Anthropic publishes more about retention than most, and it is still a promise about what a provider chooses to do with plaintext it can read. Amnesia AI® removes the choice from the provider, ours included.
The right-hand column is the category verdict from our full comparison, not a claim about Claude specifically. Brand-specific statements are below, each one linking to Anthropic's own page. No dollar figure is quoted for any product but ours.
Their words, their page, with the date we read it.
Consumer accounts that allow their data to be used for model training move to a five-year retention window; accounts that do not keep the 30-day window.
Anthropic, "Updates to Consumer Terms and Privacy Policy" (28 Aug 2025) · read 20 Aug 2026 ↗Conversations flagged by safety systems can be reviewed and retained under a separate policy from the ordinary retention window.
Anthropic privacy policy · read 20 Aug 2026 ↗Claude runs frontier models, and Anthropic's published retention terms are among the clearest in the category. The difference is who is trusted to honor them.
Product names belong to their respective owners and are used here only to identify the products compared. No affiliation or endorsement is implied. Every statement about another product links to that vendor's own published page, with the date we last read it; vendors change their terms, so check the source before relying on it.
Nothing to deploy and no contract to sign. Attach the document you could not paste anywhere else and read the report on the reply.
Experience confidential AI