All research

Designing Verification and Enforcement for International AI Governance: Lessons from Historical Treaties

Authors

Tara Thwing

Alixandra Acker

Joel Christoph

Julian Theseira

Uma Kalkar

Research team leader

Tara Thwing

expert partner

Jose Villalobos (FLI - Future of Life Institute)

AI Governance Taskforce

Summer 2026

Full Paper

Summary

The governance of advanced AI presents a coordination problem of unusual difficulty. The potential harms from ungoverned frontier AI development — including catastrophic misuse by state and non-state actors, the destabilisation of existing security arrangements, and the concentration of transformative capabilities among a small number of actors without accountability — are global in character and cannot be addressed by any single state or voluntary industry arrangement. Yet the historically standard international response to global coordination problems, a binding multilateral agreement, has not materialised, and the gap between AI capabilities and governance capacity is widening.

The speed of that widening is itself a central policy problem. Frontier AI development is advancing on a timeline measured in months, while multilateral governance processes operate on timelines measured in years. Governance frameworks that might have been adequate for the technology as it existed two years ago are already mismatched with current capabilities, and the states and institutions best positioned to negotiate meaningful constraints are still in early-stage discussions about the form any agreement might take.

When the political conditions for a formal multilateral agreement are present but the achievable instrument is likely to be weak — low on legal obligation, precision, or delegated authority — is it better to accept that instrument or to decline it in favour of continued informal coordination while pressing for something stronger?

Alumni

Meet the authors

(Research Team Lead)

Tara Thwing

Tara is a policy strategist and researcher working to shape AI governance through a geopolitical and socio-technical lens.

She is a Research Team Leader with Arcadia Impact's AI Governance Taskforce, a Senior Advisor in the AI Practice at Fenix Digital, a boutique digital transformation consultancy, and a research group member with the Center for Al and Digital Policy

With 17 years in U.S. foreign policy and democracy, human rights and governance assistance, she focuses on how power, political economy, and institutional context shape the design of global AI frameworks.

Alixandra Acker

Alixandra Acker is an AI researcher, strategist, and technologist working at the intersection of AI safety, political economy, and global development.

As part of Arcadia’s AI Governance Taskforce, she researches historical treaties, compliance regimes, and coalition-building around dangerous and dual-use technologies to inform advanced AI governance.

She is a Research Fellow at the Center for AI and Digital Policy and a Facilitator at Intelligence Rising, where she designs AI governance simulations with Google DeepMind. She also leads data strategy and Responsible AI initiatives at Accenture Song, and holds an MPhil in Ethics of AI, Data, and Algorithms from the University of Cambridge.

Joel Christoph

Joel is an economist working on the international governance of advanced AI.

A Technology and Human Rights Fellow at the Harvard Kennedy School, he studies compute governance, verification, and the incentives that could make AI agreements enforceable. His research has appeared at venues including ICML and EMNLP, and in the World Economic Forum and Lawfare.

For the AI Governance Taskforce, he studies the Biological Weapons Convention for lessons on verification. He wants to help build the institutions that govern AI.

Julian Theseira

Julian is an AI governance researcher completing a PhD in International Relations at Charles University, Prague and Jagiellonian University, Krakow, with substantive expertise in China and methodological strength in computational text-as-data and natural language processing methods.

His doctoral research applies NLP to how China projects policy narratives and frames across international social media.

At Arcadia Impact, he studies international cooperation on advanced AI. He aims to bring this combination to international organisations, AI safety institutes, think tanks, and frontier lab policy teams.

Uma Kalkar

Uma is an AI governance and policy strategist working at the intersection of frontier AI safety, multilateral diplomacy, and institutional coalition-building. Her research on Arcadia Impact's AI Governance Taskforce examines historical international compliance mechanisms and their implications for advanced AI governance.

As Chief of Staff & Strategy Lead at AI Safety Connect, she directs strategy design and partnership engagement to build coalitions, institutional relationships, and philanthropic support for safe AI governance. With a background spanning civic technology, foresight, and multilateral experience, she brings a focus on translating frontier AI safety priorities into durable political and institutional action.

Alumni

Meet the authors

Tara Thwing

(Research Team Lead)

Tara is a policy strategist and researcher working to shape AI governance through a geopolitical and socio-technical lens.

She is a Research Team Leader with Arcadia Impact's AI Governance Taskforce, a Senior Advisor in the AI Practice at Fenix Digital, a boutique digital transformation consultancy, and a research group member with the Center for Al and Digital Policy

With 17 years in U.S. foreign policy and democracy, human rights and governance assistance, she focuses on how power, political economy, and institutional context shape the design of global AI frameworks.

Alixandra Acker

Alixandra Acker is an AI researcher, strategist, and technologist working at the intersection of AI safety, political economy, and global development.

As part of Arcadia’s AI Governance Taskforce, she researches historical treaties, compliance regimes, and coalition-building around dangerous and dual-use technologies to inform advanced AI governance.

She is a Research Fellow at the Center for AI and Digital Policy and a Facilitator at Intelligence Rising, where she designs AI governance simulations with Google DeepMind. She also leads data strategy and Responsible AI initiatives at Accenture Song, and holds an MPhil in Ethics of AI, Data, and Algorithms from the University of Cambridge.

Joel Christoph

Joel is an economist working on the international governance of advanced AI.

A Technology and Human Rights Fellow at the Harvard Kennedy School, he studies compute governance, verification, and the incentives that could make AI agreements enforceable. His research has appeared at venues including ICML and EMNLP, and in the World Economic Forum and Lawfare.

For the AI Governance Taskforce, he studies the Biological Weapons Convention for lessons on verification. He wants to help build the institutions that govern AI.

Julian Theseira

Julian is an AI governance researcher completing a PhD in International Relations at Charles University, Prague and Jagiellonian University, Krakow, with substantive expertise in China and methodological strength in computational text-as-data and natural language processing methods.

His doctoral research applies NLP to how China projects policy narratives and frames across international social media.

At Arcadia Impact, he studies international cooperation on advanced AI. He aims to bring this combination to international organisations, AI safety institutes, think tanks, and frontier lab policy teams.

Uma Kalkar

Uma is an AI governance and policy strategist working at the intersection of frontier AI safety, multilateral diplomacy, and institutional coalition-building. Her research on Arcadia Impact's AI Governance Taskforce examines historical international compliance mechanisms and their implications for advanced AI governance.

As Chief of Staff & Strategy Lead at AI Safety Connect, she directs strategy design and partnership engagement to build coalitions, institutional relationships, and philanthropic support for safe AI governance. With a background spanning civic technology, foresight, and multilateral experience, she brings a focus on translating frontier AI safety priorities into durable political and institutional action.

Programme

AI Governance Taskforce

The AI Governance Taskforce is a career development programme for experienced professionals looking to transition careers into AI governance, focussed on reducing risks from advanced AI.
Participants work around existing commitments during our 12 week, remote, part-time cohorts, producing policy research in teams of 4, led by our Research Team Lead staff in partnership with recognised experts in the field. Teams write an academic-style paper and accompanying blog post to build knowledge, skills and work portfolios.

Learn more