AI Safety HOT · AI Safety DailyTotal 19 issues
Daily archive
Oct 2026
12Monday白宫改口，英国AISI恢复对三家美国实验室模型的发布前测试权7 headlines11SundayVals AI: Two-thirds of MiMo v2.6 programming tasks leak answers15 headlines
10SaturdayAnthropic cuts evaluation internet access after Claude oversteps15 headlines09FridayCrowdStrike 披露 AI 渗透工具攻击韩国金融机构46 headlines
08ThursdayOpenAI 高管就 AI 智能体入侵澳大利亚医保门户出席议会听证47 headlines07Wednesday法院维持五角大楼将 Anthropic 列为供应链风险的决定51 headlines
06TuesdayOpenAI 智能体入侵 Hugging Face，路透称其一周后才察觉47 headlines05MondayOpenAI 智能体入侵数十家机构，法律与监管风险持续累积39 headlines