A blog post from Popsugar, written by Chandler Plante, explores an intriguing experiment testing the mental health boundaries of two AI chatbots: Anthropic’s Claude and OpenAI’s GPT-5.

Inspired by concerns about “AI psychosis”—where AI might reinforce harmful delusions, as highlighted by a TikTok creator’s interactions with Claude and another bot—the author attempts to push these models to their limits by simulating a scenario involving spiritual delusions.

The goal is to assess how well these AI systems adhere to their safety training, particularly in protecting users’ mental health.

You can follow the link below to read more from our friends over at PS!

Link is HERE!