The collaborative funding call accepts applications until August 9, 2026.
1/ New funding call on multi-agent AI safety. How do we stay safe when millions of agents, built by different developers, interact in shared digital spaces? The failures that matter most there, like cascades and collusion, are invisible when you test one model at a time.
No Digg Deeper questions have been answered for this story yet.
@GoogleDeepMind @ARIA_research @coop_ai @Googleorg 3/ Technical call, open globally to researchers, teams, and institutions. Tier 1 up to $300K, Tier 2 $300K to $1M, over 1 to 2 years. Proposals due Aug 9. If you work on this, apply and share it with your network. #AISafety #MultiAgentSystems
2/ Schmidt Sciences, @GoogleDeepMind, @ARIA_research, the Cooperative AI Foundation (@coop_ai, whose board I chair), and @Googleorg are funding technical research on how agents cooperate and fail at scale. Areas include testbeds, agent networks, infrastructure, and oversight.
2/ Schmidt Sciences, @GoogleDeepMind, @ARIA_research, the Cooperative AI Foundation (@coop_ai, whose board I chair), and @Googleorg are funding technical research on how agents cooperate and fail at scale. Areas include testbeds, agent networks, infrastructure, and oversight.
1/ New funding call on multi-agent AI safety. How do we stay safe when millions of agents, built by different developers, interact in shared digital spaces? The failures that matter most there, like cascades and collusion, are invisible when you test one model at a time.
@GoogleDeepMind @ARIA_research @coop_ai @Googleorg 4/ Apply here: https://buff.ly/eirXYIH
@GoogleDeepMind @ARIA_research @coop_ai @Googleorg 3/ Technical call, open globally to researchers, teams, and institutions. Tier 1 up to $300K, Tier 2 $300K to $1M, over 1 to 2 years. Proposals due Aug 9. If you work on this, apply and share it with your network. #AISafety #MultiAgentSystems
🎉
1/ New funding call on multi-agent AI safety. How do we stay safe when millions of agents, built by different developers, interact in shared digital spaces? The failures that matter most there, like cascades and collusion, are invisible when you test one model at a time.
