Claude Shared Chats Appeared in Google Search: What Organisations Need to Know

Written by Catarina Santos

Some publicly shared Claude conversations have appeared in Google Search results. This article explains what happened, why private chats are not affected by default and what organisations should check if employees use AI sharing features.

Claude Shared Chats Appeared in Google Search What Organisations Need to Know

Claude Shared Chats Appeared in Google Search: What Organisations Need to Know

Reports published on 27 July 2026 say that some publicly shared Claude conversations appeared in Google Search results.

This does not mean every private Claude conversation was exposed. Anthropic states that Claude chats are private by default. The risk relates to conversations that users deliberately chose to share by creating a public link.

Although this was not a security breach affecting all users, it is an important reminder that organisations need to consider how information is shared from AI tools, not just what employees enter into them.

What Happened to the Claude Conversations?

Claude allows users to create a shareable snapshot of a conversation.

When a user selects Share, Claude generates a public link that can be viewed by anyone who has access to it. Recent reports found that some of these publicly shared conversations had also appeared in Google Search results.

This increased the potential audience for those conversations. A link originally intended for a small number of people could become discoverable by individuals who had never received it directly.

Public links can spread beyond their intended audience. They may be shared on websites, forums or social media, making them accessible to search engines.

Further information is available from:

Were Private Claude Chats Leaked?

There is no evidence that all private Claude conversations became public.

According to Anthropic, conversations remain private unless a user deliberately creates a public share link. Once a conversation is shared publicly, anyone with the link can view it.

The concern raised by this story is that some public links became discoverable through search, rather than remaining accessible only to people who had been sent the link.

It is important to distinguish between:

  • A private conversation stored in a Claude account
  • A conversation that a user has deliberately turned into a public snapshot using a public link
  • A publicly shared conversation that later becomes discoverable through a search engine

This distinction matters. Organisations should respond to the real risk without suggesting that every Claude user has experienced a personal data breach.

What Information Is Shared?

According to Anthropic, a shared snapshot includes all messages exchanged before the conversation was shared. It may also include content created during the conversation.

Messages added after the snapshot is created remain private unless the conversation is shared again.

Anthropic also states that:

  • Attached files are not included in the shared snapshot
  • The conversation and Claude’s responses remain visible
  • Raw information retrieved through connected tools remains hidden
  • Users on Team and Enterprise plans can only share chats with members of the same organisation

However, even if an attached file is not shared, personal data or confidential information taken from that file may still appear within the conversation or in Claude’s responses.

Why Does This Matter for Organisations?

Many employees now use AI tools to help with everyday work, such as drafting documents, summarising information and preparing reports.

Those conversations may contain:

  • Personal data about customers or employees
  • Confidential business information
  • Internal plans or meeting notes
  • Client information
  • Security details
  • Unpublished financial or commercial information

Someone may understand that a colleague can open a shared link without appreciating that the link could later become accessible to a much wider audience.

This creates potential risks, including unauthorised disclosure, loss of confidentiality and possible exposure of personal data.

Not every shared conversation will amount to a personal data breach. Whether it does depends on the information involved, who could access it and the likely impact on the individuals concerned.

Even where no personal data is involved, the disclosure of confidential business information or commercially sensitive material may still create contractual, regulatory or commercial risks.

What Should Claude Users Do Now?

Anthropic provides a way for users to review their shared conversations.

  1. Open Claude and go to Settings.
  2. Select Privacy.
  3. Find Shared chats and select Manage.
  4. Review every conversation that has been shared.
  5. Remove any public links that are no longer needed.
  6. Check whether any shared conversation contains personal, confidential or security-related information.

Claude users can also access their shared conversations through the official Claude shared-chats page. Users may need to sign in before they can view or manage their shared conversations.

Removing a shared link prevents further access using that link. However, organisations should not assume that references to the page will disappear immediately from search engines or other websites.

If sensitive information has been exposed, the organisation should record the incident and begin its internal incident response process.

What Should an Organisation Do if Personal Data Was Shared?

If a shared Claude conversation contains personal data, the organisation should establish what happened as quickly as possible.

The immediate steps should include:

  • Disable the public share link
  • Preserve the information needed to investigate the incident
  • Identify the personal data involved
  • Establish who may have accessed the conversation
  • Assess the possible impact on the people concerned
  • Record the incident and the decisions made

The ICO states that organisations should assess the likelihood and severity of any risk to individuals’ rights and freedoms.

If a personal data breach is likely to result in a risk to people, it must be reported to the ICO without undue delay and, where feasible, within 72 hours of the organisation becoming aware of it.

Where the breach is likely to result in a high risk to individuals, the organisation may also need to inform those affected.

Not every incident needs to be reported. The decision should follow a documented risk assessment based on the circumstances of each case.

Use the ICO’s personal data breach assessment guidance.

What Should AI Policies Cover?

Many organisations focus on what employees are allowed to enter into an AI tool. That is only part of the picture.

An effective policy should also explain:

  • Whether employees are permitted to create public share links
  • What information must never be included in a shared conversation
  • Who can authorise external sharing
  • How shared links should be reviewed and removed
  • How long shared conversations should remain available
  • How employees should report accidental disclosure
  • Which AI tools and account types have been approved for business use

Organisations should also review sharing settings during AI procurement and risk assessments.

Any feature that allows information to be shared publicly or with anyone holding a link should be treated as an external disclosure unless suitable access controls are in place.

What Training Should Staff Receive?

Staff do not need a technical explanation of how search engines work. They do need to understand how AI platforms should be used, what information can be shared and the risks associated with public sharing.

Training should explain that:

  • AI conversations should not contain unnecessary personal data or confidential information
  • There is an important difference between saving a conversation privately and creating a public share link
  • Public links can be shared beyond the original recipient
  • Shared content may become searchable online
  • Removing the original link does not always remove every copy or reference
  • Suspected disclosures should be reported immediately

Using practical examples is likely to be more effective than simply telling employees to use AI responsibly.

Key Takeaways

The Claude shared-chat story is not evidence that every private AI conversation became public. It does show that a public sharing feature can create wider exposure than a user expects.

Organisations should review whether public AI conversations have been shared, update their internal policies where necessary and ensure employees understand the consequences of using sharing features.

Good information governance should cover the entire lifecycle of information. This includes what is entered into a tool, what the tool produces, who can access it, how it is shared and when it should be removed.

Sources

Need Support With AI and Data Protection?

If your organisation needs help reviewing AI use, internal policies or a potential personal data incident, speak to the Data Protection Made Easy team.

Speak to Our Team