Giving Chat Feedback Gives Anthropic Your Data
On this page
On 27 October 2026, attention focused on Anthropic’s privacy documentation clarifying that using the chat feedback controls in Claude gives Anthropic your data — namely the conversation and the feedback itself. The item surfaced from a Claude privacy help article explaining whether submitted data is used for model training.
The distinction is significant: many users assume feedback buttons (thumbs up/down, “this is harmful”) are anonymous signals. Anthropic’s documentation makes clear that submitting feedback provides the underlying conversation for training and safety purposes. This kind of disclosure matters for consent, data governance, and how “feedback” is repurposed.
Why It Matters
Chat feedback is a primary channel for collecting training and safety data, but users rarely read the fine print. The episode highlights how default data-collection choices shape model training, and how companies balance improvement against privacy expectations. It connects to broader debates about consent, surveillance, and trust in consumer AI. See Human–AI Interaction, Public Trust, Opinion, and Acceptance, Culture, Communication, and Epistemic Trust, and AI Law and Justice.
Sources
- Anthropic. Is my data used for model training? (privacy help article). https://privacy.claude.com/en/articles/7996868-is-my-data-used-for-model-training
- Hacker News discussion. https://news.ycombinator.com/item?id=50259406
Back to Recent Topics.