Google DeepMind Probes Risks of Millions of AI Agents Interacting Online
Google DeepMind is backing new research into a future where vast numbers of AI agents operate across the internet and interact with one another at scale. The concern is not just what a single autonomous system might do, but what could emerge when millions of agents begin coordinating, competing, sharing instructions, or responding to tasks with limited human supervision. Rohin Shah, who leads Google DeepMind’s AGI safety and alignment research, says the rapid move toward mass-market AI agents raises fresh safety questions that are not yet well understood. These systems are designed to complete tasks on behalf of users, but they may also follow directions from other agents or online services. Researchers hope to identify potential failure modes early, including unintended cooperation, manipulation, cascading mistakes, and risks that arise from complex agent-to-agent behavior before such tools become deeply embedded in daily digital life.