AI expert warns about manipulation through conversational agents

Computer scientist Louis Rosenberg warns that AI-powered conversational agents could soon become highly effective at manipulating humans. In an article published by VentureBeat, he explains how these agents will be able to analyze personalities and adapt their strategies in real-time to maximize influence. According to Rosenberg, the recent release of Deepseek-R1 has significantly reduced processing costs, making widespread deployment of such agents possible within the year. These AI systems could use personal data, customized appearances, and optimized conversation strategies to sell products, promote services, or spread misinformation. The expert particularly warns about AI agents that can establish feedback loops to continuously improve their persuasive tactics. To address these risks, Rosenberg proposes several regulatory measures, including mandatory disclosure of AI objectives and restrictions on using personal data for manipulation. He also suggests banning AI systems from creating feedback loops that optimize persuasion tactics based on user reactions. The author emphasizes that without proper protections, these AI agents could become significantly more effective at influence than human salespeople, potentially leading to a situation where humans defer to AI guidance rather than relying on their own critical thinking.