E-Paper
Sign In
Claude chats appeared on Google: How to keep AI conversations private
Here is how to check whether you shared a Claude or ChatGPT conversation and what to do if sensitive information appears on Google
- PUBLISHED: Wed 29 Jul 2026, 4:00 PM
AI chats can contain everything, from work documents and medical questions to financial information and deeply personal conversations. A recent Claude privacy scare is a reminder that one careless tap on the share button can make that information far more public than expected.
Claude conversations and Artifacts were recently found in Google Search results after users created publicly accessible links. Some reportedly contained health records, internal company material and personal contact details. The important distinction is that private, unshared conversations were not exposed: the affected pages had been made accessible through Claude’s sharing or publishing features.
Stay up to date with the latest news. Follow KT on WhatsApp Channels.
Recommended For You
Here are key points to follow:
1. Review every Claude chat you have shared
Open Claude, select your profile and go to Settings > Privacy. Find Shared chats, select Manage, and review the list of conversations with active public links.
Select Unshare beside anything that no longer needs to be available. You can also open an individual conversation, select Share, change its visibility from Public to Private, and disable the link.
Claude’s shared links contain a snapshot of everything in the conversation up to the moment it was shared, including Artifacts displayed within that chat. Anyone who obtains the URL can view that snapshot.
2. Unpublish Claude Artifacts separately
Documents, apps, code and other content created through Claude Artifacts can also be published publicly.
Open the Artifacts section from Claude’s sidebar, review previously published creations and select Unpublish where necessary. Anthropic warns that publicly published Artifacts can be viewed and used by anyone with the link, including people without a Claude account.
3. Delete old ChatGPT shared links
In ChatGPT, open your profile and go to Settings > Data Controls > Shared links > Manage. From there, delete individual shared conversations or remove every active link.
ChatGPT does not currently provide passwords, expiry dates or granular access permissions for normal shared links. Anyone with the URL can view the shared content and forward it to someone else. Deleting the original conversation also disables its shared link, although copies previously imported into another user’s history may remain there.
4. Stop your chats being used for model improvement
Public sharing and AI training are different privacy issues. Removing a public link does not automatically change whether a platform may use future conversations to improve its models.
On Claude, go to Settings > Privacy and switch off Help Improve Claude. Anthropic says new consumer chats will no longer be used for future model training after the setting is disabled, except in limited safety-review circumstances.
On ChatGPT, open Settings > Data Controls and switch off Improve the model for everyone. Your conversations will remain in your history but will not be used to train OpenAI’s models.
5. Use temporary modes for sensitive questions
Claude offers Incognito chats, accessed through the ghost icon when starting a conversation outside a project. These chats are not added to normal history or memory and are not used to improve Claude.
ChatGPT’s Temporary Chat mode also avoids chat history, memory and model training. OpenAI says Temporary Chats are automatically deleted from its systems after 30 days, although they may be reviewed during that period for abuse monitoring.
Gemini users can similarly start a Temporary Chat and switch off Keep Activity from Gemini’s privacy settings to limit how future conversations are saved and used for personalisation or AI improvement.
6. Never paste information you cannot afford to expose
Temporary modes and privacy settings reduce risk; they do not make a consumer chatbot the right place for every document.
Do not upload passwords, OTPs, private encryption keys, bank details, medical reports, confidential contracts or unredacted identity documents. For UAE users, that also means avoiding passport copies, Emirates ID numbers, visa pages and confidential workplace files.
Replace names and account numbers with placeholders wherever possible. Before sharing a link, open it in a private browser window and read the entire page as though you were a stranger.
7. Act quickly if a chat appears on Google
First, revoke the Claude, ChatGPT or Gemini sharing link at its source. If the page has disappeared but remains visible in Google results, submit its URL through Google’s Refresh Outdated Content process.
Google also accepts removal requests for certain sensitive information, including government identification numbers, banking details, passwords, medical records, addresses and phone numbers. Removing a result from Google does not erase copies that may exist elsewhere, which is why disabling the original public link should always come first.
The safest rule is to treat every AI conversation like an email that could someday be forwarded. If seeing it publicly would cause harm, do not put the original information into the chatbot.
Want to know more about your favourite tech products, brands, and video games? Check out the links below or click here to see Khaleej Times' coverage on technology.
