only his half of those exchanges.
Why winning comes first
Weinstein's argument rests on who gets to shape the technology. "If you have strong views about the morality of superintelligence, do you have any ability to constrain it?" he asked. When Bartlett pointed to the influence of his podcast, Weinstein brushed it aside: "You, Stephen Bartlett, are losing the AI race, as am I." What either of them thinks, he said, matters only to the extent that it influences "the people at the frontier," meaning the handful of labs building the most advanced systems.
He called his advice to Altman "paradoxical": he has to win "almost at all costs." Weinstein described the alternative and his doubts about it. Rivals could try to win collectively through universal participation in a "pace-the-frontier" agreement, in which participants agree to pace advances in AI capability together. But such an agreement is open to cheating, he said, because "it's very hard to monitor who's doing linear algebra on GPUs," a reference to the mathematics and graphics chips that AI training runs on. He stressed that he was taking a moral position, not a financial or geopolitical one. Anyone who believes they are the most decent player in AI, he said, will want to win "so that you get to make the decisions at the end."
The verification problem he described also appears in proposals from people who favor slowing down. In a September 2026 essay, Amodei proposed outside evaluators, shared standards among democracies and eventually international pacing agreements. He treated verification and the incentive to defect as central obstacles. A July 2026 employee statement called Pacing the Frontier, whose listed signatories include Amodei, describes the same collective-action trap: firms and countries may want time for security and oversight but face pressure not to slow down alone.
Weinstein also argued that the race cannot simply be called off. "This is going to happen because it's linear algebra," he said. "You can't stop linear algebra."
Bartlett pushes back
Bartlett resisted the argument repeatedly. After Weinstein said "almost at all costs," Bartlett read out lines from an unnamed "he": that AI would most likely lead to the end of the world but would create great companies in the meantime, and that "the worst case is lights out for all of us." Then he asked again: "At all costs?" Weinstein answered that if humanity must avoid self-termination, "which I believe deeply," then "we are in a race with other people" and "we have to win that race."
The two also disagreed about containment. Weinstein said that if someone were "really serious about air gapping," meaning running a system on a computer with no connection to anything else and closing off the indirect escape routes, it would be safe. Bartlett countered with an analogy. These walls probably could not hold the strongest man in the world, he said, and a jail built by a lesser intelligence could not hold a greater one, "like how my dog Pablo couldn't contain me in anything that it created."
Weinstein accepted that the risk exists but said the labs had already taken it on. If you ask the system to do research, he said, "that research that it does might give it new means of escape." He compared the situation to a letter Edward Teller wrote to Leo Szilard at the close of the Manhattan Project, which he summarized as Teller saying the genie could not be contained once it had been helped out of the bottle. In Weinstein's view, the companies made that choice when they released large language models, which he called "military-grade malfunctioning experiment[s]," to the public as revenue-producing products. "I wouldn't have done it, but you did that," he said.
The reproduced letter is dated July 1945. It is Teller's reply to Szilard's request that he join objections to the bomb's immediate military use. In it Teller expresses moral unease, doubts that banning a single weapon could prevent its use in a future war, and argues for making the broad facts public so that responsibility would extend beyond the scientists who built it.
At one point Weinstein said he has been working on something for 42 years, that he is right and that others are wrong, and that this is his moment. Bartlett replied that he "might be proven right and then extinction," and asked whether being proven right was worth the chance of extinction. Weinstein answered with another private exchange. He said he had told Altman that what each of them was doing was incredibly dangerous, and that Altman replied: "What you're doing, what I'm doing should be done in a military base, but that's not how things are unfolding."
Why he stayed out of OpenAI
Weinstein said he was once in discussions to join OpenAI. When Altman later asked why he had not made a proposal, Weinstein said he answered, "I think you're going to be forced to be ruthless and I don't want to be ruthless." Asked to define ruthless, he said "taking every advantage. Playing as hard as you can," and reached for basketball: "if you're in the NBA, sometimes you have to throw an elbow into the other guy's eye socket." He did not, he said, want to be in the all-star game that much. "Somebody has to do that, but I don't want to be that guy."
He was frank about why the subject does not feel good. He called the East Bay of San Francisco "cult city" and said "there are genuine cults in the East Bay around machines." He then linked Anthropic and effective altruism, a philanthropic movement influential in AI circles, to "this idea of the ethical ends justify the unethical means," without finishing the thought. Even so, he said of Amodei: "I think he's genuine," in a weird way "a super ethical human being."
Pressed to compare the two leaders, Weinstein said he was "more confident of Sam," while noting that he knows the two men to different degrees. He called it "really dicey stuff because I'm talking about people I know." The Altman he knew before he became famous, Weinstein said, "was absolutely concerned about humanity before it made sense to be concerned," because he thought AI "was going to destroy the labor market." Earlier in the conversation, Weinstein recalled Altman coming to his house to talk about universal basic income (UBI) before Altman was widely known, and running experiments with it. He also described a courtesy call in which Altman told him "this is just coming faster than anyone imagines." Bartlett's response to the labor-market point was short: "So he built it."
Jessup on the wall
To explain how he sees the lab leaders, Weinstein turned to the film A Few Good Men, written by Aaron Sorkin, which he said is "not well understood by all of us who quote it." Audiences treat Jack Nicholson's Colonel Jessup as the villain and Tom Cruise's character as the hero. Weinstein reads it as a story about "dysfunctional characters" whose combined actions produce a moderately decent outcome, "a story about America."
In his reading, Jessup's "you want me on that wall, you need me on that wall" speech recognizes that some people have to make the most gruesome decisions. He compared them to emergency-room doctors doing triage after a building collapse, or people making decisions in war. "I view Sam and Dario as in that position," he said. "They're like generals." Their souls, he added, "are going to be scarred for sure." He said he did not want that role for himself, "selfishly," and that he was effectively telling Jessup, "I need you on that wall."
In Weinstein's telling, Jessup goes wrong when he overreaches, claiming that he himself provides the blanket of freedom, and that is how he ends up admitting to the code red. He compared this to Moses, who in his telling is punished for saying "I will bring you water from the stone" when he was only meant to be God's instrument. The lesson he drew: "Sam is a steward. Dario is a steward." They will make fortunes and own "third, fourth, fifth, sixth homes," he said, "but that's not the main story. The main story is that these people are forced to make decisions that no one should ever be forced to make."
Paying to summon the demon
Bartlett said one of them would win and "theoretically summon the demon," a phrase he attributed to Elon Musk. Weinstein turned the point back on him. He asked whether Bartlett had a subscription to any AI service. Bartlett said yes. "Okay, so you're paying to have the demon summoned. I am paying to have the demon summoned."
Bartlett objected that he was paying to have the punctuation in an essay fixed or to make a photo. Weinstein widened the charge. He asked how Bartlett's lithium-ion battery was made and where the oil that drives generators comes from. Then he asked whether Bartlett was a vegetarian and whether he ordered lamb. "You're not clean. I'm not clean," Weinstein said. "My claim is you and I are complicit in a minor way in the game in which Sam is complicit in a major way." What bothered him, he said, was that they were not having an authentic conversation.
Would he stop it?
Bartlett asked the obvious question: if Weinstein knows the risk, why not tell Altman to stop? "And have Dario win?" Weinstein replied. He said that if he asked Amodei to stop, he would add that he is not positive Altman is 100% trustworthy. When Bartlett asked about stopping both, Weinstein asked whether that would mean letting Musk and Mark Zuckerberg win, and then mentioned China's Xi Jinping.
Finally Bartlett asked whether he would stop all of it if he could. "Probably," Weinstein said. "If I could somehow stop all of it, I'd probably stop all of it, even at huge expense to myself." His reason was that "survival matters more to me than anything else. I've got two kids on this planet. I love life." He also admitted his own dependence. "It's bad for me because the AI is the only thing that I can ultimately rely on," he said, "and it's not good enough yet."