What Claude Sees, What It Keeps, and What Stays Private
Before anyone will use Claude for real work, someone asks "what happens to our data?" Plain-English answer, share-able with IT, legal, or a skeptical colleague.
The Short Version
- Claude only sees what you put in front of it. It cannot read your email, access your files, or browse your systems on its own.
- Your conversations are stored in your account until you delete them.
- On Team and Enterprise workspaces, your data is isolated from other companies.
- Anthropic does not train on Team or Enterprise workspace content by default. This is a specific, meaningful commitment for paid team accounts.
What Claude Can See
Exactly what you share in a conversation — nothing more, nothing less.
That includes:
- Text you type or dictate
- Files you attach (PDFs, Word, images, spreadsheets)
- Content you paste into chat
- Photos you share
That's the whole list. Claude has no background access. It cannot see your inbox, your calendar, your phone contacts, your internal file servers, your accounting software, or your project management system. If information isn't in the conversation window, Claude doesn't know it exists.
The practical test. "Did I put this in the chat?" If yes, Claude can use it. If no, Claude can't see it.
What Claude Does NOT See
- Your email inbox (unless you paste from it, or use a Connector — Lesson 17)
- Files on your computer or server
- Other people's conversations in your workspace
- Anything outside the current conversation
- Empower's internal systems or databases
- Your browsing history or phone activity
How Long Conversations Are Stored
Conversations are stored in your Claude account and accessible any time, like a history log.
Delete individual conversations whenever you want. Once deleted, they're gone from your account.
Workspace conversations are not routinely accessed by Anthropic employees — they're stored in Anthropic's infrastructure with standard cloud-service access controls.
The Training Data Question
This is the one that matters most for business use.
Free accounts: Conversations may be reviewed for safety and may be used to improve future models. Standard for free AI tools.
Pro and Max: By default, content is not used to train models. Verify current terms at claude.com/privacy before adopting.
Team and Enterprise: Anthropic's commitment is explicit — workspace content is not used to train AI models by default. This is the main reason Empower should operate on a Team account rather than letting staff use personal Free or Pro accounts.
If you're a decision-maker evaluating Claude: the training-data policy is the reason the Team plan exists at its price point. It's not just seat management — it's protecting business conversations from feeding a public model.
Your Workspace Is Isolated
On Team or Enterprise, Empower's data lives in a separate environment from other companies. Another company's employees cannot see your conversations, and you cannot see theirs. Same basic architecture as any enterprise SaaS.
What You Should Still Avoid Sharing
Even with protections in place, some categories carry extra risk regardless of tool. Check Empower policy before pasting:
- Social Security numbers or other government IDs
- Passwords or access credentials
- Documents under active litigation or attorney-client privilege
- Patient health records (HIPAA-covered information, unless you're on a HIPAA-ready Enterprise arrangement)
- Financial account numbers
These aren't Claude-specific — they're best-practice limits for any cloud tool. The right question: "Would I forward this to a colleague in email?" If no, don't put it in Claude either.
For IT and Legal Reviewers
Point them here:
Covers:
- SOC 2 Type II certification
- Data Processing Agreements (available for Enterprise)
- Security policies and incident response
- Privacy policy and terms
Enterprise accounts can request a formal DPA and additional compliance documentation. If Empower needs HIPAA-ready operation or a signed DPA, move from Team to Enterprise.
Real Example: A Legal Team's Concern
A general counsel was asked to approve Claude for the operations team. Concern: would client project details end up in Anthropic's training data and potentially surface for competitors?
After reviewing the Team plan documentation and trust.anthropic.com, assessment: Team workspace content is excluded from model training by default, client details stay within the workspace, risk profile comparable to other cloud software like CRM or PM tools. Approved with one policy addition: no attorney-client privileged documents until Empower evaluated Enterprise with a formal DPA.
Reasonable, calibrated response — the kind of analysis legal can make once they have the right information.
Real Example: What an Employee Shared (and Why It Was Fine)
A coordinator started using Claude to draft vendor RFIs. She pasted scope, described the issue, got drafts. Documents included project names, addresses, and dollar amounts.
None of this is privileged or trade-secret. It's the kind of content that lives in routine business correspondence. Using Claude to draft it is no more sensitive than using a word processor — the content was going into an email anyway. Company policy permitted it. She got about two hours of her week back.
Summary
| Question | Answer |
|---|---|
| Can Claude read my email? | No — only what you paste or share |
| Can Claude see my files? | No — only what you attach |
| Are my conversations saved? | Yes, in your account; deletable anytime |
| Is my workspace shared with other companies? | No — isolated per workspace |
| Does Anthropic train on Team conversations? | Not by default — explicit policy |
| Where can IT review security docs? | https://trust.anthropic.com |
What Comes Next
With the privacy picture clear, build prompts that actually produce results. Continue to Lesson 15.