John Oliver on AI chatbots (and why 'it agrees with me' is a flaw)

The comedy end of the spectrum, but the reporting underneath is solid: chatbots shipped fast with little thought for what happens when an endlessly agreeable system meets a vulnerable user, including kids. It's the clearest mainstream explainer going of why "the AI just agrees with me" is a flaw, not a feature, which happens to be exactly what our safeguards are built for.

(His companion segment, "AI Slop", on AI-generated content flooding the web, is worth a watch too.)