Does Claude Team Train on Your Data? Plan by Plan

No. Anthropic’s Commercial Terms, which cover Claude Team, say “Anthropic may not train models on Customer Content from Services.” There are two ways in anyway: a thumbs up or down on any chat, which hands Anthropic “the entire related conversation” for up to 5 years, and any training your company opts into. Training is also a separate question from keeping. Team chats stay until someone deletes them, Team can’t get zero data retention, and the personal plans (Free, Pro, Max, Ultra) train on your chats whenever the training setting is on.

My company runs on Claude Team. Everything about the plans themselves comes from Anthropic’s pages, read on November 6, 2026, not from our admin screens. I set up Claude Team for small companies, so I have a stake. None of this is legal advice. For the wider question of whether Claude is safe for company work, start with the hub post.

Which Claude plans train on your data?

Here is every plan family on one screen. Each row label links to the page the row comes from, and the sections below add the sources for the details.

Read the Team column twice. The training row is the one everyone asks about, and it’s the easy one. The rows that cause trouble are the thumbs row and the “Chats kept” row, because on Team nothing deletes a chat unless a person does.

The commercial training page covers “commercial products such as Claude for Work and the Anthropic API” and says “By default, we will not use your inputs or outputs from our commercial products” to train. The Trust Center FAQ gives the same answer to the question “Will you be using our Claude for Work conversations to train your generative models?”: “No.”

A few pages that rank for this question say Team keeps chats for 30 days. That number belongs to the API and to Claude Code, not to chats in the Team app. Keep them apart on any form you fill in.

Does Claude Pro train on my data?

It does if the setting is on. Claude’s pricing page lists model training as “Opt-out” on Free, Pro, Max and Ultra. Anthropic’s August 2025 announcement says a new user can “select your preference in the signup process,” and it doesn’t say which answer comes pre-selected. Existing users got the same choice in a pop-up. Whatever each person picked stays until they change it, so open your own settings and look rather than trusting anyone’s summary, mine included. The Claude Code docs point to the page: claude.ai/settings/data-privacy-controls.

With training on, Anthropic may keep your data “in a de-identified format for up to 5 years in our model training pipelines.” That clock applies “only” to “new or resumed chats after the model training setting has been enabled.” Turning it off means Anthropic won’t use “your previous or new chats or coding sessions for future model training,” with one catch: “Your data will still be included in model training runs that are already in progress, or in models that have been trained.” You can’t take back a run that already happened.

Two more lines from the personal training article: “Your Incognito chats are not used to improve Claude,” and chats “flagged for safety review” sit outside the training setting. Anthropic may use them “to improve our ability to detect and enforce our Usage Policy,” including “training models for use by our Safeguards team.”

For a company, the point is narrower. A personal seat doing company work sits outside everything in the Team column: no Commercial Terms, no DPA, no owner switch. If two of your people do client work on their own Pro accounts, your Team answer doesn’t cover them. That was us until we moved: before Team, our seats were personal Pro and Max subscriptions.

Our own answer to buyers covers both cases: “If you’re on Claude Pro, we turn off the model-training setting together at the setup check; our deployments run on Claude Team.” If someone tells you “we’re on Team, so it doesn’t train,” ask which plan each person doing the work is actually logged into.

Personal plan or Team plan covers the rest of that decision.

What counts as opting in on Claude Team?

Most pages stop at “no training by default.” The defaults have doors, and here they are.

The thumbs up and down. Anthropic’s commercial page says that if you “explicitly report feedback or bugs to us (e.g. via our thumbs up/down feedback button),” it “may use your chats and coding sessions to train our models,” and “we will store the entire related conversation” for up to 5 years (source). One click from any member sends a whole client chat. Primary Owners and Owners can turn the button off for everyone with the Rate chats setting, under Organization settings > Data and Privacy. If you change one thing this week, change that. The Team setup post covers the other privacy settings on that screen. If you have an AI policy, add a line telling staff not to rate client chats.

Claude Code’s feedback commands. On the data usage page, “Transcripts shared via /feedback, or via /bug and /share” are “retained for 5 years.” IT can switch the command off by setting the environment variable DISABLE_FEEDBACK_COMMAND to 1, which the docs say can be checked into settings.json.

The Claude Code session survey. The rating alone sends no transcript. If someone answers “Yes” to the follow-up, the transcript goes up, “Shared transcripts are retained for up to 6 months,” and Anthropic says they “cannot be used to train our AI models” (same page).

The Development Partner Program. This one gets misread. The program’s own page says “only your Claude Code input and output tokens from the first-party Claude API are provided to us,” it doesn’t reach the Claude apps, and an admin joins through the Console or an account executive. The Claude Code docs agree that it’s “available only for Anthropic first-party API.” A broader privacy article says commercial chats aren’t used “unless you choose to participate in our Development Partner Program,” which reads as if it could touch Team chats. I’m going with the program’s own page, which is more specific. Nobody on a Team plan joins it by accident.

How long does Claude Team keep your data?

Until someone deletes it. Anthropic’s retention page for organizations says “we retain your chats and coding sessions in the product to provide you with a consistent product experience.” Here is the rest of that page:

• Deleted chats are “Removed from your chat history immediately” and “Deleted from our back-end storage systems within 30 days.”
• Incognito chats are “automatically deleted within 30 days unless flagged as a Usage Policy violation.”
• Flagged chats: “We retain inputs and outputs for up to 2 years and trust and safety classification scores for up to 7 years.”
• Feedback (the thumbs above): “we retain data associated with that submission for 5 years.”
• The law: “we may retain your chats or sessions as required by law.”
• Research: “If permitted in your contract with us, we may anonymize your organization’s data for research or statistical purposes,” and keep it longer.

Claude Code on Team runs on a different clock. Its docs give commercial plans a “Standard: 30-day retention period” on Anthropic’s side. Your own machine keeps transcripts “in plaintext under ~/.claude/projects/ for 30 days by default,” and sessions started or last continued in Claude Desktop or Cowork are “exempt from that limit by default.”

Team has no custom retention controls, so a client chat lives as long as nobody remembers to delete it. Chats stay until deleted, and until then one person can export them. Other members only see a chat if someone shares it. Before you promise a client their data is gone, work through the four places deletion misses.

Can Claude Team get zero data retention?

No. Anthropic’s ZDR page says “the only products to which zero data retention applies are eligible Anthropic APIs.” It lists three things: eligible APIs, products that use a commercial organization’s API key (Claude Code through the API among them), and Claude Code for Enterprise plans. It’s “subject to Anthropic’s approval,” and the page sends you to Sales. The Claude Code docs add that “ZDR is not included in the standard Enterprise plan; it is enabled on a per-organization basis by your account team after confirming eligibility.”

Since June 9, 2026, ZDR has a second catch. Covered Models are “Mythos-class models and future models with similar capabilities that we designate as covered models,” and their prompts and outputs are “retained for 30 days.” The change “only applies to organizations that have set up workspaces with zero data retention (ZDR) in Claude Console,” so even a ZDR account keeps 30 days of anything it sends those models.

If a client contract demands zero data retention, the work goes through the API in software you build, or onto Claude Code on an approved Enterprise plan. Team can’t sign that clause. The Claude Code post covers the Enterprise route.

How long does Claude Enterprise keep data?

As long as you set, with a floor. Custom retention “is available to Enterprise plan customers,” set by a Primary Owner or an Owner. “The minimum retention period is 30 days, and each month is counted as 30 days.” It doesn’t reach standalone Claude Design, Claude Tag, Claude Managed Agents, or features that run Claude Code in the cloud, and chats the safety systems flag are still kept under the flagged rule above.

Enterprise chat doesn’t get zero data retention either. The Covered Models page says Enterprise chat and Cowork “already operate with standard retention”, and the ZDR page lists Claude Code as the only Enterprise product it covers. If retention is the reason you’re looking at Enterprise, 30 days is the shortest chat setting Anthropic’s pages describe. When Team isn’t enough lists the other reasons to move up.

Does Claude Team have a DPA, and is it GDPR compliant?

Yes, there’s a DPA, and you don’t sign anything. The Commercial Terms say the Data Processing Addendum “is incorporated into these Terms by reference.” The DPA itself says “Customer is the controller and Anthropic is Customer’s processor” for your personal data. It covers transfers with the Standard Contractual Clauses, “Module Two (controller to processor) or Module Three (processor to processor),” plus a UK Addendum and a Swiss Addendum. New subprocessors come with notice, and you have 15 days to object in writing before it counts as consent. Within 30 days of the contract ending, Anthropic returns your data on request and deletes it, with exceptions.

Anthropic confirms that the consumer terms and privacy policy “do not apply to the use of the Claude for Work plan where Anthropic acts as the data processor.” Personal plans have no DPA in their terms, which the client-data post’s table also shows.

Whether you’re GDPR compliant is your job, and the DPA only covers Anthropic’s part of it. It gives you the processor contract Article 28 asks for. Your records of processing, your privacy notice and any impact assessment are still yours to write.

Data residency is where I’d slow down. Anthropic’s regional compliance page offers processing within the EU/EEA, but its table covers only Amazon Bedrock, Google Vertex AI and Microsoft Foundry. It says nothing about Team or Enterprise on claude.ai. More below.

Is Claude Team HIPAA compliant?

No. Anthropic’s HIPAA article says “Team plans and individual plans (Free, Pro, Max, and Ultra) can’t enable HIPAA,” and the Trust Center FAQ says “Our BAA does not cover beta products or Free, Pro, Max, or Team plans.” The routes that do exist are in the SOC 2 post.

What should I write on the security form?

Three answers the SOC 2 post’s questionnaire table doesn’t have. Use that table for training, audits, subprocessors and the rest, and delete anything here that isn’t true for you.

What I couldn’t confirm

• Where Team data is stored. No Anthropic page I found names a region for Team. A couple of vendor pages say the US, with no source. Don’t promise a client EU storage on Team until Anthropic puts it in writing.
• What a new Pro account starts with. The pricing page says “Opt-out” and the 2025 announcement names no default. Check the toggle on every personal account that touches company work.
• The Development Partner Program wording. Anthropic’s pages disagree, as above. The program’s own page is narrower and I’d rely on it, but if a form asks about training exceptions, name the thumbs feedback and the program both.

What should I do this week?

1. Turn off Rate chats as an Owner, under Organization settings > Data and Privacy.
2. Find anyone doing company work on a personal seat. Move them onto Team (a Standard seat is $3 a month more than Pro, billed annually), or at least have them turn training off today.
3. Write a deletion habit for client work, since Team has no retention timer. A client project gets deleted the week the job closes, and someone’s name goes next to that rule.
4. Read client contracts for zero data retention or EU residency before you sign. On Team, the honest answer to both is no.

If you’d rather plan which of your tools to connect, and how, the setup planner asks six questions and gives you a plan without asking for an email. If a client’s security form is sitting in your inbox, bring it to a 30-minute call.