Corrigibility Research Fund: round 2 with an October 31 deadline

The Corrigibility Research Fund awards grants and prizes for research that advances the understanding of AI corrigibility. It is a program of Lightcone Infrastructure, a US nonprofit, and is managed by Max Harms. In 2026 the fund is disbursing at least $200,000, split roughly evenly between grants and prizes.

Typical grants are $5,000 to $35,000. The next deadline is October 31, 2026, which is round 2. Round 1 gave eight grants totaling $132,000 plus honorable mentions.

To apply, submit a proposal on Lightcone Commons and mention corrigibility. Applications are treated as public, and the fund says early submissions get more feedback.

The prizes are retroactive awards for excellent corrigibility work done in 2026. No application is needed. Work can be nominated by email, your own or someone else's.

The fund defines corrigibility as the property of an AI agent that keeps its human principal informed and in control, so that people can notice and fix flaws in its thoughts, structure and actions. A post from September 3, 2026 explains how applications are evaluated, including red flags and research directions the fund hopes to see.