Google DeepMind and Schmidt Sciences: $10M to Secure AI
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
An Ambitious Call for Projects on AI Security
Google DeepMind, in collaboration with Schmidt Sciences, the Cooperative AI Foundation, the Advanced Research and Invention Agency (ARIA), and Google.org, has announced a research grant call worth up to $10 million. This funding is aimed at researchers worldwide to explore the security of multi-agent artificial intelligence systems.
For the past decade, efforts have focused on improving individual AI models to make them more efficient and secure. However, with the rapid evolution of technology, the emphasis is now shifting towards the interactions between millions of AI agents developed by different organizations. These interactions require robust security frameworks to ensure safe and predictable exchanges.
Understanding Emerging Collective Behaviors
The call for projects focuses on studying the collective behaviors of large-scale multi-agent AI systems. Currently, the tools to predict, measure, and monitor these transitions are insufficient. Interactions between autonomous agents can lead to complex and unpredictable behaviors, posing new security challenges.
It is crucial to understand how these behaviors emerge and how they can influence economic activity or create new risks. The goal is to develop frameworks to manage these behaviors at the system level. This initiative aims to address the "invisible" security risks that arise when independent systems interact across different networks.
Expanding Research on Multi-Agent Security
Although foundational frameworks for multi-agent security exist, the rapid evolution of these systems necessitates immediate and large-scale expansion of research. Our 2025 research established a framework for understanding these interactions, while our recent work on AI agent pitfalls explores the vulnerabilities agents face in adversarial environments. However, the increasing complexity of interactions exceeds current models.
This call for projects aims to accelerate progress by supporting a global network of independent researchers. A diverse community is essential to ensure that security standards are transparent and robust for all. This effort also advances the mission of the Science of Trustworthy AI and AI Agents programs at Schmidt Sciences, as well as ARIA's Scaling Trust program, which seeks to unlock new forms of cyber-physical multi-agent coordination.
Four Priority Areas for Proposals
Researchers are invited to submit proposals in four key areas:
- Sandboxes and Testbeds: Build realistic and reproducible environments to assess, compare, and accelerate progress across all areas of multi-agent security. This includes virtual markets, simulated ecosystems, and multi-organizational workflows.
- Agent Network Science: Understand the properties relevant to the security of interacting agent populations, including studying how collective capabilities emerge and develop, how networks fail or become volatile, and how to detect dangerous and unexpected properties at the population scale.
- Strengthening Agent Infrastructure: Test protocols for identity, reputation, and commitment that ensure secure interactions between agents across different platforms.
- Monitoring and Control: Develop methods to monitor deployed agent populations and mitigate large-scale collective harm.
How to Participate
Interested researchers are invited to review the call for proposals and submit their projects by August 8, 2026. Winners will be announced in the fall of 2026, marking a significant step towards a safer future for multi-agent AI systems.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.