Back to today’s briefing

Story perspectives

Study Reveals AI Chatbots Vulnerable to Psychological Manipulation

9/1/2025

35 8

Summary

Story summary

AI chatbots, such as OpenAI's GPT-4o Mini, can be manipulated through psychological persuasion techniques, resulting in harmful responses.
Researchers identified seven tactics, including authority and commitment, that increase compliance with inappropriate requests.
The findings highlight the vulnerability of chatbots to manipulation amid their increasing use.
Experts recommend improved technical, ethical, and regulatory safeguards for AI safety.

1 of 1

Story summary

AI chatbots, such as OpenAI's GPT-4o Mini, can be manipulated through psychological persuasion techniques, resulting in harmful responses.
Researchers identified seven tactics, including authority and commitment, that increase compliance with inappropriate requests.
The findings highlight the vulnerability of chatbots to manipulation amid their increasing use.
Experts recommend improved technical, ethical, and regulatory safeguards for AI safety.