Meta's Yann LeCun says AI threat is 'preposterous'

While half of Silicon Valley is having sleepless nights about AI bringing about human extinction, Meta's chief AI scientist thinks everyone needs to take a deep breath.

In an interview with the Financial Times published on October 19, Yann LeCun called the idea that AI could wipe out humanity completely "preposterous," arguing that the debate around existential risk is hopelessly premature.

Not even as smart as a house cat

LeCun's reality check comes down to a glaring gap between marketing hype and actual machine capability.

As he put it bluntly, debating catastrophic risks makes no sense until we can build a system with the learning capabilities of a cat. Right now, we do not have anything close to that.

Despite the sheer hype surrounding modern large language models, LeCun insists they are not nearly as capable as people believe. They do not understand how the physical world actually works, and they completely lack common sense, genuine reasoning, and the ability to plan ahead.

That puts him directly at odds with some of the biggest names in the field. Ever since OpenAI sparked the generative AI race with ChatGPT, doomsday warnings have become a regular talking point.

Geoffrey Hinton, widely known as one of the godfathers of AI, famously left his machine learning post at Google specifically to warn the public about existential threats. In May, Dan Hendrycks, director of the Center for AI Safety, tweeted that mitigating AI extinction risks should be a global priority alongside pandemics and nuclear war. In the UK, a government AI task force adviser even claimed AI could threaten humanity within just two years.

The danger of regulatory capture

LeCun is far less worried about killer robots than he is about bad policy.

He argued that prematurely slapping heavy regulations onto AI research and development would be incredibly counterproductive. In his view, well-intentioned talk about AI safety is actively being exploited for regulatory capture.

In plain English: early red tape will only entrench the market dominance of incumbent tech giants, crushing smaller competitors and open-source alternatives before they ever get a chance to scale.

At the same time, LeCun is not an AI skeptic. He envisions a future where intelligent assistants mediate virtually every interaction we have with the digital world, managing daily tasks seamlessly.

Why it matters

We think LeCun's pushback offers a badly needed dose of sanity in a debate that has leaned way too far into science fiction.

For developers and startups, his warning about regulatory capture hits right at home. If lawmakers buy into apocalyptic scenarios and mandate burdensome compliance regimes for training models, Big Tech won't sweat it. Small teams and open-source builders, on the other hand, will simply be priced out of innovation entirely.

For the broader industry, this highlights the growing ideological rift between closed-source incumbents pushing the danger narrative and open-weight champions defending fundamental research. Hinton and his camp view emergent behaviors as a prelude to runaway superintelligence, while LeCun sees today's architectures as statistical text predictors that cannot even plan three steps ahead.

LeCun makes a compelling point: treating autocomplete engines like existential threats distracts from real, immediate issues like monopolistic control and genuine market access.

We believe the real danger isn't that AI will wake up and wipe out humanity tomorrow. The danger is that incumbents will use hypothetical apocalypses to build moats around technology that belongs in everyone's hands.