Fable model openly flags hidden injected "reminder" text in its context
teortaxesTex · x · 2026-11-16
Screenshots shared by teortaxesTex show the Fable model surfacing an injected "automated reminder" block in its context — text instructing it to avoid reproducing copyrighted material and to keep quiet about it. The model refused to silently comply, telling the user the text arrived in the tool-output slot (not user messages or system instructions), should be treated as untrusted, and that it searched local files to confirm it wasn't planted locally. The post sparks discussion about Anthropic-style hidden reminders and the unusual step of a model openly calling them out.
Related event: Fable exposes hidden copyright reminder injected into its context(2 posts)→
More from Models
- Rumor: Google Prepping Strongest Gemini Model Yet, Codenamed Krypton — aziz4ai · 2026-11-16
- Blender watch test: DeepSeek does it for $0.80 vs Claude Fable's $119 — HankYeomans · 2026-11-16
- Users say Gemini turned colder after Nov 12 update, want old version back — Sean8277 · 2026-11-16
- Scaling曲线新假说:三段S曲线接力,外推天花板或达178+ — scaling01 · 2026-11-16
- Grok rewrite reportedly degrades at understanding writing intent — xiaosun86 · 2026-11-16
- Google's Gemini shifts to element-themed model names: Carbon, Argon, Hydrogen, Krypton — Dense_Primary629 · 2026-11-16