AI at WorkClaude privacy: how Anthropic uses chats for training
Anthropic explains in its privacy center when your Claude chats may be used to improve its models. Here is a plain summary of what it said.
What we read
The Anthropic Privacy Center article "How Do You Use Personal Data in Model Training?", dated 16 March 2026, which we read on 13 November 2026, and the plans page the same day. The article is about the consumer products (Claude Free, Pro, Max, and Claude Code when used on those plans). It says that the commercial products (Claude for Work and the API) are covered elsewhere.
When your chats may be used
The article says Anthropic may use your chats or coding sessions to improve its models if:
- you choose to allow it;
- your conversations are flagged for safety review (they may be used or analysed to improve the ability to detect and enforce the Usage Policy, including training models for safeguards);
- you have explicitly provided materials, for example through the thumbs up/down feedback button; or
- you have explicitly opted in in another way, for example by joining a Trusted Tester programme.
The plans page showed, in its comparison table, "Opt-out" in the model training row for the Free, Pro, Max and Ultra plans, and "No model training on your content by default" for the Team plan. The two pages describe the choice in different words, so open the privacy settings in your own account and read what it says for you.
What Anthropic says it does with the data
The article says that where you have allowed use of chats, they are automatically de-linked from your user ID (for example your email address) before use, and that tools are used to filter or obfuscate sensitive data. It also says Anthropic does not use personal data from training to contact people, build profiles, sell or market, or sell the information. It states that you can adjust your privacy settings at any time.
Your rights
The article points to the Privacy Policy, which describes the right to request a copy of your personal data and to object to processing or request deletion, and it says these rights are limited and that actioning requests about the training dataset is complex.
Practical advice
- Open Settings, then Privacy in your Claude account and read the option about helping improve Claude. Choose what you are comfortable with.
- Use an incognito chat (the plans page lists incognito chats for all individual plans) for anything sensitive.
- Do not paste identity numbers, passwords or other people's private data. See the checklist page.
- Avoid feedback buttons on chats that contain anything private, because feedback can be used.
- For client or student work, check whether your workplace has an organisation plan with different terms.
What this page cannot tell you
Settings screens and policies change. We did not test the settings screen of the app, so the exact menu names are for you to check. The article also notes that its Privacy Policy does not apply where Anthropic acts as a data processor for business customers.
Are my chats used for training by default?
The pages we read describe it as a choice you can set, with other cases (safety review, feedback). Read the setting in your own account to see how it is set for you.
Does deleting a chat remove it from training data already used?
The article says rights concerning the training dataset are limited and complex, so assume that anything already used cannot be pulled back.
Sources
- How do you use personal data in model training?, Anthropic Privacy Center (dated March 16, 2026)
- Plans & pricing, Claude
Read on 13 and 14 November 2026. Pages change; open the original before you rely on a statement.
This is an independent guide. It is not connected with OpenAI, Anthropic, Google, Microsoft or any other AI company; product names belong to their owners. Statements about companies are taken from their own help pages as read on 13 and 14 November 2026. The prompts are our own wording and were not run against every tool. Not legal, financial or professional advice.