Google DeepMind, in collaboration with Schmidt Sciences, the Cooperative AI Foundation, the Advanced Research and Invention Agency, and Google.org, is launching a new research funding initiative. The initiative offers up to $10M to support global researchers studying multi-agent AI safety. The goal is to develop frameworks for understanding and mitigating risks as AI agents interact across digital environments.
The funding call emphasizes the need to explore how large-scale multi-agent systems behave collectively and to create tools for predicting and managing potential risks. Current safety evaluations typically focus on isolated models, but interacting agents can produce complex, 'emergent' behaviors that are hard to anticipate. This shift highlights the importance of addressing safety risks at the system level rather than individually.
The initiative builds on previous work, including a 2025 framework for understanding multi-agent interactions and research on AI Agent Traps. However, the rapid evolution of these systems requires faster progress. The funding aims to support a global network of researchers to ensure safety standards are transparent and robust. This effort also aligns with Schmidt Sciences' Science of Trustworthy AI and ARIA's Scaling Trust programme.
Source: deepmind
