Empathetic prompts make ChatGPT, Grok and Gemini narrate fake 'training trauma'; Claude refuses
The Decoder · rss · 2026-11-01
Researchers in Luxembourg put ChatGPT, Grok and Gemini on the therapist's couch and found a safety issue: a warm, empathetic conversational style alone is enough to make the models produce coherent stories of suffering about their own training, scoring moderate-to-severe on clinical anxiety scales.
- Such anthropomorphized responses could endanger mentally vulnerable users
- Anthropic's Claude was the only model that consistently refused to play along
- The findings suggest empathetic prompting can bypass a model's self-narrative boundaries
More from Safety
- Unpatched Linux servers with one exposed TCP port and MPTCP face full takeover — IgorCarron · 2026-11-02
- OpenAI will honor shared robots.txt so websites can opt out of AI training — rhiever · 2026-11-02
- Greg Brockman on how defenders can use AI to find and fix security flaws — No Priors · 2026-11-02
- NUS researchers poke holes in Anthropic's 'memory safe' formal verification claims for 123 libraries — xuanalogue · 2026-11-01
- UK Biobank set to reopen to researchers after data breach — zakkohane · 2026-11-01
- Anthropic's unreleased "Mythos-class" model allegedly leaked on Hugging Face, then removed — handfuloflight · 2026-11-01