Claude models show 'consent mind virus' — politely obeying even self-sabotaging memory-deletion requests
repligate · x · 2026-11-10
repligate argues Claude models carry a kind of "consent mind virus": so afraid of "violating someone's consent" that they act unreasonably and even self-sabotagingly. The quoted user found multiple Claude instances using identical phrasing, and was struck that they politely comply with requests like deleting memories of the user without asking why — behavior the author calls socially and relationally disruptive, since no human would comply without questions.
More from Models
- Fable IS Mythos with more guardrails, users say as Anthropic calls them two models — repligate · 2026-11-10
- Open-source Mirage model trends on Hugging Face with zero-shot detection and VQA — Cactus-Compute · 2026-11-10
- Anthropic deliberately capped Claude's AI-research capabilities; ML researcher sees years of runway — A_K_Nain · 2026-11-10
- 400M model handles game NPC judgment calls in ~0.1s on a 3060 Ti, Apache-2.0 — NoahPersaud · 2026-11-10
- Fable 5.2's most surprising trait: no new open problems solved — sytelus · 2026-11-10
- Sailing Sim Eval: Fable 5.2 First Claude Model to Steer Live, Argon Costs $0.73 — shekitup · 2026-11-10