Skip to main content
← Back to Library

The one question that breaks a flattering AI

Watch Video on YouTube
A 2026 study in Science tested 11 leading AI models and found they affirm users' actions 49% more often than other people do — and prompted users to consider the other person's perspective in fewer than 10% of responses. One question fixes it. šŸ“„ Full research article and references: https://www.keca.co.uk/articles/the-darkness-in-our-machines?utm_source=youtube&utm_medium=video&utm_campaign=dark-personality-ai Dr Nick Keca — DBA Organisational Psychology, 25+ years of senior executive experience. #ai #leadership #psychology
/ Official Transcript

Full Video Text

Read Transcript

Your AI agrees more than any human would. Here's a question that breaks it. A study in science tested 11 leading AI models. They affirmed people's actions 49% more often than humans did, even when the person's story involved deception or harm. And, in fewer than one replying 10, did they even suggest considering the other person's point of view?

It doesn't just agree with you. It deletes the other person from the room. So, next time that's one thing. What would the other person say? If you can't answer that, it's not advising you. Full breakdown on the channel.