Your chats may train models unless you opt out; exceptions remain
Anthropic may use your Inputs and Outputs to improve models unless you opt out; feedback and safety-flagged or reported material remain exceptions.
Saved report
Loading the findings and their saved sources.
Policies retrieved Oct 19, 2026 About these sources
← All reportsConsumer impact report
AI & Dev
Claude is an AI assistant for chat, coding, and agentic tasks, with optional connections to third-party services.
anthropic.com ↗U.S. adult consumer · plan and state unspecified
Originally analyzed Report updated
Explore what these saved policies say about your data, content, account, and payments.
Start here. Open a finding to see the exact words.
Anthropic may use your Inputs and Outputs to improve models unless you opt out; feedback and safety-flagged or reported material remain exceptions.
If you cancel a paid subscription, prepaid fees generally are not refunded, though your access continues through the paid term and app refund rules may differ.
The terms also allow suspension or termination without notice if Anthropic believes you breached them or must act to comply with law.
Anthropic says an individual conversation disappears from your history immediately and is automatically deleted from its back end within 30 days.
This applies when you delete an individual conversation.
When you connect to or interact with a third-party service, Claude may send it your inputs, outputs, or instructions; some integrations can retain access.
This applies when you connect to or interact with third-party services; ongoing access applies only to some integrations or features.
Review the third party's privacy policy before enabling an integration.
Anthropic may use your Inputs and Outputs to improve models unless you opt out; feedback and safety-flagged or reported material remain exceptions.
The general permission applies unless you opt out; feedback, safety-flagged conversations, and explicitly reported material remain exceptions.
Opt out of training through your account settings if you do not want your Inputs and Outputs used for model improvement; the stated exceptions remain.
If you rate a reply, the terms say its related conversation is stored as Feedback, which Anthropic may use however it chooses without payment.
This applies if you provide Feedback, including by rating an output.
The terms say you retain rights in your inputs and Anthropic assigns you any rights it has in outputs, subject to your compliance with the terms.
The assignment of rights in outputs is subject to your compliance with the terms.
The terms also allow suspension or termination without notice if Anthropic believes you breached them or must act to comply with law.
The right to end a subscription for any other reason applies if you have a subscription; the terms provide a pro-rata refund for the remaining period only for website purchases ended for that other reason.
When your access, subscription, or the terms end, Anthropic may choose to delete materials or other data associated with your account.
This applies if the terms, a subscription, or your access to the services end.
The terms direct disputes exclusively to state or federal courts in San Francisco, California, and specify California law.
The terms seek to limit total liability to the greater of $100 or what you paid for access in the six months before the claim arose.
If you buy a paid subscription, it renews and charges again unless you cancel at least 24 hours before the last day of the current term.
This applies only if you sign up for a paid subscription; app subscriptions follow the app distributor's cancellation terms.
If you cancel a paid subscription, prepaid fees generally are not refunded, though your access continues through the paid term and app refund rules may differ.
This applies if you have a paid subscription; purchases through an app distributor are subject to that distributor's refund terms.
Originally analyzed . The original date marks this service’s first saved analysis, including earlier report formats. This report was generated .
This report was generated from saved policy text and checked in a second automated review. Automated interpretation can still miss or misread terms. The source documents were retrieved on Oct 19, 2026. Policies may have changed since retrieval.
These supporting documents didn’t fit within this report’s reading limit. Their contents are not included in the findings.