A gap in the market: what the UK's G20 presidency could do on AI

Prime Minister Andy Burnham has said he wants to put AI at the heart of the UK's G20 presidency. The UK will take over from the US after the Miami summit in December, culminating in the leaders' summit in Manchester in November 2027. The Prime Minister has called for a single set of global principles and standards on transparency and safety. The Chancellor's finance track, meanwhile, will look at what AI means for growth and jobs.

We asked the room not to "admire the problem", but to take a view on what the UK’s presidency should actually do. 

Views ranged widely, but one thread ran through much of the discussion: there is a gap in the market for bold leadership on AI that focuses on the challenges of responsible innovation, set by neither of the two global AI leaders (the US and China). Many felt the UK is well placed to take that on. 

What we heard

Balance two major challenges faced by all G20 countries: risk mitigation and driving adoption

Two conversations are running at once. 

Far more countries now worry about AI risk: not only the long-term, existential AI safety risks, but increasingly closer-term threats like financial stability. At the same time, many countries have the opposite worry: not enough AI, and not enough investment, capacity or access to get it.

The G20 leans towards a narrower set of countries than the UN. But the outcome of the Presidency will need to speak to both worries. 

The fears in the room about what the presidency might get wrong were practical: a summit moment with a sweeping declaration and no delivery plan behind it. Yet another set of standards that aren’t possible to enforce. Months of good conversation that end in a final meeting arguing over commas. The hopes were practical too: governments, not technology companies, setting the agenda after years heavily-weighted in the reverse. And a plan the UK public can believe in, one that shows the grown-ups are in charge.

Some worried that risk driven by advanced AI development at the frontier would crowd out the risks most G20 countries actually face, from financial stability to resilient infrastructure. Others argued that capability has raced so far ahead of safety that the balance needs resetting. And one fear went largely unanswered: rising military use of AI with sophisticated cyber, biological and kinetic applications. 

Work to the UK’s strengths in responsible adoption 

With the US and China in the driving seat, much global focus has been on the race to build the most powerful models. 

But for a hospital, a bank or a small business, the most urgent need is for a reliable and locally appropriate one. Adoption is where the gains land, and it is the challenge most countries actually face.

Many saw this as the UK's natural ground. It joins up the safety track and the economic track instead of leaving them to compete, and it plays to UK strengths in assurance and standards. Frontier companies deeply understand only a handful of markets, while the UK's reach and depth of engagement, including through the Commonwealth and broad diplomatic ties, goes much wider. 

Adoption is not the race currently in the headlines. But it is where most of the value may be. 

The G20 is a critical market, and can act like one

One framing resonated with those in the room: G20 members are AI's biggest buyers, regulators and insurers. That gives them leverage, if they choose to use it.

The most concrete lever is procurement. Governments buying AI infrastructure can require suppliers to meet governance standards comparable to their own. Insurance could do similar work for open-source models, which are otherwise hard to tax, regulate or monitor, because markets that price risk can carry safety standards with them. 

Behind both sits the plumbing any standards regime depends on, such as incident reporting, verification technology and know-your-customer checks. The finance track could look to fund this. 

Navigate global principles, with focus on where they can be enforced 

Nobody wanted another unenforceable code. Yet today's standards largely reflect what frontier companies chose to focus on: cyber, biological and chemical risk. Countries also need meaningful standards which protect borders, civil resilience and its public services. Standards don’t yet reflect these wider concerns. 

Here, the room split. Some reached for models from other sectors, such as the Financial Stability Board, the International Atomic Energy Agency or international aviation regulation. Others doubted a new international regulator would ever succeed. For most, the realistic route was to build on what already exists. The OECD AI principles already sit at G20 level, and the EU has set leadership through its rules. So G20 principles need to build from those as well as UN baselines, and leave enforcement to a network of national regulators. 

There was optimism that the UK could effectively broker this: start the diplomacy now to find even a couple of concerns the US, China, the EU and India share, and build consensus from there. 

Building the evidence on AI’s economic implications is a winnable priority, and a shared G20 interest

The evidence on AI's effect on jobs, productivity and growth is thin and contradictory. The usual fiscal story is gloomy: lost jobs, lost income tax, higher welfare bills. But if AI lifts growth even modestly and raises public sector productivity, it could help with the fiscal problem rather than exacerbate it. The catch is who the gains go to, and whether they can be taxed.

Building the capacity to measure these impacts could be “low-regret” and achievable, with scope for the G20 to learn from the UK’s own approach through its AI Economics Institute. Shared observatories on adoption, jobs and safety could take this further, as could G20 governments collecting evidence from their own services. 

The topic of young people came up repeatedly, suggesting that forging a future for young people could be another feature of the G20 agenda. Each firm cutting graduate hiring is acting rationally, but together, they remove the training ground that produces the experienced people everyone will need later. New productivity gains and sectors may emerge. But in the immediate term, governments can invest in work that can’t be automated in the short term, in sectors such as care and plumbing. 

Handle the "middle powers agenda" with care

The UK inherits an agenda shaped by 4 Global South presidencies: Indonesia, India, Brazil and South Africa. Those started from discussions on growth, debt and inequality, not security alone. India's focus on digital public infrastructure showed it is possible to start the conversation outside both the US and China.