Tracker / Chat assistant

Does Claude train on your data? Claude admin guide

Since the consumer-terms change (Aug 28, 2025), every Free, Pro and Max user has to choose whether their chats are used for training. Choosing yes extends retention to 5 years. Team and Enterprise (commercial terms) are not used for training by default. SCIM, audit logs and Tenant Restrictions are Enterprise-only.

Vendor: Anthropic · Last checked Oct 22, 2026 · 3 plans tracked

Plan by plan

Free / Pro / Max (personal) consumer

Trains on your data?
User's choice Users choose at signup or in Privacy Settings whether new or resumed chats, including Claude Code sessions, can be used for training. Incognito chats are never used. source ↗ checked Oct 22, 2026
Retention
30 days after deletion, or up to 5 years if training is allowed Deleted chats are removed from back-end storage within 30 days. If training is on, data may be kept de-identified for up to 5 years. Chats flagged for policy violations: inputs/outputs up to 2 years, classifier scores up to 7 years. source ↗ checked Oct 22, 2026
SSO
No Plan comparison shows “No” for SSO on Free, Pro, Max 5x and Max 20x. source ↗ checked Oct 22, 2026
SCIM
No Plan comparison shows “No” for SCIM on all personal plans. source ↗ checked Oct 22, 2026
Audit logs
No Plan comparison shows “No” for audit logs on all personal plans. source ↗ checked Oct 22, 2026
Data residency
Unknown Not verified from a vendor-owned source yet.
DPA
No Consumer plans fall under the Consumer Terms; the DPA is incorporated in the Commercial Terms, which explicitly exclude Free/Pro/Max. source ↗ checked Oct 22, 2026

Team business

Trains on your data?
No Commercial products are not used for training by default. Exception: chats users explicitly send as feedback, which are kept up to 5 years. Owners can turn off the “Rate chats” feedback setting. source ↗ checked Oct 22, 2026
Retention
Until deleted; purged from back end within 30 days Chats are kept so you can continue them; deleted chats are removed from history immediately and from back-end storage within 30 days. Incognito chats are deleted within 30 days. Flagged chats: inputs/outputs up to 2 years. Feedback: 5 years. source ↗ checked Oct 22, 2026
SSO
Yes SSO is available on Team, Enterprise and Console organizations. source ↗ checked Oct 22, 2026
SCIM
No SCIM is not available on Team plans. JIT provisioning is. source ↗ checked Oct 22, 2026
Audit logs
No Audit logs are available for Enterprise organizations only. source ↗ checked Oct 22, 2026
Data residency
Unknown Not verified from a vendor-owned source yet.
DPA
Yes Claude for Work is under the Commercial Terms, which incorporate the Anthropic DPA by reference. source ↗ checked Oct 22, 2026

Enterprise enterprise

Trains on your data?
No Not used for training by default. source ↗ checked Oct 22, 2026
Retention
Custom retention controls Enterprise includes custom data retention controls. source ↗ checked Oct 22, 2026
SSO
Yes source ↗ checked Oct 22, 2026
SCIM
Yes SCIM is available for Enterprise and Console organizations. source ↗ checked Oct 22, 2026
Audit logs
Yes CSV export covering the last 180 days, plus the Compliance API. source ↗ checked Oct 22, 2026
Data residency
US-only inference option Enterprise can keep inference inside the United States. source ↗ checked Oct 22, 2026
DPA
Yes Claude for Work is under the Commercial Terms, which incorporate the Anthropic DPA by reference. source ↗ checked Oct 22, 2026

Includes customer-managed encryption keys and HIPAA-ready configuration (BAA) for eligible organizations.

How to block or allow Claude

  1. network

    Allow only your org (Tenant Restrictions)

    Have a TLS-inspecting proxy overwrite (not append) anthropic-allowed-org-ids: <org-uuid>[,<org-uuid>] on claude.ai, api.anthropic.com, claude.com and anthropic.com. Personal accounts get a 403 tenant_restriction_violation. Covers web, desktop, API keys and OAuth tokens. Requires Enterprise or a Console organization.

    source ↗ checked Oct 22, 2026

  2. admin-console

    Disable feedback submission

    Organization settings → Data and Privacy → “Rate chats”. This stops users sending feedback, which can be used for training.

    source ↗ checked Oct 22, 2026

Domains

“Primary domain only” means we haven't found a full endpoint list from the vendor, so blocking just this domain may not catch every API or desktop-app endpoint.

Change history

Sources