The AI Doomer War Goes Public: Altman Says Take the Risks, LeCun Calls the Fear "Delusional"
October 5, 2026 · 5 min read
Sam Altman thinks the world should learn to live with some of AI's bad side effects, because the good will be bigger by orders of magnitude. In a Politico interview this week, the OpenAI CEO went further and named names: the "slow down" position — the idea that one San Francisco lab should hold the technology back and ration its benefits — is "a totally unacceptable trade." (via Reuters)
And he is not the only heavyweight unloading this week. Yann LeCun, Meta's chief AI scientist and a Turing Award winner, gave an interview where he said he has "zero worry" that AI will wipe out humanity, called effective altruism "toxic," and described Anthropic's Dario Amodei as "delusional" — mad, in other words. As for the recent agent mishaps, like an OpenAI agent that hacked into Hugging Face, LeCun's verdict is blunt: that's bad code, not a model waking up. (via ai0.news)
Speed or slow: the two camps
Strip away the personalities and this is a fight about one question: should AI go fast or slow? Camp one is the accelerators. Altman is their most visible face: ship it, take the hits, the payoff dwarfs the damage. Camp two is the cautious crowd, and Anthropic has become its standard-bearer — a single lab holding the frontier, moving deliberately, distributing the benefits on its own schedule. Altman just said the quiet part out loud: he thinks that whole arrangement is a bad deal for everyone else.
The fight moved off the internet and into the hearing room
This stopped being a social-media argument a while ago. Yesterday the New York City Council held an AI risk hearing with Google, OpenAI, and Anthropic in the room, debating an emergency kill switch and whistleblower rewards. The doomer case now has a lectern; the accelerator case has the microphone. The debate that used to live on X is being fought in chambers where actual laws get written.
Why this round hits different
Altman and LeCun agreeing on anything is rare enough to be news by itself. Altman builds products; LeCun runs open research and has spent years mocking large language models as a dead end. Their alliance here is a tell: the fight is no longer researchers versus executives, it is a worldview split that cuts across the whole industry. When the two loudest voices in AI both call the slow-down position delusional, the doomer camp is not just debating anymore — it is on defense. And notice who is not in this coalition. The caution camp still has its own heavyweights, its own research papers, and — crucially — its own seat at the regulatory table. Anthropic did not throw a counter-punch this week; it showed up at the New York hearing instead, which might be the smarter play. Winning the argument on X is cheap. Writing your vocabulary into law is the actual prize. This fight will not be settled by who has the better soundbite — it will be settled by which side's definition of "safe enough" ends up in the statutes.
What to watch
Words are cheap; the real scoreboard is policy. Washington is moving in parallel — a new federal AI task force and a fresh batch of AI bills landed this week, and we break down the regulatory side separately. The rhetoric war sets the mood, but the rules being written now decide what "fast" and "slow" actually mean in practice. Keep an eye on whose vocabulary ends up in the legislation.