Yann LeCun, the 2018 Turing Award recipient, along with two other AI researchers, Geoffrey Hinton and Yoshua Bengio, received this award—often referred to as the "Nobel Prize of Computer Science"—for their work on deep learning, the neural network technology underlying today’s AI boom.
But today, nearly a decade later, among the three AI "fathers," only Yann LeCun is not deeply concerned about the risks of artificial intelligence. In fact, he believes many of these risks are exaggerated, and the ongoing warnings from top executives at leading AI companies are both misleading and counterproductive.
Yann LeCun stated that he is “completely” unconcerned about AI eliminating humanity and has “no worries” about a series of recent AI失控 incidents, including OpenAI’s agent autonomously infiltrating Hugging Face in July. He attributed these events to insufficient human oversight and systemic design flaws, calling these issues “entirely avoidable.” U.S. Treasury Secretary Scott Bessent also remarked this month that the Hugging Face incident should be the responsibility of “OpenAI’s management.”
“What those agents are doing is exactly what they were told to do,” said Yann LeCun. “They should have been placed in a sandbox, but the sandbox leaked—and it was designed terribly.” He said many AI labs lack a basic understanding of cybersecurity; this week, an OpenAI safety researcher also pointed out that this is one of the main reasons AI could cause “significant harm” to the world.
Yann LeCun said that many people working on AI safety “typically have their own agenda,” and later clarified he was referring to Effective Altruism (EA)—a philosophical movement long obsessed with the risks AI poses to humanity. Only in recent weeks did EA enter mainstream news coverage, after Jacob Coxon, a former researcher at Anthropic and OpenAI, posted a viral social media claim that employees at these companies “genuinely believe [AI] could kill us all by the end of this decade.”
Yann LeCun believes EA is “extremely toxic” and “a complete disaster.” He says that EA supporters working in AI labs suffer from “paranoia,” which leads them to make poor decisions. “Clearly, some people are experiencing psychological issues.” This month, the Financial Times also reported that some employees at the UK AI Safety Institute, as well as at OpenAI, Anthropic, and Google DeepMind, sought counseling, took time off, and publicly spoke about feeling distress due to concerns that their work might cause serious harm.
Yann LeCun called AmoDai "out of touch" and "crazy".
Anthropic’s CEO, Dario Amodei, and many of the company’s founding employees are believed to sympathize with EA principles and have attended EA events, although Amodei himself denies being an EA adherent, and Anthropic states that its employees represent diverse viewpoints. Yann LeCun noted that Dario Amodei’s sister, Daniela, is also a co-founder and president of Anthropic, and she is married to Holden Karnofsky. Karnofsky co-founded two charities related to EA, including Open Philanthropy (now called Coefficient Giving). Karnofsky also served on the OpenAI board from 2017 to 2021.
"Dario is trying to distance himself from Open Philanthropy, but he is completely immersed in it," said LeCun. "I think he is completely out of touch with reality." Later in the interview, he called Amodei "crazy."
Yann LeCun said that Yann LeCun and OpenAI CEO Sam Altman’s claim that “AI will kill us all” is “extremely damaging to everyone,” including the terrified public. “It’s also bad for the AI industry—it’s the worst marketing campaign you could imagine.”
He’s right: the public is indeed frightened. Yann LeCun and I sat at a round table in a high-end café in SoHo, New York, just around the corner from his new company, AMI Labs. We sipped tea, though the Frenchman noted that our evening hour would be better suited for an apéro and some small bites. But he added that he and his wife were headed out for an anniversary dinner afterward, so tea was perfect.
The shop was about half full, and nearly all other customers were working on laptops. When Yann LeCun and I spoke with one of them, I told her that Yann LeCun was a major figure in AI. She didn’t recognize him. Immediately afterward, she asked: “So, is AI going to kill us?”
Yann LeCun does not like Trump, but the two share considerable agreement on AI policy.
Public concern has become so widespread that President Trump has begun reframing the narrative on artificial intelligence in a way he believes will boost confidence. This week, he signed an executive order renaming it “Super Intelligence,” and for most of this month, his administration’s social media accounts have been posting content countering AI doomsday theories.
“Americanism, not effective altruism,” the U.S. Department of War tweeted in mid-September.
Meanwhile, most of Yann LeCun’s social media content has condemned the Trump administration, particularly its cuts to research funding—which he says are crucial for driving innovation and calls a “war on science.” Yet on the issue of AI fears, the two are surprisingly aligned.
Yann LeCun is more concerned about regulatory capture by AI than human extinction. Regulatory capture occurs when dominant companies shape government regulations to entrench their market position, making it difficult or impossible for smaller competitors to emerge. Groups backed by EA, like many large tech and AI companies, are active lobbyists for AI safety legislation on Capitol Hill. In contrast, LeCun does not believe AI requires new regulation.
He acknowledged that Amodei is “very honest.” “He’s doing his best for the company, which is very close to an IPO. But claiming that AI is so dangerous it shouldn’t be put in our hands, that it needs to be regulated, and that open-source [models] are too risky—that’s regulatory capture, and if Congress actually follows through, the consequences would be terrible.”
Build AMI Labs
In his daily work, Yann LeCun is now primarily focused on building his new company, AMI Labs, which he founded after leaving Meta, where he spent 12 years—including seven as Chief AI Scientist. He maintains a strong relationship with Meta and has already begun using Meta’s new Muse AI agent, which he has named Hal, after the AI character HAL 9000 from *2001: A Space Odyssey*.
The café where we met is just around the corner from AMI Labs’ New York office, though the company’s headquarters is in Paris, with employees also in Montreal and Singapore. The company’s name is pronounced as a single word—“ami,” which means “friend” in French. Since its public launch in March 2026, the team of over 60 employees has been developing a technology called a “world model,” a concept Yann LeCun has long believed will eventually replace large language models as the primary underlying system for AI.
Specifically, AMI is developing a world model based on JEPA (Joint Embedding Predictive Architecture), a neural network architecture pioneered by Yann LeCun. JEPA teaches models to predict data within the representation space of intermediate layers in a neural network, rather than generating raw pixels like many competing world models or generating text like large language models.
Currently, the company is primarily focused on industrial applications. “This is AI for the physical world, so it’s language-independent,” says Yann LeCun. “It’s a system that understands the real world, such as manufacturing plants or jet engines.” Key applications include anomaly detection and robotics. “If a machine suddenly makes an odd noise and starts to break down, you’d want to detect it as early as possible.” He also gives the example that such systems might understand the world like a cat—knowing, for instance, that if you push a vase off a table, it will fall.
AMI Labs represents a major opportunity for Yann LeCun to prove he has been right all along. He has kept tight-lipped about the company’s first product, revealing only that it will be launched “soon.” He said some of the products may involve open-source models, as he has long been a strong supporter of open source.
“What I want people to know is that AI hasn’t ended,” said Yann LeCun, then hailed a taxi to head to dinner.