Claude resources · Guide

Which Claude Plan Is Safe for Client Data?

The contract matters more than the features. What Anthropic's published terms say about training, retention, and confidentiality on each plan.

Checked September 17, 2026

Claude comes in consumer plans and commercial plans, and the two are governed by different contracts. That difference matters more to a law firm than any feature. This page summarizes what Anthropic's own published terms and help articles said on the date above. Terms change. Read the current versions before you decide, and have the attorney responsible for your firm's ethics compliance make the call.

Short version: Free, Pro, and Max are consumer products. Team, Enterprise, and the API are commercial products where Anthropic acts as a processor of your data, does not train on your content by default, and takes on written confidentiality obligations. For client information, start your evaluation with the commercial plans.

Claude plans compared, from Anthropic’s published terms, checked September 17, 2026
Free, Pro, MaxTeamEnterpriseAPI
Contract that governsConsumer Terms of Service and the consumer Privacy PolicyYour organization's commercial agreement with Anthropic. The consumer terms do not apply.Your organization's commercial agreement with Anthropic. The consumer terms do not apply.Commercial Terms of Service, which incorporate a Data Processing Addendum
Anthropic's roleNot stated as processor. No Data Processing Addendum.Processor. Your firm is the controller.Processor. Your firm is the controller.Processor under the Data Processing Addendum
Used to train modelsMay be, depending on each user's "Help improve our AI models" privacy setting. Each user controls it. The firm does not.Not by default. Exceptions: feedback you submit and programs you explicitly opt into.Not by default. Exceptions: feedback you submit and programs you explicitly opt into.Not by default. The Commercial Terms say Anthropic may not train on Customer Content.
How long data is keptWith training allowed: up to five years in training pipelines. With training off: 30-day back-end retention. Deleted chats are removed from the back end within 30 days.Chats stay until a user deletes them, then are removed from the back end within 30 days.Same as Team, plus owner-set custom retention periods with a 30-day minimum.Anthropic's privacy center says inputs and outputs are deleted within 30 days, with exceptions such as stored files. Zero data retention is available by agreement for eligible uses.
Safety exceptionsContent flagged for a usage policy violation can be kept up to two years. Feedback you submit is kept and can be used.Same exceptions apply.Same exceptions apply.Same exceptions apply, including under zero data retention.
Written confidentiality dutyNo equivalent confidentiality clause in the Consumer Terms.Commercial agreementCommercial agreementYes. Customer Content is your Confidential Information, used only to perform the agreement.
Who owns the outputYou. The Consumer Terms assign outputs to the user.Your firmYour firmYour firm. Anthropic assigns its rights in outputs to the customer.
Single sign-onNoYesYesYes, for Console organizations
Audit logsNoNoYes, 180 days of history. Chat content is not in the log.Not covered here
SCIM, custom retention, compliance API, IP allowlistingNoNoYesNot covered here
HIPAA business associate agreementNot availableNot listed as availableAvailable. An owner must activate the HIPAA-ready configuration first.Available for HIPAA-enabled organizations, with feature limits
SeatsOne person2 to 150Minimum 20 self-serve, 50 through salesUsage based
What it means for a firm
01

Consumer plans put the decision in each user's hands

On Free, Pro, and Max, whether conversations can be used for training is a personal setting each user picks, and the firm has no way to see or enforce it. There is no administrator, no audit trail, and no commercial confidentiality clause. If anyone at your firm is using a personal account for client work today, that is the first thing to fix.

02

Team is the realistic starting point for most small firms

Team moves the firm onto a commercial agreement: no training on your content by default, Anthropic as processor, central billing, single sign-on, and an owner who can manage members. It starts at two seats.

03

Enterprise adds the controls regulators and clients ask about

Audit logs, custom retention periods, SCIM, a compliance API, IP allowlisting, and the HIPAA-ready option with a business associate agreement are Enterprise features. Firms handling protected health information at volume, or answering client security questionnaires, usually end up here.

04

No plan removes your duties

A commercial plan addresses where the data goes. It does not address informed consent, protective orders, client instructions, supervision, or verification. ABA Formal Opinion 512 treats those as the lawyer's job regardless of the tool. Your engagement letters and your written AI policy still have to do that work.

Questions to settle before anyone uploads a file

  • Which accounts are people at the firm actually using today, and who pays for them?
  • Do any client engagement letters, outside counsel guidelines, or protective orders restrict third-party processing or AI tools?
  • Do we handle protected health information, and if so, do we need a business associate agreement in place before any of it is uploaded?
  • Who at the firm will own the admin console, add and remove users, and review settings when the terms change?
  • What retention period do we want for conversations, and does it match our file retention policy?
  • Does our written AI policy name the approved plan and prohibit client information everywhere else?