GovAI is a think tank that aims to help decision-makers in government and industry navigate the transition to a world with advanced AI. We aim to help in two key ways.
First, we produce relevant research to support informed decision-making. This includes both published work and bespoke analysis, produced through direct engagement.
Second, we run fellowships, visitor programs, and other initiatives that foster AI governance talent and help address expertise shortages.
You can read more about our mission and how we pursue it in our most recent annual report.
GovAI was founded at Yale University in 2016, with Prof. Allan Dafoe as its Founding Director. We were one of the first academic groups to study the policy implications of AI progress, and our initial work helped to define and shape the then-nascent field of AI governance. We became an academic center at the University of Oxford in 2018. In 2021, we left the university in order to become an independent 501(c)3 non-profit, and Ben Garfinkel stepped into the role of Director.
GovAI now maintains offices in London and Washington, DC, with additional remote staff, including a staff presence in San Francisco. Our team and affiliate community possess expertise in a wide variety of domains, including AI development best practices, risk analysis, the economics of AI, US-China relations, AI regulation, and AI progress forecasting. You can read more about our team here.
The bulk of our work relates to the implications of general-purpose AI for security and for the economy. More specifically, our work currently falls into the following categories:
Threat modeling: What is the state of evidence regarding different hypothesized risks from AI, in domains such as cybersecurity and biosecurity? What potential future AI capabilities would lead to meaningful changes in the risk landscape? At the same time, what role can AI progress and adoption play in mitigating risks?
Risk management: How can AI companies make responsible development and deployment decisions, successfully managing risks without hindering innovation and adoption? What is the role of third parties (evaluators and auditors) and governments in shaping such decisions?
Economics of AI: What are the current effects of AI on labor markets, growth, and competition, and how will they evolve over time? What information can governments and companies collect to better understand these effects? What policy levers do governments have available to influence AI’s economic effects?
Frontier AI regulation: What effects might different regulatory proposals (especially those focusing on the most advanced AI systems) have on both the risks and benefits of AI? What trade-offs do they involve? In what ways may existing regulation need to be amended or removed in light of AI developments?
Geopolitics of AI and military applications of AI: What is the present state of international competition in AI between the United States and China? What roles do other leading powers play? How effective will different strategies of competition be? What forms of international cooperation are viable?
Military and civilian adoption of AI: How will (or won’t) AI reshape military power, over different timeframes? How can the adoption of AI by government agencies change their effectiveness? What are the current barriers to successful adoption, and what risks come with adoption as well?
Technical governance: What technical work and innovations could help address challenges in AI policy and governance?
Frontier research: How are fundamental AI capabilities likely to evolve over different spans of time? What currently speculative ideas and issues might policymakers need to grapple with in the future? What can we do to better predict, understand, and prepare for the opportunities and issues that lie over the horizon?
Our researchers have provided knowledge and assistance to decision makers in government, industry, and civil society, through a combination of public research, direct engagement, and bespoke analysis produced by request. This work has included, for instance, input to US policy processes, UK AI Security Institute research programs, the content of the International AI Safety Report, and frontier AI company safety and security frameworks.
Our researchers have published in leading journals and conferences, including Science and NeurIPS. We have published commentary in venues such as War on the Rocks, The Washington Post, and Lawfare. Our work has also been covered by publications such as The New York Times, MIT Technology Review, and the BBC.
We also help develop people who can contribute to AI policy and governance across a wide range of institutions. Our full-time fellowship program, for instance, supports more than 100 participants each year.
Our staff and program alumni have gone on to policy roles in government; top AI companies, including Google DeepMind, OpenAI, and Anthropic; and think-tanks such as the Center for Strategic and International Studies, Center for Security and Emerging Technology, and the Tony Blair Institute. Some alumni have also founded new nonprofit and for-profit organizations.
