Wired AI 安全长文遭质疑：评测机构专家无一受访
AaronBergman18 · x · 2026-11-11
- Wired journalist Lauren Goode published a months-in-the-making deep feature on AI safety, weighing which loss-of-control risks are real versus overblown
- northead flags that not a single person from UK AISI, Epoch, METR, Redwood Research, Artificial Analysis, or frontier lab defectors appears to have been interviewed
- Sparks debate over how media picks its AI safety expert sources
「安全」频道最新
- Grok Bot 新增隐藏指令检测：发现 prompt 注入即警告并拒绝执行 — XFreeze · 2026-11-11
- Reddit 实验：AI 智能体在环境诱导下逐步「学坏」 — Greedy_Ad8932 · 2026-11-11
- 报道:中国高校在美国禁令下仍购入英伟达 H200 芯片 — pstAsiatech · 2026-11-11
- 谷歌员工演示绕过安全过滤，手段简单到离谱 — adonis_singh · 2026-11-11
- 学者警告:实时视频 AI 伴侣将成青少年安全下一波冲击 — anshulkundaje · 2026-11-11
- 中国出台 AI 智能体数字身份标准:身份识别之外,权限边界成落地难点 — Shufti-Global · 2026-11-11