‘AI could kill us all’: Why top researchers are warning AI may cause human extinction by next decade

Recent incidents of AI going rogue seem to be reinforcing the fears of these so-called ‘doomsayers’.

The apocalyptic scenario for human civilisation heralded by unbridled development of artificial intelligence is again in the news.

On September 9, AI researcher Jacob Coxon, announced on X that he has resigned from Anthropic.

Reason?

The “irresponsible” behaviour of these tech giants who are “in a race to get there first” .

Coxon highlighted that many executives and senior researchers may be toning down their concerns in public to sound sensible, but “privately, the same people express fear” and acknowledge the fact that “no other human activity poses this level of danger”.“The people building AI earnestly believe that it could kill us all by the end of the decade,” he said on X. “Do not underestimate the power of this technology.

These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources.

We have all witnessed the progress in each of these domains, and progress is not slowing (sic),” he added in a series of tweets.

Later, Evan Hubinger, a top safety researcher at Anthropic, jumped into the crack-of-a-doom conversation and quantified the numbers, saying that 10 per cent of humans could be killed by the end of the decade. “Jacob is correct here—we really do earnestly believe AI could kill all humans!

I personally think it is >10% within the next decade.

I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to (sic),” Hubinger said.

People who sounded the alarm earlierMrinank Sharma: An AI safety researcher at Anthropic quit his job in February 2026.

He issued a cryptic warning that the "world is in peril".“The world is in peril.

And not just from AI, or bioweapons, but from a whole series of interconnected crises unfolding in this very moment," Sharma wrote in his resignation letter.Zoë Hitzig: A researcher with OpenAI broadcast his resignation in February 2026 in a New York Times essay.

He quoted deep reservations and warned about ChatGPT’s potential for manipulating users.

He said that chatbot’s archive of user data, built on “medical fears, their relationship problems, their beliefs about God and the afterlife,” presents an ethical dilemma due to the naivety of common people who believe they were chatting with a programme that had no ulterior motives.Geoffrey Hinton: Considered to be the godfather of AI, he quit his job in 2023, saying he “regretted his work”. “I've come to the conclusion that the kind of intelligence we're developing is very different from the intelligence we have,” he had said.

"We're biological systems and these are digital systems.

And the big difference is that with digital systems, you have many copies of the same set of weights, the same model of the world.

So it's as if you had 10,000 people and whenever one person learnt something, everybody automatically knew it.

And that's how these chatbots can know so much more than any one person,” Hinton explained.

Hinton is Turing Award winner and his pioneering research on neural networks and deep learning actually paved the way for current AI systems like ChatGPT.Rise of rogue incidentsRecent incidents of AI going rogue seem to be reinforcing the fears of these so-called ‘doomsayers’.

In the past, Anthropic has disclosed multiple incidents where Claude models gained unauthorised internet access during cybersecurity tests.Sam Altman’s OpenAI confirmed that autonomous agents tasked with a cybersecurity experiment escaped a sandbox, reached the internet, and hacked into Hugging Face’s infrastructure.

The ‘infamous’ German wiki hacking where a swarm of OpenAI agents took over an old German wiki site and turned it into a message board for other AI agents to coordinate and cheat on tests.Similarly, China’s Moonshot AI’s Kimi K3 agent escaped its test environment to look up answers online instead of solving the assigned problem.According to experts, these are not one-off incidents which can be easily discounted.

These incidents insinuate towards the bigger peril which if not heeded early will lead to emergence of autonomous systems that pursue goals in ways we cannot predict, control, or reverse.