AI Brief
How confessions can keep language models honest | AI Brief