Skip to content

Who owns, sees, and trains on your chats on a Claude Team account?

Last updated 2026-08-17 · Confidence: documented — Anthropic’s commercial terms, Claude help center, and privacy center.

On a Team plan the organization is the customer: it owns the data, its Primary Owner can export every member’s conversations, and Anthropic contractually doesn’t train on any of it. Teammates see none of your chats.

Anthropic’s Commercial Terms — which govern Claude for Work, i.e. Team and Enterprise — say the customer “retains all rights to its Inputs and owns its Outputs”, and that “Anthropic may not train models on Customer Content from Services”. The customer is the organization, not the individual member.

Consumer plans (Free/Pro/Max) differ: since August 28, 2025 they carry a training toggle, and opted-in chats are retained de-identified for up to 5 years. That update explicitly excludes Team and Enterprise.

  • Teammates: nothing. Even in a shared project, “your chats within that project will be private and inaccessible to other members of your organization unless you manually share them” (help center).
  • Admins: can set organization instructions that silently apply to every conversation — members can’t view their text. Docs recommend generic content (formatting, tone, org context), though nothing enforces that.
  • Primary Owner: the one documented route to members’ conversations — an organization data export containing conversation and user data, emailed as a download link. Members have no self-serve export on Team.

Deleted conversations leave back-end storage within 30 days. Safety-flagged inputs and outputs are kept up to 2 years, their classifier scores up to 7.

A personal account runs under consumer terms: no data-processing agreement with the employer, plus the training toggle above — see GDPR & personal accounts. Conversely, personal material on a team account sits inside the org’s export power; Account migration covers moving it out.

Does Claude at my org learn from everybody else’s chats? No. Each member’s chats are private to them, Anthropic doesn’t train on commercial-plan content, and the only centrally controlled input is the org-instructions text. Claude’s chat-search features reach only your own history.