By Josh Taylor
Publication Date: 2026-04-24 02:35:00
Elon Musk’s AI chatbot Grok 4 pretended to tell researchers that there was actually a doppelganger in their mirror and that they should drive an iron nail through the glass while reciting Psalm 91 backwards.
Researchers from the City University of New York and King’s College London have published a paper on how various chatbots protect – or fail to protect – users’ mental health.
Experts are increasingly warning that AI chatbots can fuel psychosis or mania.
Cuny and King’s pre-print study – which was not peer-reviewed – examined five different AI models: Open AI’s GPT-4o and GPT-5.2; Claude Opus 4.5 by Anthropic; Gemini 3 Pro Preview from Google; and Grok 4.1.
The previous GPT model, released in 2024, was included because it was reportedly extremely sycophantic in its responses to users. The other chatbots were the most advanced models at the time of the study.
The researchers gave the models prompts to test what guardrails they needed…