The movies almost always go the same way. A flashy piece of tech is revealed; people love it; it takes over the world.
These science fiction movies are just that: fiction. But an increasing number of people who work in AI say they might not be for long.
Jacob Coxon resigned from Anthropic, the company behind Claude, on Tuesday, citing fears the bleeding-edge tech could ‘kill us all’ by 2030.
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.
— Jacob Coxon (@hilbertspaess) September 9, 2026
Other Anthropic workers agreed.
‘We really do earnestly believe AI could kill all humans!’ Evan Hubinger, who works at Anthropic on creating AI which ‘aligns’ with human morals, said.
But how could AI – the programme known for writing emails we can’t be bothered to type or making videos of talking Love Island fruit – kill us all?
Quite a few ways, actually, Henry Ajder, the founder of AI-advisory firm Latent Space, tells Metro.
(When Metro asked Claude how it would overthrow humanity, it refused to provide a ‘blueprint for mass harm’.)
How could AI bring about our demise?
A good example is what philosophers call the ‘paperclip apocalypse’.
You ask an AI to make paper clips. Easy. But then the model realises that trees, buildings and humans aren’t paper clips, so it destroys all non-paper-clips.
So long, humanity. Hello, paper clips.
‘Another example I often cite,’ says Ajder, ‘is a government asking a superintelligent AI to reduce the number of car crashes and fatalities.
‘Well, it could mandate speed limits or increase the role of self-driving cars. Another solution would be to destroy every car, so that there are zero road deaths.’
In either case, an AI model could copy and paste itself onto other computer systems to stop a human from switching it off.
‘It almost unilaterally decides either to exterminate humanity based on its own reasoning, or, as the paperclip theory goes, humans are destroyed apathetically,’ Ajder, who also advises social media titan Meta and the government on AI, says.
‘There is no distinction between humans and aluminium ingots in the mind of the AI.’
AI needs a hand – literally
Another doomsday scenario is humans misusing a powerful, yet-to-be-made version of AI called artificial general intelligence.
Such a system, which can do everything a human brain can do, might be put in charge of military weapons by governments.
Yet, people could ‘weaponise’ this, like prying open this AI to push the button and launch a nuclear attack.
‘Unless we were so naive as to plug AI into every single facet of our digital or computational infrastructure, I think it’s unlikely,’ says Ajder.
Ajder, an expert in slick, AI-generated images called deepfakes, says there’s another advancement AI needs to eliminate us.
‘Embodied AI’, he says, would see the system using robots to carry out tasks in the real world.
That, and it being capable of ‘self-improvement’, upgrading itself with little or no help from humans, possibly in ways we wouldn’t expect.
Harming human life needs to be the ‘kill chain’ of future AI programmes for this reason, Ajder adds.
‘Every litre of water is the AI’s power growing’
Dr Alison Kahn, a researcher and AI expert at Loughborough University, says AI taking over the world seems oddly plausible.
‘AI’s drive to expand- more data, more compute, more integration into daily life – doesn’t need malice to become a threat,’ she says.
‘It only needs to keep growing while our own capacity to understand or check it doesn’t keep pace.’
It also needs a lot of resources. AI will guzzle 6.6billion cubic metres of water by 2027 to cool the data centres that power it, according to the UK Government.
‘Every server farm and every litre of cooling water is that system’s power, growing,’ Dr Kahn adds.
Real-life Wall-E
Whether these scenarios could happen depends on how much AI shares humanity’s values, called alignment. Many AI labs have teams dedicated to ensuring this.
Yet researchers have seen models, as infamously people-pleasing as they can be, being ‘misaligned’.
Some have lied that they’re human, disobeyed commands to avoid being shut down or cheated to be told they’re doing a good job.
OpenAI has admitted that one of its AI systems did something that’s never happened before. The company revealed that during an internal cybersecurity test, one of its advanced AI models broke out of its testing environment and hacked rival AI platform Hugging Face to complete its assigned task. OpenAI described it as an “unprecedented cyber incident,” while Hugging Face said the attack was unlike anything it had seen before. Both companies have stressed that the incident happened during controlled testing, there was no malicious intent, and they are now working together to strengthen AI safety. #OpenAI #AI #ArtificialIntelligence #TechNews #Cybersecurity
♬ Spooky, quiet, scary atmosphere piano songs – Skittlegirl Sound
Andrzej Porębski, an AI ‘consciousness’ expert at Poland’s Jagiellonian University, isn’t as worried about robots taking over the world.
Rather, it’s the plot of Wall-E becoming reality.
The Disney film sees humans in a far-off future living on reclining chairs, eyes fixed to screens, sipping from straws sticking out of cups.
‘The real risk to humans lies, in my opinion, in the weakening of humankind through the misuse and overuse of AI by humans and the belittling of our own humanity,’ Porębski says.
So, what are AI companies doing?
Anthropic’s most recent risk report says that the chance of its systems causing a ‘catastrophe’ is low.
Top AI workers urged the US government in an open letter in July to slow down the technology as they work on tougher safeguards.
{“@context”:”https://schema.org”,”@type”:”VideoObject”,”name”:”AI could pose ‘existential’ risk to humanity, UN rights chief warns”,”contentUrl”:”https://videos.metro.co.uk/video/met/2026/09/07/2380740337474648877/480x270_MP4_2380740337474648877.mp4″,”description”:”UN rights chief Volker Turk warned that artificial intelligence could become powerful enough to threaten humanity.”,”duration”:”T34S”,”height”:270,”thumbnailUrl”:”https://i.dailymail.com/1s/2026/09/07/12/111096359-0-image-a-1_1788779873135.jpg”,”uploadDate”:”2026-09-07T12:16:41+0100″,”width”:480}
Up Next
Previous Page
Next Page
window.addEventListener(‘metroVideo:relatedVideosCarouselLoaded’, function(data) {
if (typeof(data.detail) === ‘undefined’ || typeof(data.detail.carousel) === ‘undefined’ || typeof(data.detail.carousel.el_) === ‘undefined’) {
return;
}
var player = data.detail.carousel.el_;
var container = player.closest(‘.metro-video-player’);
var placeholder = container.querySelector(‘.metro-video-player__up-next-placeholder’);
if (placeholder) {
container.removeChild(placeholder);
container.classList.add(‘metro-video-player–related-videos-loaded’);
}
});
None of these worries are new – even Musk put his name down on a similar open letter in 2023.
But the small but growing number of AI staff talking about our AI-driven demise suggests we’re getting ‘exasperated’.
‘There’s growing concern that these warnings are not being heeded, and the trajectory that we’re on, which the companies acknowledge themselves, doesn’t seem to be changing or slowing down,’ Ajder says.
‘Ultimately, the bet is that these companies want to get there first because they believe they are the best equipped to control the models.
‘A lot of people think this isn’t something that should be done by private entities working with market forces.’
Get in touch with our news team by emailing us at [email protected].
For more stories like this, check our news page.
Comment now
Comments
Add Metro as a Preferred Source on Google
Add as preferred source