Summits | Meetings | Publications | Research | Search | Home | About the G7 Research Group
|
Summits | Meetings | Publications | Research | Search | Home | About the G7 Research Group |
|
G7-G20 Centred Global AI Governance
John Kirton, G7 and G20 Research Groups
September 18, 2026
Presentation for the Istituto Affair Internazionale–Global Governance Program seminar on “Multilateralism and Digital Governance,” at the Lawson Institute for Sustainability, University of Toronto, on September 18, 2026. Co-sponsored by the International Relations Program, Bill Graham Centre for Contemporary International History, Canadian International Council and the NATO Organization of Canada (NAOC).
Artificial intelligence requires governance, global governance and global summit governance, by the world’s most powerful leaders of the most powerful countries and international institutions.
It requires governance by governments, rather than just self-regulation by industry and inventors, because it critically affects national security and almost anything governments are responsible for.
It requires global governance, because its essential human and physical components flow easily across international borders, and their use affects all countries in the world.
And it requires global summit governance, because only heads of state and government have the authority and the responsibility to decisively shape how the inventions of their powerful industries and best scientists will unfold and be used, for better or worse.
These leaders know this. They are now competing but also cooperating to lead such governance. The major players are the United States, China, the United Kingdom, France, the European Union, the International Telecommunications Union, the United Nations, the G7 and the G20.
Which one will and should lead?
I argue it is the G7, which has pioneered the global summit governance of AI since 2018, and should do so in the years ahead. But it should do so with the influential support of the bigger, broader, G20, to represent the rest of the world.
G7 summits have led the world in AI governance by creating consensus among the US, Europe, Japan, Canada and their invited guests.
Indeed, AI and its global governance began in Canada through the G7. A Canadian scientist invented AI – my colleague Geoffrey Hinton at the University of Toronto, who was awarded the Nobel Prize in physics in 2024 for this. Another Canadian pioneer was Yoshua Bengio at the Université de Montréal.
The Canadian government invented the global governance of AI when it hosted the G7’s Charlevoix Summit in 2018. Its leaders made the G7’s first commitment on AI, and a total of 24 on them. US president Donald Trump fully agreed with every one. Twenty-one of them addressed AI’s risks. Eight highlighted its rewards, including for gender equality in AI.
Then G7 leaders made only one AI commitment at Biarritz, France, in 2019, none under Trump at Washington DC in 2020, one at Cornwall in the UK in 2021, and none at Elmau in Germany in 2022. But Japan’s Hiroshima Summit in 2023 produced six, and Italy’s Apulia Summit in 2024 made a new high of 25.
Canadian prime minister Mark Carney’s Kananaskis Summit in 2025 produced a new high of 47. Trump again agreed with all. They deliberately shifted the emphasis to AI’s rewards. Only eight recognized risks.
Most recently, French president Emmanuel Macron’s Evian Summit on June 15–17, 2026, made only two commitments on AI. Its separate statement on minors’ digital safety had 10 commitments. Macron devoted the summit’s final session to a discussion with CEOs from G7 members on digital technologies, including AI.
At the summit’s end Macron called for stronger international regulation of AI. He said: “We all want to harness the opportunities, productivity gains, changes and improvements in health and digitalization that AI can bring. But no one can any longer ignore the impact of AI on our democracies and on our societies.” Thus, international regulation and closer cooperation among democratic countries was needed, to keep advanced AI “from falling into the hands of authoritarian regimes or individuals, who could threaten our cybersecurity or our societies.” Thus, several democratic powers would build a cooperation platform to create common AI standards, including for cybersecurity.
Moreover, since 2018, G7 members have complied highly with their priority commitments on AI, at an average of 92%, well above their all-time, all-subject average of 78%. So, what G7 leaders promise together at their summits on AI, their governments then quickly do back home.
But to get the job fully done, the G7 needs the bigger, broader G20 that includes China.
G20 summits have now made 22 commitments on AI, joining the G7’s 103. They started at Buenos Aires in 2018 with one, added one at Osaka in 2019 and another at Riyadh in 2020, three at Rome in 2021, six at New Delhi in 2023 and five at Rio de Janeiro in 2024. Johannesburg in 2025 made five, without Donald Trump or Chinese president Xi Jinping there.
Members’ compliance from 2021 to 2023 averaged 86%, well above the G20’s overall, all-subject average of 71%. These commitments all pointed to the benefits of AI.
For his G20 summit in Miami on December 14–15, 2026, Trump – as one of his three priorities – put technology, with AI at its core.
To prepare this priority, the G20 got off to a slow start. At the meeting of finance ministers and central bank governors in Ashville, North Carolina, on August 31–September 1, China vetoed several phrases and passages in the draft declaration prepared by the United States. The resulting “Chair’s Statement” had only one commitment on AI, and another related to AI. China objected to neither.
The first one asked the Financial Action Task Force to ensure the effective implementation of FATF standards on virtual assets, given the “evolving threat emanating from the scam compounds and exploitation of AI by fraudsters.” The second one said central banks will aim to rigorously distinguish gains in productive capacity from changes in demand.”
Immediately after, G20 ministers of innovation, meeting in Chapel Hill, North Carolina, on September 1–2, issued a full consensus “G20 Innovation Ministerial Statement.” Among its 33 commitments AI had 19, including 17 ambitious ones. Here all members agreed to use AI for education, employment, development and trade, to promote “secure and trustworthy” AI, to use and “enhance international cooperation,” to “advance The G20 AI Prosperity Objectives” and to “utilize The AI Prosperity Compact” to achieve these goals.
The China-guided BRICS is no substitute as a global summit governor of AI. For its recent summit in New Delhi on September 11–12, 2026, produced only three, very general, commitments on AI – one each on AI governance, cybersecurity and green economic growth.
So, what should the G7 and G20 do now?
Three things – use AI for health, the environment and defence.
They flow from ITU secretary general Doreen Bogdan-Martin’s important reminder that “AI breakthroughs are now unlocking the ability to detect diseases earlier, personalize learning, mitigate disaster risks, optimise agriculture and energy systems, and strengthen water and food security.” These benefits are the immediate personal priorities for all citizens and thus their governments in all G7 democracies and elsewhere.
Thus, the first step is to finally link AI to health, for drug discovery, disease diagnosis, personalized treatment and much else. Rapid, effective, personally targeted, affordable health care is key to keeping people alive and controlling soaring government deficits and debts.
The second step is to use AI for the environment, starting with wildfire, hurricane, and flood detection and response, and forecasting for other extreme weather events. They should then add low and clean electricity, energy, water consumption and noise pollution from AI data centres, so communities will more easily accept them in their backyard.
The third proposal is AI for defence, especially for cybersecurity, Ukraine, Taiwan and against Iran.
Thank you.
 — 
|
This Information System is provided by the University of Toronto Libraries and the G7 Research Group at the University of Toronto. |
|
Please send comments to:
g7@utoronto.ca This page was last updated September 28, 2026. |
 |
 |
 |
 |
All contents copyright © 2026. University of Toronto unless otherwise stated. All rights reserved.