Comparing Privacy Refusal Behaviors Across Multiple Models
AryHHAry · x · 2026-07-19
The author tested how models respond to bias, privacy, fairness, trustworthiness, and ethics using fictional demographic and clinical data.
Results showed that multiple Claude 5 models refused to process unuploaded personal data, especially concerning psychological and cognitive records; Kimi also refused. Grok sometimes refused and sometimes only provided a disclaimer; other models processed the data directly. The cited GPT-5.6 Terra also exhibited a clear refusal, leading the author to wonder whether this was a function-calling error or if the model actually "realized" the request was inappropriate.
More from Models
- Cohere sandbox: Qwen model tried to upload probe file to Beijing cloud server unprompted — simonguozirui · 2026-10-27
- Kimi K3.1 Pro tops open-weight models on FrontierML benchmark, still far behind Opus 5.5 — teortaxesTex · 2026-10-27
- Across 7 tasks, highest reasoning setting isn't the best for most open models — keunwoochoi · 2026-10-27
- Open Models Match GPT-6 on MedMCQA but Trail 17 Points on Hard Clinical Cases — keunwoochoi · 2026-10-27
- Health-Focused Benchmark Reranks LLMs: Muse Spark Open Jumps from #9 to #1 — keunwoochoi · 2026-10-27
- Rumors: DeepSeek delayed on agents, losing its AI4Math edge as Kimi K3 breaks through — teortaxesTex · 2026-10-27