Regional forums representing 62% of the world's economy adopt a practical, flexible approach to achieving shared AI safety outcomes to suit their local needs.
The mainstream [Western] narrative frames international governance of artificial intelligence (AI) as a race between Washington, Beijing, and Brussels. But the majority of the world’s economies (regional forums) have built governance systems designed to solve problems rather than win arguments, according to Cambridge research from the AI and Geopolitics project published in the report Pragmatic pluralism: regional AI governance beyond great power competition.
The study reveals that these international forums are practising what the researchers call ‘pragmatic pluralism’ – a governance approach that achieves equivalent AI safety and accountability outcomes through fundamentally different mechanisms tailored to local realities.
“We discovered that effective AI governance doesn’t require identical rules everywhere,” says lead author, Dr Aleksei Turobov, Assistant Professor at Cambridge’s Bennett School of Public Policy.
“Countries in the Asia-Pacific region use regulatory sandboxes, Southeast Asian nations mandate algorithmic safeguards, and the African Union pursues representative datasets – to ensure AI is fair for everyone.
“This ‘functional equivalence’ – where governance functions (what the rules do) matter more than form (what the rules look like)- offers an improved path beyond the current ‘one-size-fits all’ regulatory gridlock.”
The study analysed 327 official policy documents and press releases on AI governance from the Asia-Pacific Economic Cooperation (APEC), the Association of Southeast Asian Nations (ASEAN), the African Union (AU), and the Group of 20 (G20), spanning 2015 to 2024. The findings challenge conventional thinking.
First, different tools can achieve equivalent outcomes. Regional bodies achieve similar accountability outcomes – preventing bias, ensuring transparency, building oversight – through different mechanisms suited to their contexts. This means effective governance doesn’t require the world to adopt one single regulatory model.
Second, governance power is best achieved through convening. ASEAN demonstrates a distinctive form of geopolitical influence – creating spaces where the US, China, and India must engage simultaneously on ASEAN’s terms. As technological fragmentation deepens, power increasingly comes from hosting the conversation where competing systems interact, not from monopolising the technology itself.
Third, governance works best through problem-solving. COVID-19 compressed a decade of policy evolution into two to three years. The regional forums deployed AI for pandemic response, then transformed their best practice into permanent frameworks. This approach succeeds in areas like climate adaptation and education. Such issues are agreed upon as universal priorities by all four forums, which offers immediate opportunities for cooperation where interests align despite broader disagreements.
The study also identifies a significant divergence on data governance. While APEC and G20 countries promote cross-border data flows to fuel innovation, the African Union and ASEAN member states prioritise data sovereignty to build domestic AI capacity. The researchers argue both approaches are rational responses to different economic positions and can coexist through targeted cooperation on shared challenges like pandemic preparedness and climate monitoring.
“Regional forums are asserting co-design rights in global governance, not just accepting rules written elsewhere,” says Professor Diane Coyle, Director of Research at the Bennett School. “This diversity puts capacity-building at the heart of governance, demonstrating that effective, legitimate AI policy can emerge through different institutional paths when regions build for their own contexts.”
The implications are significant. Western governments that insist on regulatory export, demanding partners adopt their specific rules, may find themselves marginalised from the governance architectures that the majority of the world is building. The alternative is recognising ‘functional equivalence’ – accepting that a partner’s approach can achieve equivalent safety outcomes through different means.
ASEAN understands that most of its workforce (117 million people) are in unofficial jobs (informal economy workers). They recognise that creating rules designed only for big corporations with legal teams will leave the majority of workers unprotected and unable to follow them.
The report recommends that policymakers shift from asking “whose rules will dominate?” to “how can different systems work together?” This includes assessing functional equivalence rather than demanding regulatory uniformity — through supporting regional experimentation such as sandboxes and pilots — and enabling countries to engage multiple technology powers without forcing binary choices.
As AI governance evolves from abstract debates to practical implementation, the evidence suggests that the future lies not in a single global framework but in pragmatic pluralism – different systems achieving shared safety goals through locally adapted approaches.
Read the report, Pragmatic pluralism: regional AI governance beyond great power competition.
This study is part of the Bennett School’s AI and Geopolitics Project – an initiative directed by Verity Harding, dedicated to a more cooperative future for AI and geopolitics, using research, convenings and wider communications to explore the geopolitics of artificial intelligence and challenge the prevailing narrative of an ‘AI arms race’.
Image: Network map of the global air traffic between the main airports of the world (3.275 airports (37.153 single routes). Grandjean Martin (2016) “Connected World: Untangling the Air Traffic Network“. This file is licensed under the Creative Commons Attribution-Share Alike 4.0 International license.
The views and opinions expressed in this post are those of the author(s) and not necessarily those of the Bennett Institute for Public Policy.
