June 11, 2026
Google DeepMind, with Schmidt Sciences, the Cooperative AI Foundation, the UK’s Advanced Research and Invention Agency, and support from Google.org, opened a research funding call of up to $10 million for work on the safety of multi-agent AI systems. Applications are open until August 8, with decisions expected in the autumn.
The partners said the call targets a gap in current safety work: most evaluations test a model in isolation, but they expect a near future in which many AI agents, built by different organizations, interact, negotiate, and transact with one another. Interacting agents can produce collective behaviors that are hard to predict from any single model, and the funders framed the goal as studying those dynamics before such systems are widely deployed.
According to DeepMind, the call invites proposals in four areas: sandboxes and testbeds for evaluating multi-agent systems; the science of agent networks, including how collective capabilities emerge and how networks fail; agent infrastructure such as identity, reputation, and commitment protocols; and methods for overseeing and controlling deployed agents. The Cooperative AI Foundation said funding runs in two tiers, up to $300,000 and up to $1 million, open to academic and independent researchers.
The program is small relative to the sums spent building the agents themselves, and it funds research rather than any standard or requirement. Its significance is as a signal: several of the organizations shaping agent deployment are now paying to study the failure modes of agents acting together, a problem that has so far attracted far less work than the alignment of models on their own.