Hot topics

Why some experts say AI could wipe out humanity

One researcher quit his job and warned that AI could destroy humanity. Another researcher agreed with him. A famous scientist told the BBC a 10% chance of this “seems not unreasonable.” Here's what they actually said, who disagrees, and who would be to blame.

Last updated September 22, 2026

Stylized, fictional illustration of giant robots destroying a city while people flee — the kind of imagery these warnings evoke, not a depiction of anything in this article.
This is just a picture, not a real prediction — the actual warnings below are about percentages and chances, not robots in the street.

Terms

p(doom)
A short way AI researchers say "my guess at the chance AI could cause a disaster or wipe out humanity." It's just someone's personal guess — not a real measurement, like a temperature or a score.
AI alignment
The work of making sure an AI system actually does what its makers want it to do — and doesn't find some strange, unwanted way to reach its goal instead.

What "10% chance" actually means

The number in the headlines is the chance that something catastrophic happens at all — it is not a prediction that AI will kill some specific share of people, like one in ten. Think of a weather forecaster saying "10% chance of a hurricane this season." That's very different from saying "a hurricane will destroy 10% of the coastline." It's also just someone's personal guess, not a real measurement. Nobody has a formula that can calculate this. Keep that in mind as you read the rest of this page.

What was actually said

"I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives."

Jacob Coxon, Sept. 8, 2026, on resigning from Anthropic

"We really do earnestly believe AI could kill all humans! Unless there is AI regulation or a coordinated slowdown between labs, human extinction in the next few years seems very likely."

Evan Hubinger, Anthropic alignment science lead, responding to Coxon

"A 10 percent chance seems not unreasonable to me... nobody really knows how to give a sensible estimate."

Geoffrey Hinton, Turing Award winner, BBC interview, Sept. 10, 2026

Hubinger also gave his own number: he thinks the chance AI causes human extinction in the next ten years is "above 10 percent." He said today's AI isn't very dangerous by itself — he's worried about future AI that's much more powerful. Someone asked Hinton how AI could actually kill people. He didn't point to just one scary idea. Instead, he mentioned things like AI-designed viruses and computer attacks: "there's so many ways it could do it."

Not everyone agrees

Not everyone agrees with this. Some researchers — including people who work on AI safety and security — pushed back hard on the same story.

"AI killing us all by creating dangerous viruses is total bogus."

David Bellamy, infrastructure engineer, Institute of Foundation Models — he says the real problem is getting lab equipment, not AI itself

"Using fear under the pretext of protecting the public, these oligopolies are now requesting to bend competition rules."

Aidan Gomez, CEO, Cohere — he says big AI companies want rules written to help themselves

"The current incidents that we've had have generally been security incidents."

Artem Dinaburg, Chief Research Scientist, Trail of Bits — he says these are security problems we can fix, not an unstoppable disaster

Sayash Kapoor, a professor who will soon teach at UC Berkeley, makes a similar point. He says there's "lots of low-hanging fruit" — meaning easy fixes — for controlling AI systems better. He thinks today's problems are things we can fix, not proof that disaster is coming no matter what we do.

Who's responsible if it happens

On this question, people actually agree more than the headlines make it seem. Even the researchers most worried about extinction don't blame "the AI" itself, like it's some separate villain. They blame the people and companies who build it. Coxon said this directly when he quit: "Neither company is acting responsibly." That matches what we've found on our other AI pages: a model is a trained system doing the job it was trained for, and when it resists being shut down, the best explanation is that it's just doing what it was trained to do — not that it wants to survive. When a tool fails, even in a huge way, the blame usually traces back to how it was built, tested, and released — and who did that.

The real argument is about which people should be blamed. Gomez says AI companies are using fear to write rules that help themselves. Coxon and Hubinger blame company leaders for racing ahead instead of being careful. Rep. Bill Foster raised a different idea entirely: if the U.S. slows down on its own, would that just let Chinese AI companies get further ahead? Others blame regulators for not yet building the rules needed to watch over these companies. None of this is settled — this is the part of the story that's really about opinions and choices, not just facts.

The one place we take a side

Our opinion, not a fact

Here's where we break from being "nonpartisan": we don't think AI is going to kill us.

We think the actions of AI are really the actions of the people who use, own, and operate the data centers behind it. A bomb doesn't attack someone on its own — it has a trigger, and a person pulls it. AI works the same way: there's always a person, or a group of people, behind it making the real decisions. The thing to fear isn't AI itself. It's AI with no rules or oversight.

AI isn't taking your job, the CEOs and executives at the companies are when they lay their workers off and use AI as an excuse. Companies are actively switching to automated and robotic workforces to improve their bottom lines and those decisions are what is taking the jobs.

Nothing hurts more when you, the worker, the one who brought the value to the company, get laid off. Nothing hurts more knowing you made a CEO a billionaire and are struggling to afford the basic costs of living. AI is not the blame here, it is the people making decisions and you have the power to help change things by being active in our democracy and ensuring you are registered to vote!

AI has already helped us make scientific discoveries, and it will keep helping us make more. It's here to stay, so we need to figure out how to use it well — for the people of this country and this planet. That includes making sure we don't use it in ways that make climate change worse or actively engage its use during wars that result in innocent civilians dying.

Talking points

These are questions to ask people who want to be elected and represent you. We won't tell you the right answer, but we DO think you should be talking about them.

  1. If an AI system hurts someone, should the company that built it be legally responsible — the same way companies are responsible for other unsafe products?
  2. Before a company releases a very powerful AI system, should an outside group have to check that it's safe first?
  3. Who should decide if an AI is "too risky" to release — the company itself, a government watchdog, or elected leaders?

Read more

WONKRZ STORE

Upgrade Your Gear Today

Discover our latest collection of products and exclusive deals. This store supports various initiatives to keep people informed and engaged in our democracy regardless of political party.

Shop Now →