---
title: Is Claude, Copilot or Gemini safe for confidential information?
description: Is Claude, Microsoft Copilot or Gemini safe for confidential information? Each plan's training, retention and human review, from each provider's own terms.
canonical: https://privatesuperintelligence.si/blog/is-claude-copilot-gemini-safe-for-confidential-information/
last-updated: 2026-10-15
---

# Is Claude, Copilot or Gemini safe for confidential information?

Claude, Microsoft Copilot and Gemini can each be a reasonable home for confidential information on their work plans, and each is a weaker one on a personal account. On the business versions (Claude for Work and the Anthropic API, Microsoft Copilot signed in with a work Entra ID account, Gemini in Google Workspace), all three providers say they do not train on your prompts by default. On personal accounts, each provider may use your conversations to improve its models or have people review them, and you have to change settings to limit that.

The answer turns on which account you are signed in with, more than on which brand you picked. A chief of staff who pastes a trust summary into a personal Gemini account and one who pastes it into Gemini through the family office's Workspace domain are under two different sets of terms.

## The short answer, plan by plan

The table summarizes each provider's own help and privacy pages, accessed 2026-10-15.

| Product and account | Trains on your chats by default? | How long the provider keeps them | Human review |
|---|---|---|---|
| Claude Free, Pro, Max (personal) | Only if you allow it in Privacy Settings | 30 days after you delete a chat; up to 5 years in de-identified form if you allow training | Conversations flagged for safety review can be analyzed |
| Claude for Work (Team, Enterprise) and the Anthropic API | No | API inputs and outputs deleted within 30 days; Enterprise admins can set custom chat retention (minimum 30 days) | Content flagged as violating the Usage Policy kept up to 2 years |
| Microsoft Copilot, personal Microsoft account | Yes, unless you opt out (some groups are excluded) | Conversation activity stored 18 months by default | Some conversations get automated and human review; you can't opt out of it |
| Microsoft Copilot and Copilot Chat, work Entra ID account | No | Stored with your other Microsoft 365 content; admins set retention in Purview | Copilot services have opted out of Azure OpenAI abuse monitoring, which includes human review |
| Gemini apps, personal Google account | Yes, when Keep Activity is on | 18 months by default (3 or 36 months selectable); 72 hours with Keep Activity off; reviewed chats up to 3 years | A subset of chats is reviewed by people, including Google's service providers |
| Gemini in Google Workspace, work or school account | No, without your permission | Admins set it: 90 days to indefinite in Workspace apps, up to 36 months in the Gemini app | Not human reviewed outside your domain without permission |

## Is Claude safe for confidential information?

Anthropic splits Claude into consumer and commercial products, and the rules differ.

**Claude Free, Pro and Max.** Since Anthropic updated its consumer terms on August 28, 2025, consumer users choose whether Anthropic may use their chats and coding sessions to train models. If you allow it, Anthropic keeps that data for up to five years. If you don't, the 30-day retention period continues. Three exceptions apply even with training off: conversations flagged for safety review, any thumbs up or down rating (Anthropic stores the whole related conversation for up to five years, de-linked from your user ID), and programs you join on purpose, such as the Trusted Tester Program. Incognito chats are not used to improve Claude, whatever your setting.

**Claude for Work and the API.** Anthropic says it does not use inputs or outputs from its commercial products to train models by default. The consumer terms change did not apply to Claude for Work, the API, or Claude through Amazon Bedrock and Google Cloud Vertex AI. API data is deleted within 30 days. On Enterprise, chats are kept until deleted unless an owner sets a custom retention period, and the minimum is 30 days. Feedback you submit is still kept for five years, and flagged content for up to two years.

If you need more, Anthropic offers zero data retention, but only on eligible APIs and Claude Code for Enterprise, not on the Claude apps. Our guide to [zero data retention AI](/blog/zero-data-retention-ai/) covers its exceptions.

## Is Microsoft Copilot safe for confidential information?

Microsoft sells two Copilots under one name, and the account you sign in with decides which terms apply.

**Copilot at work.** When you sign in with an organizational Entra ID account, Microsoft acts as a data processor under its Data Protection Addendum. Microsoft states that prompts, responses and data accessed through Microsoft Graph "aren't used to train foundation LLMs." Copilot only surfaces content you already have at least view permission to, which also means it surfaces everything you do have permission to: an over-shared SharePoint folder becomes an over-shared Copilot answer. Two exceptions are worth knowing. Web search queries go to Bing, which Microsoft runs as an independent data controller outside the Data Protection Addendum. And Anthropic models inside Copilot sit outside the EU Data Boundary.

**Copilot on a personal Microsoft account.** Microsoft uses data from Copilot, Bing and MSN for AI training unless you opt out under Privacy, then "Training on conversation activity." Users signed in with Entra ID, Microsoft 365 Personal and Family subscribers, signed-out users and under-18s are excluded. Microsoft stores conversation activity for 18 months by default, and some conversations get human review that you cannot turn off. Opting out of training does not stop use for advertising, safety and other product improvement as the Microsoft Privacy Statement describes.

For a family office running on Microsoft 365, our comparison of [Copilot for family offices](/blog/copilot-for-family-offices/) goes further.

## Is Gemini safe for confidential information?

**Gemini on a personal Google account.** Google's Gemini Apps Privacy Hub says a subset of chats is reviewed by human reviewers, including trained service providers, and that chats saved with Keep Activity on are used to develop models. Google's own advice: don't enter anything confidential or that you "wouldn't want a reviewer to see." Activity auto-deletes after 18 months by default. With Keep Activity off, Google still keeps chats for 72 hours, and chats a reviewer has seen stay for up to three years, disconnected from your account.

**Gemini in Google Workspace.** With a work or school account, Google treats prompts as customer data under its Cloud Data Processing Addendum. Google says your content is not human reviewed or used for model training outside your domain without permission, and Gemini retrieves only files and emails the user can already open. Admins set retention.

## Does ChatGPT train on my data?

On Free, Go, Plus and Pro, it can, unless you turn off "Improve the model for everyone." ChatGPT Business, Enterprise, Edu and the API do not train on your data by default. Our guide [Is ChatGPT safe for confidential information?](/blog/is-chatgpt-safe-for-confidential-information/) covers retention, memory and the New York Times preservation order.

## A checklist before your team uses any of them

- Confirm everyone signs in with the work account. A personal Claude, Microsoft or Google account on the same laptop carries consumer terms.
- Ask your admin what retention is set: Purview for Copilot, the Workspace admin console for Gemini, organization settings for Claude Enterprise.
- Tell staff not to use thumbs up or down on sensitive chats in Claude, since a rating keeps the conversation for five years.
- Fix file permissions before you switch on Copilot or Gemini across your mail and documents, because both answer from whatever the user can open.
- Decide whether web grounding is acceptable. Copilot's Bing queries fall outside the Data Protection Addendum.
- Read the data processing terms, and for legal or regulated work, your professional rules. This is general information, not legal advice.

## When the plan settings aren't the whole question

Each work plan above improves one thing: what the provider does with the text you send. The data still sits on the provider's servers for a period, its staff can reach it in defined cases, and a court can order it preserved. The broader design goal, where the provider keeps no copy and cannot see the outputs, is called [private superintelligence](/blog/what-is-private-superintelligence/).

Private SuperIntelligence from Mitosis Labs gives you the most capable AI, trusted with everything: an AI concierge for enterprises, family offices and private clients that works across your own email, documents and data and answers with citations. Each deployment is built for one client, and it captures none of that client's data. Read how it works for [family offices](/family-offices/), or [register interest](/#waitlist).

## Frequently asked questions

### Is Claude safer than ChatGPT for confidential information?

On consumer plans the two are close: both let you turn off training, and both keep rated conversations. Claude's consumer retention is 30 days with training off, and 5 years with training on. On business plans, neither trains by default, so the deciding questions become retention settings and contract terms.

### Does Claude train on my data?

On Claude Free, Pro and Max, only if you allow it in Privacy Settings, or if you rate a response, or if a chat is flagged for safety review. Anthropic says it does not train on Claude for Work or API data by default.

### Is Microsoft 365 Copilot safe for confidential data?

Microsoft says prompts, responses and Graph data are not used to train foundation models, and Copilot respects your existing permissions and sensitivity labels. The main risks are over-shared files and web queries sent to Bing, which sit outside the Data Protection Addendum.

### Do Google employees read my Gemini chats?

On a personal account, a subset of chats is reviewed by people, including Google's service providers, and reviewed chats are kept for up to three years. On a Workspace work account, Google says content is not human reviewed outside your domain without permission.

### Which is the most private: Claude, Copilot or Gemini?

On work accounts, all three commit to no training by default and give admins retention controls, so the answer depends on which suite your firm already runs and how it is configured. On personal accounts, all three can use your chats in some cases, so check the training setting on each before anything confidential goes in.


## Related reading

- [Is ChatGPT safe for confidential information?](/blog/is-chatgpt-safe-for-confidential-information/)
- [What is zero data retention AI?](/blog/zero-data-retention-ai/)
- [How to keep company data private when using AI](/blog/keep-company-data-private-with-ai/)
- [Private AI vs public AI](/blog/private-ai-vs-public-ai/)
- [Private ChatGPT for business: the options](/blog/private-chatgpt-for-business/)
- [What is confidential AI?](/blog/confidential-ai/)
- [What is private superintelligence?](/blog/what-is-private-superintelligence/)

## Sources

- [Updates to Consumer Terms and Privacy Policy (Anthropic, August 28, 2025)](https://www.anthropic.com/news/updates-to-our-consumer-terms), accessed 2026-10-15
- [Is my data used for model training? Consumer products (Anthropic Privacy Center)](https://privacy.claude.com/en/articles/10023580-is-my-data-used-for-model-training), accessed 2026-10-15
- [How long do you store my data? (Anthropic Privacy Center)](https://privacy.claude.com/en/articles/10023548-how-long-do-you-store-my-data), accessed 2026-10-15
- [Is my data used for model training? Commercial products (Anthropic Privacy Center)](https://privacy.claude.com/en/articles/7996868-is-my-data-used-for-model-training), accessed 2026-10-15
- [How long do you store my organization's data? (Anthropic Privacy Center)](https://privacy.claude.com/en/articles/7996866-how-long-do-you-store-my-organization-s-data), accessed 2026-10-15
- [Configure custom data retention controls for Enterprise plans (Anthropic Privacy Center)](https://privacy.claude.com/en/articles/10440198-configure-custom-data-retention-controls-for-enterprise-plans), accessed 2026-10-15
- [I have a zero data retention agreement with Anthropic. What products does it apply to? (Anthropic Privacy Center)](https://privacy.claude.com/en/articles/8956058-i-have-a-zero-data-retention-agreement-with-anthropic-what-products-does-it-apply-to), accessed 2026-10-15
- [Enterprise data protection in Microsoft Copilot and Microsoft Copilot Chat (Microsoft Learn)](https://learn.microsoft.com/en-us/copilot/microsoft-365/enterprise-data-protection), accessed 2026-10-15
- [Data, Privacy, and Security for Microsoft Copilot (Microsoft Learn)](https://learn.microsoft.com/en-us/copilot/microsoft-365/microsoft-365-copilot-privacy), accessed 2026-10-15
- [Privacy FAQ for Microsoft Copilot (Microsoft Support)](https://support.microsoft.com/en-us/microsoft-copilot/privacy-faq-for-microsoft-copilot), accessed 2026-10-15
- [Gemini Apps Privacy Hub (Google)](https://support.google.com/gemini/answer/13594961), accessed 2026-10-15
- [Generative AI in Google Workspace Privacy Hub (Google)](https://knowledge.workspace.google.com/admin/gemini/generative-ai-in-google-workspace-privacy-hub), accessed 2026-10-15
- [Enterprise privacy at OpenAI (OpenAI, updated January 8, 2026)](https://openai.com/enterprise-privacy/), accessed 2026-10-15