Use & Processing · Context
How Anthropic collects and uses personal information is in a separate Privacy Policy
The Terms point to Anthropic's Privacy Policy for how it collects and uses personal information. Unlike the Acceptable Use Policy, the Terms ask you to read the Privacy Policy but do not state that it is included in the Terms.
Supporting quotations
Please read our [Privacy Policy](https://anthropic.com/privacy), which describes how we collect and use personal information.
Not stated here
- The Privacy Policy is linked but not supplied in this source, so its collection, retention and sharing practices are not analyzed here.
Use & Processing · Company practice
Anthropic may train its models on your chats unless you opt out, with two exceptions that apply even after opting out
Anthropic may use Materials (your Inputs and Outputs) to provide, maintain and improve the Services and to develop other products and services, including training its models, unless you opt out of training through your account settings. Even if you opt out, Anthropic will use Materials for model training in two cases: (1) when you give Feedback about any Materials, or (2) when your Materials are flagged for safety review, to improve its ability to detect harmful content, enforce its policies or advance its safety research.
Supporting quotations
We may use Materials to provide, maintain, and improve the Services and to develop other products and services, including training our models, unless you opt out of training through your account settings.
Even if you opt out, we will use Materials for model training when: (1) you provide Feedback to us regarding any Materials, or (2) your Materials are flagged for safety review to improve our ability to detect harmful content, enforce our policies, or advance our safety research.
Qualifications & exceptions
- The opt-out is described as an opt-out of training; the sentence does not say it stops other listed uses such as providing, maintaining and improving the Services.
- Section 5 says rating an Output (for example thumbs up or down) stores the related conversation as Feedback, so a rating brings that conversation within the first exception.
Not stated here
- How long Materials are kept, and how flagging for safety review is decided, is not stated here; the Privacy Policy that may cover this is not supplied.