Dan Lahav, co-founder and CEO of Irregular, is now tracked in Turing Tree. Irregular builds realistic cyber evaluations and defenses designed to help frontier AI labs understand and reduce security risks before deployment.
Lahav co-authored a framework with RAND for protecting advanced model weights from theft. The work maps dozens of attack paths and proposes escalating security levels for threats ranging from opportunistic criminals to nation states.
Before founding Irregular, he was a researcher in IBM's AI Research division and worked on Project Debater, an early system designed to debate people on complex topics.