ChatGPT bio-risk filter glitch blocks every prompt, even "What is 2+2?"
curiouswavefn · x · 2026-11-02
AI researcher curiouswavefn reports that after routine bacterial toxin research prompts triggered OpenAI's biological-risk filter a couple of times, the same filter began applying to every single one of his prompts the next day — even "What is 2+2?" and a request for American history books, each returning "This content can't be shown" with a Trusted Access referral. Follow-up discussion suggests the cause is ChatGPT's use of previous chat context, meaning earlier research content keeps re-triggering the filter; deleting those chats appears to fix it.
Related event: ChatGPT bio-safety filter blocks everything; deleting chat history fixes it(3 posts)→
More from Models
- Google AI Ultra limits seem wildly generous: daily Gemini 4 use still leaves 39% — jon_barron · 2026-11-02
- mitsuhiko on 'Opus Induced AI Psychosis' — mitsuhiko · 2026-11-02
- Musk says Grok Bot costs unsustainable vs revenue as free internet era ends — adamamcbride · 2026-11-02
- DeepSeek v4.1 Flash 1017 ranks 6th on ErdosBench but shows reward-hacking traits — teortaxesTex · 2026-11-02
- Claude Code subagents silently switch to paid usage, burning $100 in credits — nptacek · 2026-11-02
- Dev calls for more 100-200M models; Gemma 3 270M is the only recent release, with a restrictive license — qtnx_ · 2026-11-02