When Does Randomized Oversight Align AI Agents That Can Conceal? -- by Joshua S. Gans, Richard T. Holden

Thanks to Refine.ink, ChatGPT 6 Astra, and Claude Opus 5.5 for valuable research assistance, and Gans thanks the SSHRC for funding. Responsibility for all errors remains our own. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.

Joshua Gans has drawn on the findings of his research for both compensated speaking engagements and consulting engagements. He has written the books Prediction Machines, Power & Prediction, and Innovation + Equality on the economics of AI for which he receives royalties. He is also chief economist of the Creative Destruction Lab, a University of Toronto-based program that helps seed stage companies, from which he receives compensation. He conducts consulting on anti-trust and intellectual property matters with an association with Keystone Strategy and his ownership of Core Economic Research Ltd. He also has equity and advisory relationships with a number of startup firms. Joshua is also a co-founder of All Day TA.