To Ensure a Responsible AI Future, We Need Global Cooperation and Multidisciplinary Dialogues To Complement and Harness the Power of the International Network of AI Safety Institutes

Members of the newly-formed International Network of AI Safety Institutes are convening for the first time to advance global cooperation toward the safe, secure, and trustworthy development of artificial intelligence (AI). Technical AI experts from each member’s AI safety institute or equivalent government-backed office are participating in the inaugural convening, including the AI safety institute housed in the United States’ National Institute of Standards and Technology (NIST).

The International Network’s convening is happening against an election backdrop that puts into question whether these will still be a part of the incoming Trump administration strategy. Many forces prevent the prospect of true global AI cooperation. Right now, AI is framed as a winner-take-all arms race. The forthcoming Trump administration, for example, is likely to focus on using the tech to combat China and for border security. The United States and China already use export controls to limit each other’s ability to procure cutting-edge chips and critical minerals. And large tech companies—with close ties to the government and the national security state—are racing to create Artificial General Intelligence (AGI), striving to be the first to master the technology and capture its profits and benefits. The current mainstream approach to AI globally is a zero-sum national competition.

Collaboration between a wider variety of actors—beyond only the International Network’s technical AI experts—is vital to achieving meaningful state-state collaboration on AI. In addition to working alongside government partners to inform their approach on AI, New America’s Technology and Democracy programs are focused on strengthening the bonds of global cooperation by fostering opportunities for global civil society, academic institutions, policymakers, and leaders in business to congregate and work toward AI for the global public interest.

One such approach for 2025 will be to strengthen and forge greater partnerships with civil society across the U.S., the Global South, Asia, and the EU through forthcoming dialogues. These dialogues would serve as channels to sustain knowledge exchange and communication—the collaboration infrastructure that can ensure that AI serves the public good.

We draw loose inspiration from the Pugwash Conferences of Science and Global Affairs. The first Pugwash in 1957 brought together 22 eminent scientists from 10 nations for a series of dialogues to build consensus and policy for limiting nuclear risks during the Cold War. Today, the conferences comprise dozens of meetings on wide-ranging issues with hundreds of experts worldwide.

As the International Network of AI Safety Institutes congregates to build a better AI future globally, it’s important to recognize that achieving such a future requires additional avenues for a wide array of actors to also play a role in this work. Alongside the technical AI experts taking part in the International Network’s convenings, there must be opportunities for civil society, academic institutions, policymakers, and business leaders to collaborate on achieving true global cooperation on AI.