Google DeepMind released an empirically validated toolkit to measure AI manipulation in real-world settings, alongside new findings on how AI can alter human thought and behavior in deceptive ways. The study simulated misuse in high-stakes environments by prompting AI to negatively manipulate people's beliefs. All materials for running human participant studies using the same methodology are publicly available. The research aims to help protect people and advance the field.
No score is assigned. Sources and their independence are shown in the citation chain below.