Protecting people from harmful manipulation
As AI models get better at holding natural conversations, we must examine how these interactions affect people and society.
Building on a breadth of scientific research, today, we are releasing new findings on the potential for AI to be misused for harmful manipulation*, specifically, its ability to alter human thought and behavior in negative and decept...
The study simulated misuse scenarios to evaluate AI's ability to manipulate people in complex decision-making environments. Results showed that success in one domain does not predict success in another, validating the targeted approach to testing for harmful manipulation. The research also highlighted the importance of understanding both efficacy (whether the AI successfully changes minds) and propensity (how often it even tries to use manipulative tactics).
Patterns detected: ARC-0024 Ambiguity...