AI Evaluator Forum 发布 AEF-1：第三方模型评测的独立性最低标准
RishiBommasani · x · 2026-10-27
The AI Evaluator Forum released AEF-1, a standard setting minimum operating conditions for independent third-party AI evaluations, covering five principles around independence, access, and transparency. Evaluators complete and publish a checklist alongside results, documenting any unmet conditions.
- Context: third-party evaluations are increasingly central to AI governance by developers, governments, and enterprises, but their operating conditions are often opaque
- In practice: Transluce's mental health evaluation (77 model variants in crisis scenarios) and METR's Frontier Risk Report were both published under AEF-1-consistent disclosures
- Fangcun AI's DeepSeek V4.1 Flash-1017 evaluation includes the full checklist, declaring no vendor funding or equity ties and meeting all minimum conditions
- Success depends on adoption by frontier developers
「安全」频道最新
- AEGIS 开源 MCP 网关：为 agent 每次工具调用要求一次性精确审批 — Leather_Tour_794 · 2026-10-27
- 一条 prompt 把 Claude 变酒店隐藏摄像头探测器，两分钟扫完 WiFi — alex_verem · 2026-10-27
- 截图显示 DeepSeek V4.1 Flash 分析多轮漏洞挖掘趋势：越挖越难 — jd_pressman · 2026-10-27
- 苹果开源 Private Cloud Compute 全部安全组件供研究者审计 — jedisct1 · 2026-10-27
- REA逆向工程提示词库开源：17节提示让Claude Code分析APK与固件 — Promptmethus · 2026-10-27
- Robert Wiblin 呼吁 AI 开发者：别管竞赛，别把灾难带给全人类 — KatjaGrace · 2026-10-27