The odds are against Rishi Sunak’s AI summit. He might just pull it off

LONDON — Optimism is in the air.

It’s September 2019, and at a dinner involving diplomats and tech execs at a private dining club in Paris’s eighth arrondissement, hopes are high that the world’s rival superpowers — the United States and China — will put aside their differences to discuss the future of a rapidly emerging field important to both: artificial intelligence.

At the summit the next day, those hopes quickly fizzle. 

A U.S. official opens the debate by saying Washington will never cooperate with Beijing on AI for as long as China remains authoritarian. It goes downhill from there, descending into a mud-slinging match between the two global titans.

“Here we are trying to sort out this huge thing and China and the U.S. are squabbling like two divorced parents,” said Neil Lawrence, a leading academic on AI at Cambridge University recalling the 2019 event he was present at.

More than three years on, Rishi Sunak will be hoping his own AI summit, to be held in historic Bletchley Park next week, will go better.

With senior representatives of both the U.S. and China invited, the stage is set for Sunak to achieve where others have failed: to get Washington and Beijing talking and even — whisper it — signing a shared communiqué outlining the risks of AI.

It would be quite the diplomatic coup — and one that could mark the difference between success and failure for Sunak’s much-hyped summit. The British PM will be desperate to burnish his reputation on the world stage after a bruising set of by-election losses to the resurgent opposition Labour Party last week. His party is widely expected to lose the next general election, to be held before the end of 2024.

POLITICO spoke to U.K. government officials, foreign diplomats, tech company insiders, academics and civil society representatives for this piece, many of whom were granted anonymity to speak candidly about sensitive discussions. The picture that emerged was of a conference beset by shambolic organization, and that has at times irked Britain’s allies and faced criticism of skewing the global conversation on AI away from present risks to as yet unrealised apocalyptic scenarios.

Yet it’s a gathering that could also, against the odds, see Sunak hailed as far ahead of the pack on AI governance. 

“Demonstrating alignment would be a powerful step forward in the global governance of AI,” Rumman Chowdhury, a leading expert in global AI policy, said of the possibility of an agreement involving the big geopolitical rivals.

Last-minute scramble

With less than a week to go, details of the summit remain alarmingly sketchy. 

Some participants were told at first that there would be a gala dinner at Buckingham Palace but were then informed this was no longer taking place.

The government is having to erect a prefabricated conference centre at Bletchley Park because the building itself, which was chosen for its historic significance, is not fit for purpose. Bletchley Park was a base for World War II codebreakers.

The guestlist also risks being an embarrassment for Sunak. While senior tech executives such as Anthropic CEO Dario Amodei are coming and European Commission President Ursula von der Leyen is also likely to attend, so far Italy’s Giorgia Meloni is the only confirmed national head of government.

U.S. President Joe Biden was an early confirmed no-show, with the U.S. is sending his VP Kamala Harris instead. Canadian Prime Minister Justin Trudeau and Germany’s Olaf Scholz are also unlikely to attend.

French President Emmanuel Macron will not attend, an Élysée official told POLITICO. He will be in central Asia at the time of the summit.

Some countries, like Japan, are exploring the possibility of attending virtually.

The lack of star names could point to an uncomfortable reality for team Sunak: his summit has strained relations with allies. Some share widespread disquiet with the British summit’s focus on longer term risks of AI, such as the tech being used by terrorists to create bioweapons, at what critics say is the expense of current dangers, such as bias in AI. 

There are officials in the EU and the U.S. who see the British summit as an irritating distraction from their own arguably more substantial governance initiatives: the EU’s AI Act and a White House executive order — expected on Monday — which will both go much further than anything Britain’s summit can achieve, but that focus on much more mundane risks of the technology. 

Washington in particular has become increasingly uncomfortable with what it sees as British posturing on AI as a global leader, according to a U.K. official, an ex-U.S. official and industry figures familiar with the U.S.’s thinking. They are putting pressure on Britain to scale down its ambitions for the summit, according to the U.K. official and industry figures, and prefer existing governance initiatives like the G7 Hiroshima Process. The U.N. also set up its own advisory board that would recommend on how to govern AI on Thursday.

The U.S. is keeping a close eye on Sunak’s plan to use the summit to set up an AI Safety Institute, which aims to build on the work of Britain’s existing Frontier AI Taskforce in gaining access to and evaluating the risks of the models of leading — predominantly American — AI companies like Google DeepMind, Anthropic and OpenAI.

The U.K. plans for the institute to publish some of its findings publicly, but would reserve more sensitive national security intel to a smaller group of like-minded governments, a U.K. official with close understanding of the summit said. It is expected the U.S. will keep a tight leash on who gets access to that information, according to the ex-U.S. official cited above. 

Japan is also worried that Britain’s summit risks distracting from the G7 process, which it leads. “We are concerned there may be some duplication,” said a Japanese official.

Sunak’s bid for a legacy

China’s controversial invite to the summit — already attacked by some U.K. Conservatives — has only added to allies’ disquiet at the summit. But it is also where Sunak can break new ground, and perhaps even forge a legacy. 

Existing initiatives like the G7 Hiroshima Process don’t involve Beijing, which is keen to be more involved in global policymaking around AI. Last week, China launched its own “Global AI Governance Initiative,” which warned about the risks of AI falling into non-state actor hands, the need to develop human-centric AI and how to assess the technology — thus indicating areas of convergence with the West on high-level principles.

“This has the potential to show clearer lines on what the West and China can work on together on AI,” said Sihao Huang, an AI governance expert at Oxford University. “I think there is room for collaboration.” 

Similarly, Britain’s plan to invite China to a new AI research body modeled on the Intergovernmental Panel on Climate Change, could prove canny. Current AI research networks, such as the Franco-Canadian Global Partnership on AI (GPAI), have a much narrower membership than the U.K.’s proposed new body.

“We don’t currently have the governance bodies for AI set up, nor the links between the experts and the policymakers,” said Chowdhury. “If done well, this advisory body could really help with global policy making.”

Clea Caulcutt and Laura Kayali contributed reporting from Paris; Hans von der Burchard from Berlin; Mohar Chatterjee from Washington; Gian Volpicelli from Brussels; and Mark Scott, Joseph Bambridge, Tom Bristow and Laurie Clarke from London.