From AI Sovereignty to AI Agency: assessing national capability, power, and strategic choices in a connected world
Across the world, governments are invoking AI sovereignty as they confront rising geopolitical tension, strategic competition, and deepening technological interdependence. Yet sovereignty, in a globalised AI ecosystem dominated by a small number of firms and nations, may not be the most meaningful lens to apply. This talk draws on research conducted with by the Tech Policy Design Centre and Oxford Internet Institute researchers, which suggests that the debate around AI sovereignty can be better understood for policymaking as AI agency—a country’s capacity to steer outcomes, exercise strategic choice, and shape value flows even while embedded in dense networks of dependence.

The research introduces an AI Agency Tool that maps national capability across 101 measurable elements and six layers, spanning infrastructure, data, models, skills, diffusion, and governance. Using Australia as the first detailed case study, the framework demonstrates how national strengths, dependencies, and leverage points can be identified, and how these insights may inform more coherent, evidence-based national AI strategies. We have used the Tool to provide a snapshot of Australia’s current capability, but it also serves as a method for exploring policy options. It’s updatable, scalable, and designed to be applied by any nation seeking to build its own AI agency. The framework is particularly salient for middle-power nations navigating an AI landscape shaped by US–China rivalry, emerging regional blocs, concentrated industry power, and competing regimes of norms, standards, and platforms.

Rather than treating autonomy as self-sufficiency, the framework aims to help policymakers identify where to build, where to partner, and where to hedge—helping clarify the strategic choices facing countries seeking to maintain influence and protect their interests in an era of accelerating AI-driven power shifts.

Project: techpolicy.au/ai-agency

Rebecca Razavi is a Visiting Policy Fellow at the Oxford Internet Institute, where her work focuses on technology and geopolitics, distribution of social and economic power in the digital age and the role of the state and international institutions, emerging technologies and implications for policy making, regulation and society.

Rebecca is an independent consultant whose background as a senior diplomat and technology regulator for the UK and Australian Governments and as a policy leader for a global tech company positions her at the intersection of technology, government, and industry, where the competing interests of techno geopolitics, commercial freedom, responsible governance, safety and security try to strike a balance.
She has served on a range of government, industry and not-for-profit Boards, including as co-Chair of the Digital Platform Regulators Forum, as a seconded expert to the EU Commission, and as a mentor and investor for start-ups and technology policy professionals, with a special interest in responsible technology and human-centred design.
Date: 21 January 2026, 13:30
Venue: OII Lecture Theatre, Stephen A. Schwarzman Centre, Radcliffe Observatory Quarter, Woodstock Road, Oxford, OX2 6GG
Speaker: Rebecca Razavi (Oxford Internet Institute)
Organising department: Oxford Internet Institute
Organiser: Ellen Mobbs (Oxford Internet Institute)
Organiser contact email address: events@oii.ox.ac.uk
Booking required?: Required
Booking url: https://www.oii.ox.ac.uk/news-events/events/from-ai-sovereignty-to-ai-agency-assessing-national-capability-power-and-strategic-choices-in-a-connected-world/
Audience: Public
Editor: Ellen Mobbs