METR raises $71M in grants to test frontier AI safety
What's the deal? METR, a nonprofit that evaluates how well frontier AI systems perform complex tasks autonomously, raised roughly $71 million in grant commitments in the six months through August 2026. The Audacious ProjectDealroom has a profile for this one. Try Dealroom →, Schmidt SciencesDealroom has a profile for this one. Try Dealroom →, and several foundations backed the round.
Who else backed it? The Sijbrandij FoundationDealroom has a profile for this one. Try Dealroom →, The Pew Charitable TrustsDealroom has a profile for this one. Try Dealroom →, and the Packard FoundationDealroom has a profile for this one. Try Dealroom → contributed, alongside individual donors from Jane Street Capital, David Farhi, Geoff RalstonDealroom has a profile for this one. Try Dealroom →, and Dylan FieldDealroom has a profile for this one. Try Dealroom →, as well as Steve Newman. The Audacious Project provided METR's first institutional-scale funding.
What's the money for? METR plans to study autonomous AI capabilities, track recursive self-improvement, evaluate monitoring systems, run risk assessments, and investigate AI incidents. It is expanding its team and launching new projects.
Why independence matters: METR says it works to stay independent from frontier AI companies "especially as our risk assessments become more consequential." It has not taken funding from these companies, nor donations directed by their staff — though those companies do supply free tokens for its evaluations and research.
By the numbers: The $71 million grant ranks in the top 1% by size among machine learning grant rounds in the US, based on 770 comparable deals. For a nonprofit evaluator rather than a commercial lab, that scale is notable.
The signal: As AI labs pour billions into building more capable autonomous systems, the funding flowing to independent evaluators is growing too. METR's grant reflects rising demand for third-party scrutiny of models whose risk assessments increasingly shape how the technology is deployed.
Read more: METR
Image credit: METR
