Google DeepMind and partners have announced a $10 million funding call aimed at studying the safety risks that emerge when large numbers of AI agents interact with one another — a problem the company believes is arriving faster than the research community is prepared for.
According to MIT Technology Review, the initiative is driven by concern about what happens as AI agents — software that can carry out tasks with minimal human oversight — move from research labs into mass-market deployment. Rohin Shah, who directs AGI safety and alignment research at Google DeepMind, is cited as a central voice behind the push.
The worry, as MIT Technology Review describes it, is not just about a single rogue agent but about emergent, hard-to-predict behavior that could arise when millions of agents begin operating and interacting across the internet simultaneously. Individual agents might be well-behaved in isolation; the concern is what happens in aggregate.
The Google DeepMind blog describes the announcement as a call for research proposals, positioning it as a collaborative effort with partners rather than purely internal work. The $10M is intended to draw in outside researchers to help map and address these risks before they become entrenched.
This matters because the AI industry is racing to deploy autonomous agents across everything from customer service to coding to financial tasks — and the safety science for how those systems behave at scale, interacting with each other rather than just with humans, is still in its earliest stages.