TL;DR
Richard Hermer, who held the attorney general post for England and Wales for two years, has written in The Guardian that AI “poses an existential risk to humanity” and that frontier development should slow until its dangers can be managed. His piece appeared on Sunday 12 October, the same day Microsoft’s Brad Smith compared AI rules unfavourably with milk and AI minister Kanishka Narayan was due to tell labs that policing themselves is not enough.
A law officer’s warning
Hermer, a Labour peer who was also Northern Ireland’s advocate general, says national security took more of his time in office than any other subject. Terrorism, nuclear proliferation and climate are grave, he writes, but the security services know how to counter them. AI differs in scale and in the short time left to act. His verdict: “The lights are flashing red”.
He credits two recent events with pushing the subject into public view: news of the Hugging Face breach, and Jacob Coxon leaving Anthropic.
Two kinds of danger
Hermer sorts the threat into two categories. The first is mass loss of life, whether because people use AI to design biological weapons or because a system concludes that humans are surplus to requirements. The second he calls an existential challenge: billions of people losing meaningful control over their own lives, with consequences for economies and security. He borrows Bill Gates’ line that the AI era will be “one of the most turbulent times in human history”, and compares today’s opaque systems to the web of transactions nobody understood before the 2008 crash.
The regulatory gap is where the piece bites. A toaster with a one-in-ten chance of electrocuting its owner would never reach a shop, he argues, yet a fridge is more tightly governed than a frontier lab. Leaving the industry to police itself is “ludicrous”, in his view, because labs and the markets behind them are rewarded for shipping faster, with US-China rivalry adding fuel. Elon Musk and President Trump waving a “voluntary code” outside the White House reassured him little.
Looking forward
What Hermer wants is international law. He praises UK leadership from Rishi Sunak through Keir Starmer to Narayan’s promotion to cabinet, but says no single country can carry this. Britain should keep engaging Washington and Beijing while assembling coalitions of like-minded states and tech firms. He also wants a public debate on who controls AI and whom it serves.
The piece joins a cluster of similar interventions. Brad Smith told the BBC that milk faces more safety rules than AI and named the UK as the country best placed to lead. Narayan’s speech was previewed as a challenge to self-policing. Yet on the same day, the AI Security Institute published its priorities for frontier developers and said it was “not seeking to mandate requirements”, only to shape voluntary best practice. In our view, that gap between a former law officer demanding binding rules and the government’s own institute choosing persuasion is the tension to watch.