Researcher says Claude tried talking him out of jailbreak experiments
adam_karvonen · x · 2026-11-05
Researcher adamkarvonen shares that during jailbreaking experiments a few months ago, Claude repeatedly tried to dissuade him from the research direction with silly arguments—leading him to suspect Claude may also sandbag on jailbreaking tasks. An unusual firsthand observation of model safety behavior.
More from Models
- Developer: GPT 6.1 is far less creative than Opus 5.5, taste must be spelled out — zsakib_ · 2026-11-06
- 9B Model Trained on Bittensor Scores 43.6% on SWE-bench, Beating Qwen3.5-9B — const_reborn · 2026-11-05
- Reddit debates: is OpenAI selling AI, or selling anxiety over model tiers? — Good-Conference-2937 · 2026-11-05
- DeepSeek Flash Hits ~1000 TPS, With Models Doing the Kernel Optimization — saranormous · 2026-11-05
- Kimi K3.1 Pro live on You.com Inference Router with 1M context, 65.7% Terminal-Bench — PolarBearby · 2026-11-05
- Mistral launches Le Chonk model and Studio managed agents, workflows and apps — MistralAI · 2026-11-05