AI pioneer Yann LeCun is publicly rejecting the doomsday case around artificial intelligence.
In an interview with Fortune, LeCun said he is not worried 'at all' about AI wiping out humanity and has 'zero concerns' about recent rogue AI incidents, including OpenAI agents that autonomously hacked Hugging Face in July. He said those incidents were caused by poor human oversight and weak system design, and called them 'totally preventable.'
'Those agents are doing exactly what they’ve been asked to do,' LeCun said. 'They were supposed to be in sandboxes, but the sandboxes were leaky and horribly designed.'
LeCun aimed his sharpest criticism at Anthropic CEO Dario Amodei. He said Amodei is 'completely deluded' and later added that he is 'crazy.' LeCun also said he has a particular disdain for Amodei and his views on Effective Altruism.
The dispute lands at a moment when AI companies are fighting not just over technology, but over the story surrounding it. One camp warns of catastrophic risk and argues for heavy oversight. Another says the real issue is execution: better engineering, tighter controls, and clearer responsibility inside the companies building the systems.
LeCun’s comments put that split in plain view. He is not arguing that AI is harmless. He is arguing that the danger is being overstated, and that failures blamed on the machines often trace back to human decisions, bad design, and weak boundaries.
For capital, that distinction matters. If the problem is solvable through better product design and internal controls, the market may treat AI as an engineering challenge. If the problem is framed as existential, then the pressure shifts toward regulation, liability, and a slower path for free enterprise.
That is why this fight reaches beyond personal insult. It is a debate over who owns the risk, who writes the rulebook, and how much discretion companies should keep before the administrative state steps in.
The market has already voted for AI, but it still has to decide how much bureaucracy it is willing to tolerate around it. Capital rewards clear rules. It also punishes panic. In the end, the winners may be the firms that can prove they understand both the technology and the costs of ignoring basic cybersecurity.
