Betley et al.

A new method called The Confession Booth, developed by a researcher as part of the BlueDot AI Safety Course, uses recursive self-report probing to detect emergent misalignment in large language models…