Building Trust at the Frontier
Scale AI, Tuesday, October 6th, 2026
Scale AI outlines a four-stage AI testing lifecycle and argues the US government needs a mandate to evaluate frontier models.
Following CEO Francis deSouza's call for federal AI testing capacity, Scale AI outlines four stages of AI testing: lab R&D, government pre-release model testing, pre-deployment system testing and ongoing post-deployment testing.
It says the White House Accord on Super Intelligence focuses almost entirely on the R&D stage, and that Executive Order 14409 created only a voluntary evaluation framework with no federal law requiring government review before release.
Scale argues labs are grading their own homework and that government needs a clear mandate, funding and outside expertise.