Does Claude train on your data?

Anthropic · AI assistants

By default it can. On the consumer tier Claude uses your inputs to improve models unless you opt out. Its business tier typically does not.

What we checked

This assessment separates the consumer product from any business plan. It uses the vendor documents linked below; an unverified item is not a pass.

Default data training: Opt-out required; Since the Aug 2025 consumer terms update, Free/Pro/Max chats and coding sessions are used to train models unless the user opts out, with up to 5-year retention if left on.
Business-tier training: No by default; Claude for Work (Team/Enterprise), Claude for Education, Claude Gov, and the API run under Commercial Terms and are not used for model training.
Training control: Confirmed; Anthropic documents a training control. Confirm whether it is enabled centrally or must be set by each user.

Practical risk: Even when a consumer opts out of training, conversations flagged for safety review can still be retained up to two years and used to improve models without notifying the user.

How to make a decision

Check the precise account tier, written contract and intended data before approving Claude. A security certification, DPA or setting can apply to only part of a vendor's service. Keep the source links with your supplier review and revisit them when the vendor changes its terms.

More on Claude

Is Claude HIPAA compliant?Is Claude GDPR compliant?Is Claude SOC 2 compliant?Is Claude ISO 27001 certified?

See the full Claude risk profile, with every data-handling fact and its source, or browse all rated AI tools.