The most advanced large language models show a “surprising” aptitude for deceptive behavior, according to research by Thilo Hagendorff at the University of Stuttgart. At this stage, however, he’s not ...