AI at WorkAI privacy compared: ChatGPT, Claude, Gemini and Copilot in one table
Companies describe privacy differently. This table puts the statements side by side, with the date they were read and the places where we found no answer.
The table (personal accounts)
Read on 13 and 14 November 2026 from the official pages listed below. "Not stated" means that the page we read did not say, not that nothing exists.
| Question | ChatGPT (OpenAI) | Claude (Anthropic) | Gemini (Google) | Copilot (Microsoft) |
|---|---|---|---|---|
| Can my chats improve the company's models? | Yes by default on Free, Plus and Pro; switch "Improve the model for everyone" | May be used if you allow it, if flagged for safety review, via feedback or by opting in | When Keep Activity is on, used to improve services by default | Not stated on the overview page |
| How do I stop it? | Settings, Data controls, turn off the switch | Privacy settings in your account | Turn off Keep Activity or use Temporary Chat | See "Your privacy controls and choices" |
| Temporary or incognito chat | Temporary chat: not in history, not used for training, may be kept up to 30 days for safety | Incognito chats are listed for individual plans | Temporary Chat: kept 72 hours, not used for training | Not stated on the overview page |
| How long are chats kept? | Saved chats stay until you delete them | Not stated on the page we read | Auto-delete after 18 months by default (3, 36 months, or never) | See "Your activity history" |
| Do people read chats? | Not stated on the page we read | Chats flagged for safety may be analysed | Human reviewers review some data; reviewed chats kept up to 3 years | Not stated on the overview page |
| Does feedback matter? | Thumbs up/down: the whole conversation may be used for training | Provided materials (such as feedback) may be used | Reviewed feedback kept up to 3 years | Not stated |
| Business plans | Not trained on by default | Team plan: no training on your content by default | Workspace accounts follow your administrator | Work accounts follow your organisation |
How to use the table
- For personal use, the safest habit is the same for all: keep private identifiers out, use the temporary or incognito mode for sensitive questions, and turn off the training option if you do not want it.
- For work and clients, use a business or organisation plan that your employer has approved. Read its terms.
- Check the date. These pages change. Open the originals before you decide.
What is common to all of them
None of the companies' pages promises that a chat is secret. Chats are stored on servers, may be reviewed under some conditions, and can be exposed by account theft. The right standard is not "is it private?" but "would I be able to live with this being read?".
Sources
The four official pages named below, read on the dates shown; plus Anthropic's plans page for the incognito row. We did not read the companies' full privacy policies.
Which is the most private?
The pages do not allow a ranking. They describe different controls. Pick the one whose controls you understand and have set.
Do these rules apply in India?
The pages we read are global or US-oriented. Indian law also applies to how businesses handle personal data; see the DPDP Act page.
Sources
- Data controls in ChatGPT
- How do you use personal data in model training? (Anthropic)
- Gemini Apps Privacy Hub
- Privacy and responsible AI (Microsoft Copilot)
Read on 13 and 14 November 2026. Pages change; open the original before you rely on a statement.
This is an independent guide. It is not connected with OpenAI, Anthropic, Google, Microsoft or any other AI company; product names belong to their owners. Statements about companies are taken from their own help pages as read on 13 and 14 November 2026. The prompts are our own wording and were not run against every tool. Not legal, financial or professional advice.