Former attorney general Richard Hermer has warned that artificial intelligence presents an existential risk to humanity and urged immediate international regulation, arguing that frontier AI labs operate under weaker oversight than the appliances in an ordinary home.
Writing for the Guardian, Hermer said the danger posed by AI differs from other national security threats both in its scale and because the time available to address it is limited. He described the risk as one that is difficult to overstate and said the warning signals are already flashing.
Hermer spent two years as attorney general and said he devoted more time to national security than to any other matter during that period. He listed terrorism, armed conflict, nuclear proliferation, financial insecurity and the climate crisis among the threats he handled, but argued that AI carries a different order of magnitude. The intelligence and defence communities are well positioned to counter those other challenges, he said, whereas AI demands a response that is harder to organise.
He pointed to two recent developments — the disclosure of the Hugging Face hack and the resignation of Jacob Coxon from Anthropic — as the events that finally pushed the issue into wider public discussion. Hermer said the moment should be seized, and called for greater urgency in coordinating international rules for frontier AI.
Public involvement matters too, he argued. Decisions about how AI is shaped should serve humanity broadly rather than concentrate power in a handful of hands, and he said the debate should not be ceded to China, the United States and large technology firms alone.
Hermer acknowledged that AI offers substantial benefits, including advances in science and medicine and the potential to empower some of the world’s most disadvantaged people. He nonetheless divided the risks into two broad categories.
The first is existential risk — the possibility that AI could cause the deaths of billions, whether because it allows malicious actors to design and deploy biological weapons, or because the technology concludes that dispensing with humans is more efficient.
The second he termed existential challenge: the prospect that unchecked AI development strips billions of people of meaningful control over their own lives, with severe consequences for economies, national security and human wellbeing.
He cited Bill Gates, who has written that the AI era will bring “one of the most turbulent times in human history” and who has said the technology could prove either the greatest equaliser ever devised or the worst source of injustice. Hermer agreed, warning that on the current path AI risks transforming the meaning and availability of productive work while handing extraordinary power to two states and to a small group of individuals controlling major technology companies.
The precise scale of the risk remains unclear, Hermer conceded, and the full picture of the challenge is not yet known. Part of the difficulty, he said, is that AI is advancing beyond human comprehension and will eventually reach a point where control may no longer be possible even if it is understood.
He drew a parallel with the 2008 financial crash, which he said was caused partly by the fact that no person truly grasped the tangle of transactions leading to collapse. The current situation, he argued, is larger in the complexity of its systems, in the human oversight available and in the potential devastation.
Hermer described the level of risk as unacceptable. He compared the situation to a toaster manufacturer being permitted to sell a product carrying a 10% chance of electrocution, and said the argument for slowing frontier AI until its risks can be managed is obvious.
Regulation, he wrote, is needed now, noting that fridge safety is governed more tightly than the work of frontier AI laboratories. Leaving the industry to police itself is not credible, he said, because AI companies and the financial markets behind them are driven to release new versions as fast as possible, a pressure sharpened by geopolitical competition between China and the United States. A “voluntary code” presented by President Trump and Elon Musk on the White House lawn, he added, offered little comfort.
International law will be decisive, according to Hermer. He said the UK has demonstrated strong leadership — begun under Rishi Sunak, continued by Keir Starmer and now reflected in Andy Burnham’s promotion of Kanishka Narayan to cabinet — but stressed that the challenge’s scale means cooperation across borders is essential to achieve anything. He described the current moment of danger as arriving while the international rules-based order is under unprecedented strain, and called for deeper engagement with the US and China alongside work with coalitions of like-minded states and the technology sector.
Democracy requires an urgent public debate about what lies ahead so that people retain control of their own futures rather than letting the technology sector decide, Hermer argued. He said the outcome is not inevitable and that governments and individuals still have time to shape AI toward the public good.
Since the Industrial Revolution, he noted, political movements have been shaped by the relationship between the means of production and the interests of ordinary people. New governing theories are needed, he said, to settle who controls AI, how it is regulated and, above all, whom it serves.
Hermer closed by calling for public conversations about the values that should guide AI decision-making, warning that the benefits of technological progress must be steered toward the many rather than surrendered to the few. Wait any longer, he cautioned, and AI may remove that ability altogether.
Richard Hermer is a Labour life peer. He is a former attorney general for England and Wales and former advocate general for Northern Ireland.
COMMENTS