Nick Clegg has dismissed warnings that artificial intelligence could acquire a “godlike power to exterminate humanity”, saying the talk shows tech bosses are “breathing their own fumes”.
The former deputy prime minister, who went on to run global affairs at Meta before leaving in 2022, made his remarks on BBC Radio 4’s Today programme, and they quickly recirculated through the Guardian and wider tech press. His target was not AI itself but the industry’s doomsday script: the claim, pushed by several frontier lab leaders, that catastrophic or existential risk from advanced models deserves priority over near-term harms. Many within the sector, he said, are “winding themselves up into a lather”.
Clegg’s own position is not neutral, and he does not pretend it is. He is a shareholder in UK datacentre firm Nscale and co-founded an AI startup, AMI Labs, which has an interest in the industry not being strangled by precaution. He has called the apocalyptic framing “self-serving” on the grounds that it paralyses practical regulation while entrenching the small oligopoly of labs grandstanding about survival risk.
What he wants attention paid to instead is the list of concrete dangers: cyber attacks, bioweapons misuse, and the degradation of information systems. It is, in other words, the risk register a bank’s compliance team would recognise, as against the one a lab’s safety team presents at a parliamentary hearing.
He also offered a cold reading of Britain’s ambitions in the space. The UK should be realistic about how much it can shape global AI rules, he said, a caution aimed at the government’s repeated framing of Britain as a standard-setter rather than a rule-taker.
The debate Clegg has restarted is the one that has run under the AI safety conversation since 2023, now with a fresh wrinkle: the labs that warned of extinction in the hearing rooms have spent the intervening years raising prices, cutting safety staff and racing each other, which makes the existential talking point easier for sceptics to read as marketing. Neither the doomers nor the dismissals should be taken at face value. The useful question is what a country like Britain actually legislates between a risk nobody can measure and harms that are already arriving, one phishing email at a time.