Investments in Safety Research for Multi-Agent AI Systems

Google DeepMind, in collaboration with Schmidt Sciences, the Cooperative AI Foundation, the Advanced Research and Invention Agency (ARIA), and with support from Google.org, has announced a new competition for technical research with a total budget of up to $10 million for scientists worldwide. As AI technologies scale, a new era is emerging: millions of AI agents created by various organizations will interact in digital environments, communicating, negotiating, and conducting transactions. It is essential to ensure the safety and predictability of such interactions. The competition aims to study the collective behavior of large-scale multi-agent systems and create frameworks for understanding and mitigating potential risks. Currently, there are no tools for forecasting, measuring, or monitoring emergent effects arising from the interaction of autonomous agents. Although basic foundations for multi-agent system safety exist, rapid evolution requires immediate expansion of research. The competition supports scientific programs by Schmidt Sciences and ARIA and invites researchers to submit proposals in four priority areas: sandboxes and testbeds, agent network science, strengthening agent infrastructure, and oversight and control. The application deadline is August 8, 2026, with winners announced in fall 2026.