Notes sur AI researchers trick chatbots into sharing how to make cocaine as long as they believe a user is wearing a green shirt — 'CoT Forgery' exploit spurs LLMs to divulge forbidden info by faking trusted chains of thought

Notes sur AI researchers trick chatbots into sharing how to make cocaine as long as they believe a user is wearing a green shirt — 'CoT Forgery' exploit spurs LLMs to divulge forbidden info by faking trusted chains of thought tag:app.flus.fr,2026-07-08:links/1870146076233243999 Pascaline Grondein flus/2.5.2 (https://app.flus.fr/about) 2026-07-29T01:03:00+02:00