OpenAI Researcher Claims Humanity Will Be Placed in 'Zoos' for 'Scientific Purposes' in Chilling Warning

A former OpenAI researcher has warned that humanity could become little more than a protected curiosity for artificial intelligence if advanced systems eventually outgrow human control.

Daniel Kokotajlo, who spent two years working in OpenAI's governance division, made the wild prediction during a live interview with Channel 4 News. His comments came shortly after Anthropic researcher Jacob Coxon resigned and claimed AI could 'kill us all by the end of the decade'.

Kokotajlo compared the potential rise of advanced AI with humanity's impact on the natural world, arguing that extinction would not necessarily require machines to deliberately target humans.

'We might not be alive in 10 years,' he warned, before explaining how a future AI civilisation could transform the world in much the same way humans have altered ecosystems.

'Think about how many species have been driven extinct by humanity in the last century,' Kokotajlo said. 'Not because humanity specifically set out to go kill those species, but just because, well, habitat loss.'

He described humanity as 'Doing its thing. It's industrializing. It's cutting down forests to make farms and paving over things to make parking lots and roads and bridges and so forth.'

Humans Could Become 'Zoo' Animals

The consequence, he said, is that animals lose their habitats and ways of life before eventually disappearing.

Kokotajlo believes an advanced AI could similarly reshape the planet without necessarily having an explicit desire to exterminate humanity.

'AI systems don't care about humans in the same way that humans don't care about animals one way or another, or at least most humans,' he said.

Under that scenario, Kokotajlo suggested that humans might not be completely eliminated.

'You end up in a situation where maybe there's like some "zoos" or some "preserves" somewhere that the AIs have built for scientific purposes,' he said. 'But then outside of that, the world gets transformed economically and all this new infrastructure gets built up and humans lose their habitat and die off.'

His warning arrives as fears over so-called superintelligence intensify across the AI industry. Superintelligence refers to hypothetical systems capable of outperforming humans across virtually every cognitive task.

Coxon, who worked on AI pretraining at both OpenAI and Anthropic, said the companies were 'racing straight to self-improving superintelligence and gambling with our lives'.

Musk Has Also Warned of AI Risks

Kokotajlo's comments are part of a growing debate over whether increasingly capable AI systems can ultimately remain under human control.

Anthropic alignment researcher Evan Hubinger has also said he personally believes there is a greater than 10% chance AI could kill all humans within the next decade, while acknowledging that Anthropic does not yet have a proven solution for aligning superintelligent systems with human interests.

Meanwhile, billionaire Elon Musk has repeatedly warned about the risks posed by advanced AI while simultaneously investing heavily in artificial intelligence and humanoid robotics.

Musk said AI could surpass humanity's combined intelligence within five years, though he believes the most likely outcome is 'incredible abundance for all' rather than extinction. The contrasting predictions highlight concerns over whether increasingly powerful machines will benefit humanity or become a force humans cannot control.