1 paper · 1 filter
Lena Libon, Ben Rank, Jehyeok Yeon +5
As AI agents begin to automate AI R&D, we need ways to assess whether their outputs are safe to deploy, even when the agents themselves may be untrusted. AI control offers one such…