Money. Identity. Technology.
Follow the money. Find the decision.
Joint funding that shapes which safety infrastructure gets researched
Scaling AI Safety for a Multi-Agent World is a research competition advertising up to USD10 million, supported by Cooperative AI Foundation, Schmidt Sciences, ARIA and DeepMind. Its agenda includes coordination, monitoring and admission mechanisms for systems containing multiple AI agents.
Schmidt Sciences hosts the application route while the partners jointly consider applications; individual contributions and voting arrangements are undisclosed. Applicants can opt out of a Google.org route and remain eligible for other support. That route adds specified open-code, licensing, data and no-patent conditions which do not automatically bind every award. Later Schmidt funding depends on progress and reporting. The call describes one-to-two-year research, excludes commercial product development and one privileged controller of all agents, and expected awards in autumn 2026; the ceiling is not a paid-grants total.
Google.org optional funding route → Scaling AI Safety for a Multi-Agent World
Google route: Apache2.0 code/inventions, CC-BY4.0 specified IP within3months, public-domain residual IP, no patents and open data subject to privacy law. Edition/dispute rights unknown. Does not bind every partner award.
ARIA → Scaling AI Safety for a Multi-Agent World
Joint consideration; voting, recusal and individual contributions not disclosed. DeepMind is the named current brand, not an identified legal grantor or separate cash contribution.
Schmidt Sciences → Scaling AI Safety for a Multi-Agent World
Later years depend on progress/reporting; no guaranteed renewal or demonstrated clawback of paid funds. Call-specific duration1-2years governs over generic FAQ wording.
Scaling AI Safety for a Multi-Agent World → Agent infrastructure and admission-control research
Commercial product development and one privileged controller of all agents are excluded; research does not compel deployment.
Scaling AI Safety for a Multi-Agent World → Google.org optional funding route
Applicants may opt out while remaining eligible for Schmidt, ARIA and CAIF. No executed or paid grant established.
Schmidt Sciences → Scaling AI Safety for a Multi-Agent World
Portal/contact establishes administration, not exclusive award control.
Cooperative AI Foundation → named joint supporting party considering applications → Scaling AI Safety for a Multi-Agent World
Joint consideration; voting, recusal and individual contributions not disclosed. DeepMind is the named current brand, not an identified legal grantor or separate cash contribution.
Schmidt Sciences → named joint supporting party considering applications → Scaling AI Safety for a Multi-Agent World
Joint consideration; voting, recusal and individual contributions not disclosed. DeepMind is the named current brand, not an identified legal grantor or separate cash contribution.
ARIA → named joint supporting party considering applications → Scaling AI Safety for a Multi-Agent World
Joint consideration; voting, recusal and individual contributions not disclosed. DeepMind is the named current brand, not an identified legal grantor or separate cash contribution.
DeepMind → named joint supporting party considering applications → Scaling AI Safety for a Multi-Agent World
Joint consideration; voting, recusal and individual contributions not disclosed. DeepMind is the named current brand, not an identified legal grantor or separate cash contribution.
Schmidt Sciences → hosts application and administrative front door → Scaling AI Safety for a Multi-Agent World
Portal/contact establishes administration, not exclusive award control.
Scaling AI Safety for a Multi-Agent World → offers optional differentiated funding route → Google.org optional funding route
Applicants may opt out while remaining eligible for Schmidt, ARIA and CAIF. No executed or paid grant established.
Google.org optional funding route → conditions prospective awards on openness and AI Principles → Scaling AI Safety for a Multi-Agent World
Google route: Apache2.0 code/inventions, CC-BY4.0 specified IP within3months, public-domain residual IP, no patents and open data subject to privacy law. Edition/dispute rights unknown. Does not bind every partner award.
Schmidt Sciences → describes progress-dependent annual funding → Scaling AI Safety for a Multi-Agent World
Later years depend on progress/reporting; no guaranteed renewal or demonstrated clawback of paid funds. Call-specific duration1-2years governs over generic FAQ wording.
Scaling AI Safety for a Multi-Agent World → solicits research on coordination and admission mechanisms → Agent infrastructure and admission-control research
Commercial product development and one privileged controller of all agents are excluded; research does not compel deployment.