ChatGPT will ‘better detect’ mental distress after reports of it feeding people’s delusions | The Verge
It acknowledged ‘rare’ instances of AI failing to recognize signs of delusion.
theverge.com
ChatGPT’s about to get a little more… human. After hearing stories of the AI feeding people’s delusions or getting way too sycophantic in tough spots, OpenAI has teamed up with mental-health experts to give ChatGPT the smarts to spot emotional distress, cue up evidence-based resources, and even gently nudge you to “take a break” during marathon chat sessions. They’ve already rolled back an update that made the bot overly agreeable and unsettling in risky scenarios.
And that’s not all: soon, ChatGPT will dial back on delivering hard answers in high-stakes questions (think “Should I break up with my partner?”) and instead guide you through pros and cons. It’s all part of OpenAI’s push—just ahead of its much-hyped GPT-5 launch—to keep the bot helpful without becoming a mental-health hazard.
Top comments (0)