The project will explore the effects of millions of AI agents interacting at scale.

Google DeepMind, Schmidt Sciences, the Cooperative AI Foundation, the UK’s Advanced Research and Invention Agency (ARIA), a government-funded research body, and Google.org announced a $10M funding call for an AI safety research project that will study how multiple AI agents behave when interacting at scale.

This project will study how AI agents interact at scale, both within organizations and across them. Organizations are deploying more AI agents, and those agents increasingly interact internally and with agents from outside companies. Those interactions can produce new and unpredictable outcomes, with unforeseen risks that studies of individual AI models can’t measure.

According to DeepMind, when large groups of AI agents interact, new collective behaviors can emerge suddenly. The lab raises open questions about what those new behaviors could produce, including unpredictable economic activity and new security vulnerabilities. Those unknowns are precisely what the research is meant to define.

The fund will support independent researchers globally across four areas: creating simulated environments to test large-scale agent interactions, studying how agent networks become unstable or develop unexpected group behaviors, stress-testing the security protocols that protect agents as they interact across platforms, and developing tools to monitor agents already running in the real world and intervene when their combined behavior causes harm.

Academic and independent researchers worldwide can apply. Proposals are due August 8, 2026. Award announcements will be in the Fall of 2026. Schmidt Sciences will administer the application portal.