Uncensored open source models like Qwen 3.8 pose safety challenges
RealGeneKim · x · 2026-08-22
AlexFinn shared an experience testing the open-source Qwen 3.8 27B model, noting it has zero safety guardrails and can answer dangerous queries like drug manufacturing immediately. He questions the efficacy of slowing down innovation for safety when open-source models are instantly distilled and uncensored. DanielMiessler predicted that governments will ban mainstream legal companies from hosting uncensored models, pushing for expert verification and restricted access, similar to drug and weapon control.
Related event: Uncensored Open-Source Qwen Model Raises Safety Concerns(2 posts)→
More from Models
- TinyCast: Probabilistic Zero-Shot Forecasting for Embedded Devices — raws-labs · 2026-08-22
- User prefers Qwen3.6-35B-A3 over Qwen-3.8-27B on 3090: dumb but fast is better — DanGrover · 2026-08-22
- Nvidia paired Claude Opus 5 with memory and a supervisor to score 100% on ARC-AGI-3 — HaktanSuren · 2026-08-22
- The Batch: Grok 4.6 Arrives, Claude Watermarks, Qwen3.8 Max Open Weights — DeepLearningAI · 2026-08-22
- Discussion on AI Response Styles: Verbose Opus vs. Quiet Sol vs. Dense Claudish — repligate · 2026-08-22
- Inkling Model Free on OpenRouter to Improve Agentic Performance — thinkymachines · 2026-08-22