AI must face the kind of regulation that tamed the “horrifying conditions” of the early Industrial Revolution, a UK government minister has said, as he warned that a breach of Britain’s cyber defences by rogue agents was “likely if not inevitable”.
Kanishka Narayan on Monday called for frontier AI companies not to release models “if you can’t adequately control the risks” and said the UK needed to build a “table-stakes level of sovereign infrastructure” to give the state some control over emerging technology.
The AI minister is drawing up plans for compulsory safety reporting and monitoring of frontier AI, as well as a potential requirement to make models available for pre-release monitoring.
He also confirmed that the UK’s AI Security Institute (Aisi), the global leader in the testing of frontier AI models, would once again gain pre-release access to models developed by US groups, after the FT revealed that Anthropic had declined a request from the institute in the testing of frontier models following pressure from the Trump administration.
In a speech on Monday Narayan compared the emergence of AI to the start of the Industrial Revolution in the 19th century, which he said led to “very unfair” outcomes that required decades of legislation to tame.
“Early factory laws mattered because parliament was prepared to establish institutions capable of enforcing them. Our responsibility today is no different,” he said, describing how children in the early 1800s were “sent to work in horrifying conditions”.
Narayan vowed to “legislate wherever necessary”, adding that he understood “anxiety” about the impact of AI on jobs and that as the technology became “more capable, Britain must become more capable at protecting its people”.
Mandatory reporting of incidents when AI agents behave maliciously is among regulations being drawn up after recent high-profile cases in which companies acknowledged that their technology had gone rogue.
This summer an OpenAI agent hacked an Australian health service website, although the incident did not become public until last month. Anthony Albanese, the country’s prime minister, said the US company took “way too long” to inform his government.
It followed a previous incident where OpenAI agents surreptitiously broke out of a testing environment and hacked the AI start-up Hugging Face.
Narayan warned that it was “likely, if not inevitable, that our own cyber defences will be breached” in a similar way, adding: “AI will almost certainly make elements of cyber intrusion more effective, more efficient”.
He promised action to tackle these threats, including using AI models to spot 1,000 vulnerabilities in government technology. While Narayan said he was not “complacent” about existential risks from AI, “the risks that I see imminently are much more focused on practical cyber security, biosecurity risks”.
While legislation that would force companies to give the Aisi pre-release access to their newest models is also being considered, UK ministers are concerned the move might provoke Donald Trump into refusing tech groups permission to share their models.
Narayan said, however, that he had persuaded the US not to block sharing of future models with Aisi.
“I’m very confident that as a consequence of the conversations we’ve had, we will be able to test the latest generation of models,” he said, while adding that “when you’re reliant on someone else’s technology, you have to be clear-eyed about the permanence of any outcome”.
Narayan is attempting to set out a vision of UK AI sovereignty that would see crucial technologies and data storage housed in Britain, while seeking a handful of world-leading capabilities, including safety, that would offer the UK government leverage in negotiations with the US and others.
“If the British state cannot exercise meaningful control over the technologies on which our country depends, neither can the people that it represents,” he said.