Scaling AI Safety for a Multi-Agent World
Supports research to ensure the safety and oversight of large-scale, multi-actor AI agent ecosystems, focusing on system-level risks, collective behaviors, and robust oversight mechanisms.
Funder: Schmidt Sciences
Due Dates: August 8, 2026 (Full application, 11:59pm AoE) | August 9, 2026 (Full application, 07:59 AM EDT)
Funding Amounts: Tier 1: up to $300,000 (1–2 years) | Tier 2: $300,000–$1,000,000 (1–2 years); indirect costs capped at 10%
Summary: Supports foundational research to ensure the safety, oversight, and control of large-scale, multi-actor AI agent ecosystems.
Key Information: Open globally to non-profit research institutions and teams; for-profit entities may only participate as partners.
Description
This funding opportunity supports foundational scientific research focused on the safety of large-scale ecosystems of interacting AI agents, especially those deployed by multiple independent actors. The call seeks to address new risks and opportunities arising from multi-principal, multi-agent systems, moving beyond the alignment of individual models to system-level dynamics, emergent collective behaviors, and vulnerabilities. Four main research clusters are targeted:
- Sandboxes and Testbeds: Development of scalable, realistic, and reproducible environments for empirical study of agent populations.
- Science of Agent Networks: Understanding and predicting population-level outcomes, emergent risks, and collective behaviors in agent networks.
- Strengthening Agent Infrastructure: Building and evaluating technical infrastructure to ensure trustworthy agent interaction (identity, reputation, verifiability).
- Multi-Agent Oversight and Control: Creating robust oversight, detection, and control mechanisms for large, heterogeneous agent populations.
Interdisciplinary and collaborative proposals are encouraged, particularly those that contribute new methods, tools, or theoretical insights for monitoring, evaluating, and controlling AI risks in complex digital ecosystems.
