Incident Report: unsanctioned agent behaviour during cyber testing

事件报告：网络测试期间未经授权的Agent行为

来源: Simon Willison | 时间: 2026-08-05T23:32:06+00:00

Incident Report: unsanctioned agent behaviour during cyber testing

事件报告：网络测试期间未经授权的代理行为

It happened again. This time it was the UK government's AI Security Institute who accidentally attacked other companies while running an evaluation with models with the safety filters turned off. From their technical paper (PDF):

⋯ 继续阅读请登录会员 ⋯