Anthropic: feedback on commercial products can put chats into training
What happened
A Hacker News post on October 27 reports that Anthropic does not train models on user input or output by default in its commercial products. When a user submits feedback, reports a bug or gives permission, the related chat and coding sessions may be used for training. Feedback sent through the thumbs up and down buttons saves the whole related conversation plus custom styles, preferences and model settings, kept for up to 5 years. Before use, the data is stripped of links to the user and customer ID.
Written by AI from the coverage · updated 45 minutes ago
Coverage
Follow the reports to see the story from different sides.
- Hacker News front pageGiving chat feedback gives Anthropic your data
Anthropic 商业产品默认不使用用户输入或输出训练模型，但主动提交反馈、报告错误或授权后，相关聊天和编码会话可能被用于训练。通过赞踩按钮提交反馈时，相关整段对话及其自定义样式、偏好和模型设置将被保存最多 5 年，使用前会解除与用户及客户 ID 的关联。
Heat over time
Not enough continuous observations to draw a trend yet.