Can people see your ChatGPT chats?

Whether people can see your ChatGPT chats comes down to two questions: did you share a link to the conversation, and does the account or the computer belong to your employer? The same two questions decide it for Claude, Gemini and Copilot, though each tool's rules differ.

Are ChatGPT shared links public?

A link created from a personal ChatGPT account (Free, Go, Plus or Pro) opens for anyone who has the link. Shared links have no per-person access controls and no expiry dates. A personal link is a snapshot: messages you add later are not included unless you open the conversation, select Share, then Update link. Links from ChatGPT Business (formerly called Team), Enterprise and Edu open only for members of the same workspace, and they can include messages added after sharing.

Shared links are anonymous by default, though some older sharing experiences may show the creator's name. Supported images and uploaded files can appear in a shared conversation, which is why OpenAI says not to include health information, financial details, passwords or account numbers in shared content. Treat a share link like an attachment anyone can forward.

OpenAI says shared link pages are not intended for search-engine indexing, but that does not make a link private, and anyone who has it can forward it. If someone continues your shared conversation, ChatGPT creates a separate, private copy in their account, and deleting your link does not remove that copy. Deleting the conversation removes its link. To review your links:

1. Open Settings. Select your profile, then Settings.
2. Go to Data controls. Beside Shared links, select Manage.
3. Remove links. Delete one link, or select More actions, then Delete all shared links.

Can my employer see my ChatGPT history?

Yes, if you use a work account, and possibly on a work laptop or network. On ChatGPT Business, other members do not automatically see your chats, and usage analytics do not let admins read them. But the workspace owner can export your data, and OpenAI's privacy page says Business admins can view, access, export and delete conversations in the workspace.

On ChatGPT Enterprise and Edu, admins can get an audit log of conversations through OpenAI's Compliance API, which companies connect to their eDiscovery and security tools. Only the workspace owner can grant access to conversation messages. In Enterprise, temporary chats stay available through the Compliance API for 30 days.

Work computers and networks are a separate route. Microsoft documents tools (Microsoft Purview) that let a company audit prompts and responses sent to ChatGPT, Gemini and the consumer version of Copilot from company-managed devices or at the network level. So on a work laptop or work network, your company may be able to see what you type into AI sites even if you use your personal account. Keep personal questions on your own phone or computer.

Can OpenAI see your chats?

In limited cases, yes. On personal plans (Free, Plus and Pro), a limited number of authorized OpenAI staff and trusted service providers may access your content to investigate abuse or a security incident, give support you asked for, handle legal matters, or improve model performance unless you opted out. OpenAI's advice is not to enter sensitive information you would not want reviewed or used. OpenAI says it does not sell your data or share your conversations with advertisers.

To stop training on a personal plan, go to Settings, then Data controls, and turn off "Improve the model for everyone". It applies to new conversations. One exception: if you give a thumbs-up or thumbs-down, the whole conversation tied to that feedback may be used for training. By default, OpenAI does not train on ChatGPT Business, Enterprise or Edu content.

Temporary Chat keeps a conversation out of your history and memories, is not used for training, and may be kept for up to 30 days for safety purposes. When you delete a regular chat, OpenAI schedules it for permanent deletion within 30 days, unless it was already de-identified or OpenAI must keep it longer for security or legal reasons. Archiving a chat does not delete it. If Memory is on (Settings, then Personalization, then Memory), deleting a chat does not necessarily delete a memory made from it.

Who can see your Claude chats?

Claude chats are private unless you share them, your organization runs the account, or you allow training on a consumer plan. By default, Anthropic employees cannot access consumer conversations unless you send feedback or a review is needed to enforce Anthropic's Usage Policy.

You can share a Claude chat with specific people by email (the default, on all plans), or by public link on Free, Pro and Max. A public link is a view-only snapshot anyone can open without a Claude account; later messages stay private, and attached files are not included. The page tells search engines not to index it, but Anthropic's own advice is to treat a public link as public. Turn sharing off from the Share menu, or Settings, then Privacy, then Shared chats.

On Team and Enterprise, chats can only be shared inside the organization, and the Primary Owner can export the organization's data, including conversations. Enterprise also has a Compliance API that pulls chat data. Incognito chats are not saved to your history or used for training, but on Team and Enterprise they are included in those exports.

On Free, Pro and Max, Anthropic trains on your chats only if you allow it (Settings, then Privacy, "Help improve our AI models"), or if a chat is flagged for safety review. If you allow it, de-identified data may be kept for up to 5 years. Anthropic does not train on Team or Enterprise data by default.

Who can see your Gemini and Copilot chats?

On a personal Google account, human reviewers (including Google's service providers) read a subset of Gemini chats. Reviewed chats are disconnected from your account and kept for up to three years, even if you delete your activity, so Google says not to enter confidential information. With Keep Activity on, Google uses your activity to improve its services, including training AI models; turning it off stops future chats from being reviewed for that. Anyone with a Gemini share link can read that snapshot; delete links under Settings & help, then Your public links.

On Google Workspace editions where the Gemini app is a core service, chats are not reviewed by humans or used to improve AI models. But admins decide how long chats are kept (up to 36 months), and admins can keep, search and export Gemini conversations with Google Vault, even ones you deleted.

Microsoft 365 Copilot with a work account stores your prompts and Copilot's responses, admins can search them with Microsoft Purview, and they are not used to train the underlying models. On a personal Copilot account, some conversations are reviewed by people, activity is kept for 18 months by default, and you can opt out of training under Privacy, then Training on conversation activity (for future conversations only).

Who can read your chats, tool by tool

Here is the same answer for each type of account.

How do you keep your AI chats private?

A few settings and habits cover most of it. Before you share anything, also read the output itself: AI output can repeat names, numbers or details you did not mean to share.

• Delete old shared links. ChatGPT: Settings, then Data controls, then Shared links. Claude: Settings, then Privacy, then Shared chats. Gemini: Settings & help, then Your public links.
• Turn off training on personal accounts. ChatGPT: "Improve the model for everyone". Claude: "Help improve our AI models". Gemini: Keep Activity. Copilot: Training on conversation activity.
• Mind work devices. Company tools such as Microsoft Purview can audit prompts sent from work laptops and networks, whichever account you use.
• Leave sensitive details out. OpenAI and Google both advise against entering confidential information you would not want reviewed.

One more habit: a shared chat link shows the whole conversation up to that point (in ChatGPT, including supported images and uploaded files), when usually you only meant to share the result. If your audience is your own team inside one ChatGPT or Claude workspace, the built-in sharing is fine; see how to share a ChatGPT or Claude project with your team. If you only want people to read the finished report, SecondPage publishes it as a page with its own link, set to Restricted (only people you add), Your organization, or Anyone with the link, and nobody needs a ChatGPT or Claude account to read it. (Disclosure: SecondPage is our product, and you are reading a SecondPage page.)