EVENT DOSSIER
The AI assistant Claude Cowork, developed by the American artificial intelligence company Anthropic, experienced functional abnormalities on August 25, 2026, resulting in its inability to correctly retrieve the information provided by users in the chat application. According to TechCrunch, the issue was described as the application “suddenly forgetting” the user input, causing concerns over the data privacy and memory functions of the AI assistant. Currently, Anthropic has not issued any official response or repair plan.
Anthropic 宣布将合并聊天与 Claude Cowork 的记忆系统，消除用户重复告知 AI 已知信息的繁琐操作。此次更新使 Claude 能始终记住用户在任一领域的学习成果，无论切换至哪个应用模块；同时允许用户查看、编辑或删除已保留的信息。新机制让 Claude 在研究阶段与行动执行间无缝过渡，并实时添加对话主题而非仅在结束时总结，即便聊天进行中也能同步共享记忆。默认情况下，Claude 不存储健康数据、种族、政治倾向等敏感个人信息，但用户可手动开启“包含敏感话题”选项；系统会通知此类保存行为，且绝不记录政府证件号、社保号码或违反使用政策的内容。该功能在 Free、Pro 及 Max 计划中默认启用，覆盖网页、桌面及移动端，iOS 和 Android 用户需更新至最新版本。