The Claude Team Data Protection File

The Commercial Terms of Service, plus the documents they incorporate by reference: the Service Specific Terms (whose Section A is titled for the Team and Enterprise plans), the Data Processing Addendum, the Usage Policy, the Supported Regions Policy, and the Model Pricing Page. There is no standalone "Claude Team terms" document.

These Terms (including the Usage Policy, Supported Regions Policy, Service Specific Terms, DPA, Model Pricing Page and other documents or terms that are incorporated by reference by these Terms) constitute the parties’ entire understanding as to the Services’ provision and use.Section M.9 · document effective 17 June 2025 · accessed 22 July 2026

A. Claude for Work (Team Plan; Enterprise Plan)Section A heading · document effective 8 June 2026 · accessed 22 July 2026

Anthropic Ireland, Limited, for a customer in the UK, the EEA or Switzerland. Anthropic, PBC contracts with customers elsewhere.

“Anthropic” means Anthropic Ireland, Limited if Customer resides in the European Economic Area (“EEA”), Switzerland or UK, and Anthropic, PBC if Customer resides anywhere else.introductory paragraph, preceding Section A · document effective 17 June 2025 · accessed 22 July 2026 · re-read against the live page 28 July 2026

Not by default. The Commercial Terms state the prohibition without qualification. Two documented exceptions run alongside it, and both require somebody to switch them on: deliberately submitted feedback, and Development Partner Mode, which the Service Specific Terms make an organisation-level choice the customer has to enable.

Anthropic may not train models on Customer Content from Services.Section B · document effective 17 June 2025 · accessed 22 July 2026 · re-read against the live page 28 July 2026

By default, we will not use your inputs or outputs from our commercial products (e.g. Claude for Work, Anthropic API, Claude Gov, etc.) to train our models. If you explicitly report feedback or bugs to us (e.g. via our thumbs up/down feedback button), or otherwise choose to allow us to use your data, then we may use your chats and coding sessions to train our models.https://privacy.claude.com/en/articles/7996868-is-my-data-used-for-model-training · accessed 22 July 2026

If Customer enables Development Partner Mode, Anthropic may use the data that Customer submits to the Services (e.g., Customer Content) in connection with Anthropic’s products and services, including to train models.Section E · document effective 8 June 2026 · accessed 22 July 2026

In the event of any conflict between the Terms and these Service Specific Terms, these Service Specific Terms control for the relevant Services.preamble · document effective 8 June 2026 · accessed 22 July 2026

It sends the entire related conversation, not the single message, and that conversation may be stored for up to five years and used for training. Anthropic de-links the feedback from user and customer IDs before it is used; the content of the conversation is stored regardless. An Owner can switch the button off for the whole organisation.

When you provide us feedback via our thumbs up / down button, we will store the entire related conversation, including any content, custom styles, conversation preferences, or model settings, in our secured back-end for up to 5 years.https://privacy.claude.com/en/articles/7996868-is-my-data-used-for-model-training · accessed 22 July 2026

We de-link your feedback from your user and customer IDs before it’s used by Anthropic.https://privacy.claude.com/en/articles/7996868-is-my-data-used-for-model-training · accessed 22 July 2026

As a Primary Owner or Owner of a Team or Enterprise plan, you can disable the ability for members of your organization to submit feedback to Anthropic via the thumbs up / down button using the Rate chats setting, under Organization settings > Data and PrivacyDisabling Feedback · accessed 22 July 2026

In practice. Switching Rate chats off stops this from that point on. It does not reach back to a conversation that has already been rated, and Anthropic’s documentation does not describe a self-service way to withdraw an earlier rating.

The DPA fixes the roles: the customer is the controller and Anthropic is the processor. Comparing the DPA against the ICO’s checklist for processor contracts, the required content is present, each at a named section: processing on documented instructions (B.2), notice of an infringing instruction (B.5), assistance with impact assessments (B.6), confidentiality (B.7), sub-processor terms (Section C), assistance with data-subject requests (D.1 and D.2), security measures (E.1 and Schedule 2), audit rights including SOC 2 reports on request (F.1 and F.2), breach notice (G.1), and return or deletion within 30 days of termination (H.1).

With respect to Customer Personal Data, Customer is the controller and Anthropic is Customer’s processor.Section B.1 · document effective 24 February 2025 · accessed 22 July 2026 · re-read against the live page 28 July 2026

Unless required by applicable law to which Anthropic is subject, Anthropic will only process Customer Personal Data to provide or maintain the Services, and in compliance with Customer’s documented instructions...Section B.2 · document effective 24 February 2025 · accessed 22 July 2026 · re-read against the live page 28 July 2026

In practice. This is a comparison of documents, which shows the required terms are present. Two of the provisions are quoted here; the rest are section pinpoints to the same document, each due the same word-for-word re-read on the File’s review dates. Whether a processor performs them in practice is a separate question, answered through the audit artefacts the DPA itself provides for.

Within 48 hours of becoming aware of it. A UK controller’s own clock to the ICO, where a breach is notifiable, is 72 hours, so a 48-hour vendor commitment leaves that timetable workable.

Anthropic will notify Customer in writing without undue delay, but in any event within 48 hours, after becoming aware of any Security Breach, and will assist Customer in complying with Customer’s obligations under Applicable Data Protection laws by reasonably cooperating with Customer’s investigation of the Security Breach.Section G.1 · document effective 24 February 2025 · accessed 22 July 2026 · re-read against the live page 28 July 2026

“Security Breach” means a breach of Anthropic’s security leading to the accidental or unlawful destruction, loss, alteration, unauthorized disclosure of, or unauthorized access to, Customer Personal Data.Section A.7 · document effective 24 February 2025 · accessed 22 July 2026 · re-read against the live page 28 July 2026

In practice. The commitment is tied to personal data. A pure service outage that never touches personal data falls outside this clause.

The DPA incorporates the EU Standard Contractual Clauses together with the UK Addendum in its approved version. That is the safeguard route this contract adopts; UK law also recognises others, including the standalone International Data Transfer Agreement and, for US importers holding a UK Extension certification, the UK-US data bridge, but none of those is what this DPA uses.

The parties agree that, to the extent required by Applicable Data Protection Laws, the terms of the SCCs Module Two (controller to processor) and/or Module Three (processor to processor), as completed as described in Schedule 3 of this DPA, are hereby incorporated by reference and will be deemed to have been executed by the parties.Section I.1 · document effective 24 February 2025 · accessed 22 July 2026 · re-read against the live page 28 July 2026

“Approved Addendum” means the template addendum, version B.1.0 issued by the UK Information Commissioner under S119A(1) Data Protection Act 2018 and laid before the UK Parliament on 2 February 2022Schedule 3.B · document effective 24 February 2025 · accessed 22 July 2026 · re-read against the live page 28 July 2026

By default, we may route customer traffic to select countries in the US, Europe, Asia and Australia, unless otherwise agreed upon or at your instructionsfull article · accessed 22 July 2026

data is stored in the USfull article · accessed 22 July 2026

process data for internal processes (such as safety-related review, product support, or incident response) in countries where we or our affiliates operatefull article · accessed 22 July 2026

In practice. The mechanism is only half of the job; the customer, as exporter, still owes a documented assessment that the mechanism gives adequate protection for its own transfer. The contract does not supply that.

A file the tool never reads stays where it is. A file a session reads into context is transmitted to Anthropic, processed under the DPA, and held on that surface for 30 days. A plaintext transcript of each session is also written on the local machine.

Claude Code runs locally. To interact with the LLM, Claude Code sends data over the network. This data includes all user prompts and model outputs, encrypted in transit via TLS 1.2+.https://code.claude.com/docs/en/data-usage · accessed 22 July 2026

Commercial users (Team, Enterprise, and API): Standard: 30-day retention periodData retention · accessed 8 August 2026

In practice. The paths by which session content can go back to Anthropic for product improvement, the feedback command and the after-session transcript prompt, have documented off-switches. Before fetching a web page the tool sends the hostname, and only the hostname, for a safety check.

Yes, on Anthropic’s published plan pages. The Team plan card lists it and the pricing FAQ states it is included in all paid plans. The Service Specific Terms, which are the contract document dealing with the Team plan, do not mention Claude Code at all.

Claude Code is included in all paid plans. It shares the same usage limits as the rest of your plan, so your work in the terminal and your chats draw from one pool.Is Claude Code included in my plan? · accessed 28 July 2026

Commercial users: (Team and Enterprise plans, API, 3rd-party platforms, and Claude Gov) maintain existing policies: Anthropic does not train generative models using code or prompts sent to Claude Code under commercial terms, unless the customer has chosen to provide their data to us for model improvement […]Data training policy · accessed 22 July 2026

In practice. The Claude Code documentation points Team users to the same commercial contract that governs the rest of the plan, so the data-protection position follows the Commercial Terms and the DPA rather than anything separate.

A public project exposes its contents and knowledge files to everyone in the workspace. Chats inside it stay private unless a user shares one, and sharing stays inside the organisation.

Everyone in your organization can view and use the project.Understand the two project visibility options · accessed 10 August 2026

Even if a project is public, your chats within that project will be private and inaccessible to other members of your organization unless you manually share them.https://support.claude.com/en/articles/9519189-manage-project-visibility-and-sharing · accessed 22 July 2026 · re-read against the live page 28 July 2026

Archiving a project doesn’t reset its sharing permissions or remove members. All members, permission levels, and project knowledge are preserved, and everything is restored exactly as it was when you unarchive the project. To revoke someone’s access, remove them in the project’s sharing settings before or after archiving.What happens when archiving a shared project? · accessed 28 July 2026

Disabling the public project feature allows Team and Enterprise owners to restrict the creation of public projects across their organization while maintaining internal sharing capabilities.introduction · document effective 16 March 2026 · accessed 28 July 2026

All existing public projects will be converted to private projects.When you disable public projects · document effective 16 March 2026 · accessed 28 July 2026

In practice. The documentation does not describe a routine way for an administrator to browse members’ private chats. It does record that an organisation-level export run by the Primary Owner can contain conversations and uploaded files. Removing a member ends that member’s access, and ends other members’ access to that member’s own chats and private projects, but projects they had shared stay available. Removal is not a deletion event.

A snapshot of the messages sent up to that moment, and it stays inside the organisation. Team and Enterprise plans cannot share a chat publicly. Later messages stay private unless somebody updates the snapshot. Uploaded files are excluded from it, and so is the raw data behind tool calls.

Users on Team and Enterprise plans can only share chats with other members of the same organization, not publicly.Share chats · document effective 15 June 2026 · accessed 22 July 2026

If you share a chat that contains an attached file, the file itself is not included in the shared snapshot and remains private.Attached files · document effective 15 June 2026 · accessed 8 August 2026

When sharing chats that use MCP integrations, the raw data retrieved from MCP tool calls remains hidden in the shared snapshot.MCP tool calls · document effective 15 June 2026 · accessed 8 August 2026

The chat snapshot includes all messages that were sent prior to sharing the chat, including any artifacts. All messages sent after sharing a chat will remain private by default. However, you can choose to update the shared snapshot to include new messages.Share and unshare chats · accessed 22 July 2026 · re-read against the live page 28 July 2026

In practice. Unsharing is a visibility change on the same chat, and the documented effect is on the link: the chat moves from Shared back to Private, which disables the direct link.

Anthropic publishes a sub-processor list and commits to it in the DPA. On a new appointment it gives "reasonable notice", with no fixed number of days, and a fifteen-day window to object on reasonable data-protection grounds.

Anthropic’s list of subprocessors is available at https://www.anthropic.com/subprocessors.Schedule 4 · document effective 24 February 2025 · accessed 22 July 2026 · re-read against the live page 28 July 2026

C.3. In the event that Anthropic wishes to appoint an additional Subprocessor: (a) Anthropic will provide Customer reasonable notice of the new Subprocessor prior to giving the Subprocessor access to Customer Personal Data; and (b) Customer may, on the basis of reasonable data privacy or data security concerns, object to Anthropic’s use of such Subprocessor by providing Anthropic with written notice of the objection within fifteen (15) days of the date of such notice, or Customer is deemed to consent to the new Subprocessor. In the event Customer objects to Anthropic’s use of a new Subprocessor, Customer and Anthropic will work together in good faith to find a mutually acceptable resolution to address any objection raised by Customer.Section C.3 · document effective 24 February 2025 · accessed 22 July 2026

In practice. That address redirects to Anthropic’s Trust Center, which carries the current list and a subscription option for change notices. If an objection cannot be resolved, the practical backstop is the customer’s right to terminate on notice under the Commercial Terms.

Nothing in the documents prohibits it, though the Usage Policy treats healthcare decision-making as a high-risk use that requires review by a qualified professional and disclosure that AI was involved. Schedule 1 is the standard description of the processing that the Standard Contractual Clauses require. It records the categories of data subjects and of personal data as matters the customer determines, then records special categories as “None”, which will not match a customer whose ordinary work involves health or other special-category material.

B.1. Categories of data subjects: Determined by Customer (in accordance with the Agreement). B.2. Categories of personal data: Determined by the Customer (in accordance with the Agreement). B.3. Special categories of personal data (if applicable): None.Schedule 1, Part B · document effective 24 February 2025 · accessed 22 July 2026

When using our products or services to provide advice, recommendations, or in subjective decision-making directly affecting individuals or consumers, a qualified professional in that field must review the content or decision prior to dissemination or finalization.High-Risk Use Cases · accessed 10 August 2026

In practice. The protections themselves are unaffected. The substantive obligations are drafted around “Customer Personal Data” as defined, with no special-category carve-out: processing on documented instructions at B.2, confidentiality at B.7, security at E.1 and Schedule 2, breach notice at G.1, and deletion or return at H.1. Those apply to special-category data on the same terms as any other personal data.

No. Privilege is not addressed in the Commercial Terms, the Service Specific Terms or the DPA. What those documents do provide is a confidentiality obligation on every person Anthropic authorises to process customer personal data.

Anthropic will ensure that each person it authorizes to process Customer Personal Data is subject to an appropriate duty of confidentiality.Section B.7 · document effective 24 February 2025 · accessed 22 July 2026

In practice. Anthropic maintains a separate article on how it handles governmental requests for user information (support.claude.com/en/articles/9519291, read 10 August 2026). A firm sending privileged material to any cloud service will want to read that page itself and reach its own view. The contract documents do not speak to it.

Not established from the public documents. A UK customer contracts with Anthropic Ireland, Limited, and the DPA incorporates the ICO’s Approved Addendum for UK transfers. No Article 27 representative is named in any of the legal documents reviewed, and no ICO registration number appears in them.

“Anthropic” means Anthropic Ireland, Limited if Customer resides in the European Economic Area (“EEA”), Switzerland or UK, and Anthropic, PBC if Customer resides anywhere else.introductory paragraph, preceding Section A · document effective 17 June 2025 · accessed 22 July 2026 · re-read against the live page 28 July 2026

With your own dated copy. No Anthropic legal document publishes a version number. Each page shows an effective date and a link to a previous version, which gives two points rather than a numbered history. Privacy centre articles are dated inconsistently: some carry a fixed date, and at least one, as read on 22 July 2026, carried only a relative timestamp such as “Updated over 3 weeks ago”.

Anthropic may update these Terms at any time, to be effective 30 days after the updates are posted by Anthropic or Customer otherwise receives Notice, except that updates made in response to changes to law or regulation take effect immediately upon posting or Notice. Changes will not apply retroactively.Section M.3, Amendment and Modification · document effective 17 June 2025 · accessed 22 July 2026

In practice. The consequence falls on the customer. Point-in-time evidence of the terms you contracted under depends on a copy you took and dated yourself, because the published page shows whatever is current when you look.

Almost none of them are contractual. The Commercial Terms contain no retention or deletion timetable at all. Every row in the retention table in this File cites a page Anthropic can revise without amending the contract: privacy centre articles, the Claude Code documentation, and a support centre article. The single retention commitment written into a contractual document is the deletion-on-termination clause in the DPA, quoted below.

H.1. Within thirty (30) days of the date of termination or expiration of the Agreement, Anthropic will: [...] delete all copies of Customer Data (including Customer Personal Data) processed by Anthropic or any Subprocessors, except to the extent (i) Applicable Data Protection Laws or other applicable legal or regulatory requirements requires storage of the Customer Data, (ii) retention of the Customer Data by Anthropic is necessary to resolve a dispute between the parties, or (iii) retention of the Customer Data is necessary to combat harmful use of the Services.Section H.1 · document effective 24 February 2025 · accessed 22 July 2026

In practice. Anyone writing a retention schedule from these documents should record where each period comes from, because the two kinds of source behave differently. A contractual term changes by agreement. An article changes when the vendor edits the page.

No such control is documented for Team. A custom retention period is documented as an Enterprise feature, and the documented Team controls are chat deletion, the Rate chats toggle, and an organisation export.

This feature is available to Enterprise plan customers.https://privacy.claude.com/en/articles/10440198-configure-custom-data-retention-controls-for-enterprise-plans · accessed 22 July 2026

In practice. The sentence "By default, data is retained indefinitely unless a custom retention period is set" appears in that same Enterprise article and describes the Enterprise default. It should not be quoted as a statement about Team.

Not settled on the text as written. Feedback submissions are held for five years. That five-year period is not among the stated exceptions to the DPA’s deletion-on-termination clause, and the public documents do not reconcile the two.

Where you have provided feedback to us (e.g. by submitting feedback through our thumbs up/down button or sent bug reports), we retain data associated with that submission for 5 years.https://privacy.claude.com/en/articles/7996866-how-long-do-you-store-my-organization-s-data · accessed 22 July 2026

In practice. One more reason to keep the Rate chats switch off where client material is in play.

What is kept, and for how long

How the dates work. Accessed is when a quotation was first captured. Re-read is set only where the quotation was actually checked against the live page again on that date; its absence means it has not been re-read since it was captured. The whole-set verification date at the top is the date the entire File was checked together. It is never advanced to match a single source's later re-read, because that would claim a full sweep that did not happen.

Machine-readable version: https://companyscope.io/api/files/claude-team (JSON, CC BY 4.0).