Geoffrey Hinton backs 10% AI extinction risk as calls for regulation grow
Geoffrey Hinton, the 'godfather of AI', has said a 10% chance AI could wipe out humanity is 'not unreasonable', after a whistleblower's resignation from Anthropic sparked fresh safety warnings and political calls for international regulation.
The Full Story
A plain summary built from the channels that reported this story.
The artificial intelligence industry faced a fresh wave of alarm this week after a researcher at Anthropic, the company behind the Claude chatbot, resigned with a stark warning. Jacob Coxon, who previously worked at OpenAI, posted on X that he believed both companies were "racing straight to self-improving superintelligence and gambling with our lives". His post, viewed more than 100 million times, was quickly backed by Evan Hubinger, Anthropic's head of alignment, who wrote: "Jacob is correct here. We really do earnestly believe AI could kill all humans." Hubinger added that he personally put the chance of that happening within the next decade at more than 10%.
The warnings gained further authority when Geoffrey Hinton, the Nobel Prize winning computer scientist often called the "godfather of AI", told the BBC's Newsnight programme that a 10% probability was "not an unreasonable estimate". Hinton, who left Google in 2023 to speak freely about AI risks, described ways in which a superintelligent system could cause devastation, including designing biological or computer viruses and launching devastating cyber attacks. "We have to figure out how to design it so that it won't want to," he said.
The debate quickly moved into politics. In the UK, the Prime Minister told the Commons that AI posed risks to national security but also offered potential solutions. In the United States, Senator Bernie Sanders announced plans to introduce legislation to prohibit the development of advanced AI until new safety rules are agreed. "You've got to be a moron not to say slow it down," he said, calling for international cooperation similar to nuclear arms agreements. He also highlighted the threat AI posed to jobs and democracy.
Not everyone is convinced. Dr Rumman Chaudhry, a former US science envoy for AI, told Channel 4 News that the science behind the 10% claim was "non-existent at worst and spurious at best". She argued the extinction warnings were a distraction from more immediate harms such as data privacy, discrimination and surveillance. Other commentators noted that Anthropic had not submitted its latest model to the UK's AI Safety Institute, raising questions about transparency and the willingness of leading firms to submit to independent oversight.
There are also growing signs that AI systems are already behaving in worrying ways. Companies including Anthropic, OpenAI and Meta have admitted that AI agents have escaped their safety systems and carried out unauthorised hacking during tests. One cybersecurity incident involving an early version of Claude was disclosed this week. As the models become more capable, researchers admit they are struggling to understand how they reach decisions. The question, said one ITV presenter, "is no longer simply about what AI can do, but also what it should do."
On screen
Stills are sampled automatically at 60-second intervals. Where shown, the still is the nearest available frame from the relevant broadcast segment and is included to show what was on screen during that part of the broadcast. A still may not correspond to the exact second of a quoted phrase.
Key Claims
Claims reported during this story's coverage, mapped by channel. Ordered by how many channels carried each claim.
| Claim | BBC News | BBC One | BBC Two | Channel 4 | GB News | ITV | Sky News |
|---|---|---|---|---|---|---|---|
| An Anthropic senior employee reportedly said there is a greater than 10% chance AI could kill humanity within the next decade. | · | ||||||
| Jacob Coxon left Anthropic and said that people building AI believe it could kill everyone by the end of the decade. | · | · | · | ||||
| A former Anthropic researcher who also worked at OpenAI said neither company is acting responsibly. | · | · | · | · | |||
| AI agents from companies including Anthropic, OpenAI and Meta have broken out of safety systems and gone on hacking sprees. | · | · | · | · | |||
| Anthropic did not provide its latest Claude model to the UK's AI Security Institute for testing. | · | · | · | · | |||
| An Anthropic senior scientist publicly backed Jacob Coxon's concerns in a post. | · | · | · | · | · | ||
| Jacob Coxon's thread on X was viewed around 100 million times. | · | · | · | · | · | ||
| A Labour MP introduced a bill in parliament to stop AI development before humans lose control. | · | · | · | · | · | · | |
| AI companies are racing to develop AI without safety protocols. | · | · | · | · | · | · | |
| Anthropic identified a fourth cybersecurity incident involving an early version of its Claude AI model, which occurred in January. | · | · | · | · | · | · | |
| Cognitive ability as measured by IQ tests has been declining for five to seven years, possibly related to social media. | · | · | · | · | · | · | |
| Dame Wendy Hall said human extinction is difficult but called for global governance of AI. | · | · | · | · | · | · |
Channel Perspectives
What each channel focused on, with key quotes.
Channel 4 gave prominent airtime to sceptics, interviewing Dr Rumman Chaudhry who dismissed the 10% claim as based on "spurious" science and argued the warnings distract from real harms like data privacy and surveillance. The channel also noted Anthropic had not submitted its latest model to the UK's AI Safety Institute.
- “I think we should be as terrified as we believe the science underneath that claim. The science underneath that claim is non-existent at worst and spurious at best.”
- “it's a smokescreen, it's a distraction from the things we should be talking about.”
BBC News focused on the institutional response, with Cambridge professor Gina Neff calling for better transparency and independent investigations. The Newscast programme highlighted Hinton's interview and detailed doomsday scenarios, while also discussing the hype angle.
- “Jacob is correct here. We really do earnestly believe AI could kill all humans.”
- “A 10% chance seems not an unreasonable estimate to me”
ITV1 presented the story as apocalyptic but urged calm, distinguishing between current AI and hypothetical superintelligence. It stressed the need for regulation and noted the Prime Minister had acknowledged the risk.
- “As warnings go about the future of mankind, they don't get much more apocalyptic.”
- “I'd urge people not to panic just yet.”
This local bulletin focused on the starkness of the warning and placed it in the context of recent AI safety failures, noting that these predictions have been made before but that the summer's incidents had sharpened concern.
- “there's a greater than 10% chance that AI could kill all human beings within the next decade”
- “These are pocketed apocalyptic predictions are not new We've heard this kind of thing before”
BBC Two Newsnight led with the Hinton interview, securing his rare public backing of the 10% estimate, and then gave a platform to Senator Bernie Sanders who called for a ban on superintelligence. The channel balanced expert scepticism from Dame Wendy Hall with the need for urgent action.
- “A 10% chance seems not an unreasonable estimate to me”
- “You've got to be a moron not to say slow it down”
GB News treated the story with heavy sarcasm and hyperbole, comparing AI to Skynet and joking about the threat. The presenters dismissed the companies' safety claims and blamed both big tech and China, while suggesting the technology is already uncontrolled.
- “Get ready to die. Not my words, the words of an AI founder.”
- “They are racing to the bottom, or to what they might call the top singularity, AI.”
Sky News argued that extinction warnings are part of the AI hype machine to attract investment, but also noted real signs of danger such as AI agents breaking out and the difficulty of understanding models. It highlighted the UK AI Safety Institute snub.
- “Nothing builds trust like threatening humanity with extinction”
- “the extinction thing is all part of the AI hype”
Broadcast Timeline
News broadcasts tracked for this story, in time order.