Investing in multi-agent AI safety research
DeepMind Blog · 2026-06-10
Google DeepMind, Schmidt Sciences, and partner organizations announce a $10M research funding call targeting safety risks that emerge when large populations of independent AI agents interact across digital environments.
Appears in
Extraction
Topics: multi-agent-aiai-safetyresearch-fundingemergent-behaviorai-governance
Claims
- Google DeepMind and partners are launching a $10M funding call for global researchers focused on multi-agent AI safety.
- Current safety evaluations analyze AI models in isolation and cannot predict emergent behaviors that arise when large populations of agents interact.
- Millions of AI agents built by different organizations will soon interact across digital environments, creating novel collective safety risks that no single lab can address alone.
- The funding call targets four research areas: sandboxes and testbeds, the science of agent networks, agent infrastructure integrity, and oversight and control at scale.
- The application deadline is August 8, 2026, with awardees expected to be announced in Autumn 2026.
Key quotes
When large groups of AI agents interact, new collective behaviors and capabilities can emerge suddenly. Currently, we lack the tools to predict, measure and monitor these transitions.
We are at a critical juncture where the complexity of multi-agent interactions is outpacing existing safety models.
No single lab can solve multi-agent safety alone.