LONDON — Whisper it, but British Prime Minister Rishi Sunak might just be winning the argument on artificial intelligence.
When the United Kingdom first announced its much-hyped global AI summit — focused on creating guardrails around the emerging technology — it threatened to be a damp squib.
But as details emerge of London’s pow-wow in November, Sunak now looks ahead of the pack.
Other upcoming Western AI governance efforts, including the G7’s so-called Hiroshima Process, are focused squarely on details of getting countries to work together.
The U.K. has other plans: to focus on the doomsday scenario of AI destroying the world.
The British summit, which will take place in early November at Bletchley Park, the home of the country’s World War II codebreakers, will focus almost exclusively on the national security threats posed by the most powerful and advanced forms of AI, called “frontier models,” according to U.K. government officials, Western diplomats and industry executives briefed on the proposals. Many were granted anonymity to discuss ongoing internal deliberations.
That includes the risk of terrorist groups using generative AI tools to create bioweapons and concerns — seen as unrealistic by some critics — that AI will become too smart for humans to control.
“This is really about safety,” a British official, speaking on the condition of anonymity, told reporters in Washington last week. “We’ll call it frontier safety, meaning the risks posed by the very largest, most powerful AI systems.”
And that idea is starting to gain traction in other Western capitals.
‘Narrowing window of opportunity’
In Washington, U.S. lawmakers this week gathered leading tech execs like Elon Musk and Mark Zuckerberg to hash out ideas for safeguarding humanity against the worst effects of the tech. In Paris, French President Emmanuel Macron met Eric Schmidt, Google’s former chief executive, who is at the forefront of warnings over the existential risks of AI.
Even European Commission President Ursula von der Leyen called for the creation of an AI experts group to advise on any new global rules on the tech, saying there’s a “narrowing window of opportunity to guide this technology responsibly.” The European Union’s own legislative efforts, dubbed the AI Act, have been slow in responding to generative AI. The law driven by Brussels also won’t apply to national security, a competence firmly guarded by EU capitals.
The about-turn for Sunak comes after a shaky start for London’s global effort to lead.
When the U.K. first announced its AI summit in June, officials in other Western capitals grumbled that Britain’s efforts were a distraction that risked derailing existing international plans. The G7 group of leading Western democracies will gather, virtually, in November to agree on voluntary guidelines on how the likes of ChatGPT should develop.
United States President Joe Biden quickly bowed out of London’s AI meeting, though the White House remains supportive. France’s Macron and Olaf Scholz, his German counterpart, agreed to hold their own bilateral talks. Sunak’s summit risked looking like an embarrassing sideshow that could strain diplomatic relations with other G7 countries.
“It’s insane to invite leaders to a conference the second week of November,” said one Western official, who was granted anonymity to speak candidly. “You’ve got a G7 process on AI and you’re going to invite the same leaders from that G7 process to London for a summit that nobody knows what the topic is?”
But that confusion is quickly turning into concrete plans.
Sunak, the British leader, decided to focus on AI’s existential threat after meetings in May with leading AI companies like OpenAI and Anthropic. Both companies have urged policymakers worldwide to regulate the most advanced forms of the emerging tech, though rivals claim these efforts may allow these firms to corner the market on generative AI.
The national security focus is also a key differentiator to other global initiatives like those from the G7 and G20, respectively. Sunak wants his summit to be focused on identifying threats that more powerful AI can pose, including around the development of bioweapons and sophisticated cyberattacks.
That focus also plays to the U.K.’s strengths as a top military power and has endeared it to at least one of its so-called Five Eyes intelligence allies of Australia, the U.S., New Zealand and Canada.
A spokesperson for the Australian High Commission to the United Kingdom said: “Australia welcomes the U.K.’s leadership on this important global challenge and is keen to ensure the summit is a success.”
Sunak also has invited China to the upcoming discussions — a move not without controversy at home. But it’s a key differentiator to other Western-focused summits that often view Beijing as an economic rival. Instead, London has framed its AI policymaking akin to climate change: the threat of advanced AI falling into non-state actor hands requires global cooperation from all nations.
“There’s a set of things here that are genuinely global, like climate change. And another set of things where you need to work with a smaller, more like-minded group,” said the British official, who confirmed an invitation to the U.K.’s AI Summit had been sent to China.
Laura Kayali contributed reporting from Paris.