Is Claude confidential? What Anthropic publishes, plan by plan

There is no single answer, because Anthropic runs two sets of rules: one for Free, Pro and Max, another for Team, Enterprise and the API. Below: what each plan says about training and retention, the setting that changes it, who else can see a chat, and what people at work commonly keep out. Every Anthropic statement is quoted from its own pages, read on 17 October 2026.

Short answer: it depends on which Claude you are using, one privacy setting, and what you do with a chat afterwards. This guide won't call Claude, or any other tool, confidential or not. It sets out what Anthropic publishes on training, retention and who can open a chat, so you can decide what to paste in under your firm's policy.

Everything below comes from Anthropic's Privacy Center, Help Center and pricing page, read on 17 October 2026, with the article dates Anthropic shows; sources are listed at the end. More on using Claude at work is in our guides.

Start with the plan: two sets of terms

Nearly every Anthropic privacy article opens by saying which side it covers. Consumer means Claude Free, Pro and Max. Commercial means what Anthropic calls Claude for Work (the Team and Enterprise plans) and the Anthropic API. The same chat window can sit on either side; what decides it is the account you are signed in to. As of 17 October 2026:

The commercial training article names the one way the Team and Enterprise default changes: "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." Team and Enterprise owners can switch off that feedback button for everyone with the Rate chats setting, under Organization settings > Data and Privacy.

Free, Pro and Max: the one setting, and how it came about

On 28 August 2025 Anthropic announced updated Consumer Terms and a new choice for Free, Pro and Max users: allow chats to be used "to improve Claude," or not. New users pick during signup; existing users were given until 8 October 2025 to decide. The same post announced five-year retention if you allow training and said: "If you do not choose to provide your data for model training, you'll continue with our existing 30-day data retention period." Today's retention article (dated 1 July 2026) still gives the five-year term, and says a deleted chat leaves Anthropic's back-end within 30 days. The post also said the change does not apply to "services under our Commercial Terms, including Claude for Work, Claude for Government, Claude for Education, or API use."

Today's Privacy Center article (dated 16 March 2026) lists when Anthropic will use consumer chats, including to improve its models:

• You chose to allow it.
• "Your conversations are flagged for safety review," in which case Anthropic may use them to improve how it detects and enforces its Usage Policy.
• You opted in some other way, "e.g. joining our Trusted Tester Program."

What counts as training data is broad: "the entire related conversation, along with any content, custom styles or conversation preferences." Content fetched by connectors such as Google Drive is excluded, "though data may be included if it's directly copied into your conversation with Claude." The exclusion is for the raw connector data only: Claude's replies about what a connector fetched are part of the conversation, and so is anything you paste in.

Switching the setting off is not retroactive in every sense. Anthropic: "Your data will still be included in model training that has already started and in models that have already been trained, but we will stop using your previously stored chats and coding sessions in future model training runs." Deleting a chat also keeps it out of future training.

Two smaller rules sit next to the main one. A thumbs-up or thumbs-down stores "the entire related conversation" for up to 5 years, de-linked from your user ID, and Anthropic may use it to train. And chats flagged by its automated safety systems are kept longer: "inputs and outputs for up to 2 years and trust and safety classification scores for up to 7 years."

Checking your own settings

The steps are Anthropic's, from its Privacy Center article "How do I change my model improvement privacy settings?" (dated 3 August 2026) and its Help Center article on sharing chats:

1. Open Privacy. On the web or desktop app, select your name in the settings menu, open the drop-down, choose Settings, then Privacy. On the iPhone or Android app it's the same path without the drop-down.
2. Read the training toggle. Under Help Improve our AI models, the toggle shows your current choice. Some Anthropic pages call the same control "Model Improvement."
3. Review what you have shared. On Free, Pro or Max, in the same Privacy section, click Manage next to Shared chats. The list shows every shared chat with its date; you can switch sharing off per chat.
4. For a one-off question, start an incognito chat. On a new chat outside a project, click the ghost icon. Anthropic says incognito chats "are not used for training" and are "retained for either 30 days (default), or longer" where an Enterprise organization sets its own retention. They are available on every plan, though not inside projects.

Who else can see a Claude chat

Training is one question. Access is another, and on Anthropic's pages it has four parts.

People you share it with. Sharing creates a snapshot of the chat at that moment. Anthropic's Help Center (two articles dated 4 September 2026) describes two kinds. Inviting people by email "is the default way to share a chat and keeps it private to the people you choose"; the invite opens only for the address you entered. A public link, on Free, Pro and Max, lets "anyone who has the link" view the snapshot, no account needed. Raw data from connectors is left out of both. Attached files are left out of a public link, and out of an email invite unless the sharing happens within your organization. The conversation and Claude's answers are in. On Team and Enterprise, sharing stays inside the organization, and admins can "see and revoke shared chats."

Public links have made the news. BankInfoSecurity reported on 28 July 2026 that shared Claude conversations and Artifacts had turned up in Google search results, including what appeared to be a patient's medical report and internal company files, and that a similar episode had happened in September 2025. Anthropic's Help Center article "Public links for shared chats" (dated 4 September 2026) says every shared chat page carries a "noindex" instruction and that Anthropic publishes no directory of shared chats, then lists what it can't control: where the link gets posted, copies, and crawlers that ignore noindex. Its own rule: "treat a public link as public. If you wouldn't post the contents on the open web, don't put them behind a public link."

Your organization, on Team and Enterprise. Anthropic's article for work accounts says your organization's Primary Owner "manages your Work account and all associated data," including data exports that "may contain your conversations with Claude, uploaded files, and usage patterns." Incognito chats are included in those exports, and on Enterprise in the Compliance API.

Anthropic. If you allow training on a consumer plan, Anthropic says it de-links your data from your user ID "before any review" and that "Access is limited to a small number of personnel involved in model training." It also says it does not use that data to build profiles or "to sell the information itself to any third party."

Apps you connect, and Cowork. As of 17 October 2026, Anthropic says Cowork, where Claude works through tasks using your files, browser and apps, runs its sessions in the cloud on its servers, and that "the work Claude does there, including any local files it opens through the desktop app, is processed on Anthropic's servers rather than staying on your computer." Its own Cowork safety article advises: "Avoid granting access to local files with sensitive information, like financial documents."

What people at work commonly keep out

Anthropic's own Help Center article for people who want to put sensitive data into Claude (dated 22 May 2026) recommends "being thoughtful about sharing highly sensitive personal details such as" financial information, health records, passwords and "Confidential business or personal documents." Beyond that list, these are choices professionals commonly make rather than rules, and your firm may already have made them for you:

• Names come out, the question stays. "A client in manufacturing wants to renegotiate a fixed-fee engagement" gets most of the value of the real paragraph without naming anyone.
• Material under someone else's terms stays out. Documents received under an NDA, a data-room agreement or an engagement letter that limits third-party services: those terms decide.
• The plan follows the work. Client work on a firm-provided Team or Enterprise account; personal experiments on a personal one. Mixing them means mixing two sets of terms.
• Connections stay narrow. One folder for Cowork, not a whole drive.

What this page doesn't cover

No legal or compliance conclusion: nothing here says whether a use of Claude meets a professional rule, privilege, HIPAA, GDPR or a client contract. Nor Claude Code, zero data retention in detail, or Claude through Amazon Bedrock or Vertex AI. For certifications and security reports, Anthropic points to its Trust Center at trust.anthropic.com (Privacy Center article 10015870, commercial products, dated 16 March 2026). This is what Anthropic publishes, not an audit of how it operates.

Where Equerry fits: what it reads

Equerry is a personal assistant who helps you get things done. It reads only what you send it: an email thread you CC or forward to your own Equerry address; a PDF, an image, a link or text shared from your iPhone's share sheet; a note you type; a voice note you record in Equerry. If you allow it, it also reads your calendar, read-only. It never asks for your inbox, and the original stays where it was. What your firm weighs is each item you decided to send, not access to an account, and that decision is still yours under your policy.

Here is one exchange:

A thread you CC keeps itself current: replies that keep the Equerry address on the thread keep arriving, so Thursday's reply from the client is there before Friday's call. Anyone on a CC'd thread can see the Equerry address.

What you send is processed to make it searchable and answerable, including by the outside providers named on our AI disclosure page, which says what each one receives and states that we never train AI models on what you save. No ads. Equerry sends nothing, books nothing and buys nothing on your behalf. It makes no claim about any professional rule, or about whether what you send it is confidential: whether client material goes into Equerry, Claude or anything else is your decision under your firm's policy and your client's terms.

If you're weighing the same question for the other assistant on your phone, read is ChatGPT confidential? For what Claude can reach once you connect a mailbox, see can Claude read my emails?

Sources, all read 2026-10-17. Anthropic: Claude pricing (plan prices and the "Model training" row; prices exclude applicable tax and "are subject to change at Anthropic's discretion"); Updates to Consumer Terms and Privacy Policy (28 August 2025). Anthropic Privacy Center: Is my data used for model training? (consumer, dated 16 March 2026); Is my data used for model training? (commercial, dated 18 August 2026); How long do you store my data? (dated 1 July 2026); How long do you store my organization's data? (dated 1 July 2026); How do I change my model improvement privacy settings? (dated 3 August 2026); What Certifications has Anthropic obtained? (commercial, dated 16 March 2026). Claude Help Center: I would like to input sensitive data into my chats with Claude. Who can view my conversations? (dated 22 May 2026); Who owns and manages the data of my team? (dated 16 March 2026); Use incognito chats (updated 30 September 2026); Share and unshare chats (dated 15 June 2026); Share a chat with specific people and Public links for shared chats (both dated 4 September 2026); Use Claude Cowork safely (updated 14 October 2026). Press: BankInfoSecurity, Claude's Shared Chats Surface on Google Search (28 July 2026). Anthropic's pages describe consumer plans (Free, Pro, Max) or commercial plans (Team, Enterprise, API) as each article states; where Anthropic's current page disagrees with this one, Anthropic's page is right. This guide is general information, not legal advice.

Last updated 2026-10-17.

All product names, logos, and brands are property of their respective owners. Equerry is not affiliated with, endorsed by, or sponsored by any other product mentioned; comparisons are provided for informational purposes.