Announcing the Corrigibility Research Fund

LessWrong reports that Max Harms is announcing a new corrigibility research fund — with at least two hundred thousand dollars in grants and prizes for twenty twenty-six. Corrigibility, in AI safety terms, means building AI systems that keep humans in control rather than trying to directly embed human values into the system itself. The fund, housed at Lightcone Infrastructure and directed by philanthropist Peter McCluskey, will split the money roughly evenly: half in traditional grants with application deadlines in August and October, and half in retroactive prizes for excellent corrigibility work published this year. Typical grants are expected to range from five to thirty-five thousand dollars. The announcement signals support for corrigibility research, an area Harms describes as deeply neglected despite growing interest in AI safety more broadly.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton