Claude Privacy: Shared Chats and Artifacts May Have Been Indexed by Google

On July 27, 2026, TechCrunch reported that conversations and Artifacts that Claude users had shared via link may have ended up in Google's search index, making them accessible to people other than the intended recipients of the link.

Where the problem came from

The problem appears to stem from Claude's "share chat" feature (a public link to a chat). It lets users create a URL that anyone who has the address can use to open a conversation or an entire project. That is how the original news source describes the mechanism.

• TechCrunch published its report on July 27, 2026
• The source of the problem is the "share chat" feature in Claude from Anthropic
• The link gives access to anyone with the URL, with no password
• Both the conversations themselves and the Artifacts created in them were indexed
• The platform where the content surfaced is Google Search

"The problem appears to stem from

Claude's 'share chat' feature, which lets users create links that allow anyone with the URL to view a conversation or project." — from TechCrunch's report, July 27, 2026

Why this matters for users

A shared link in Claude is not password-protected: anyone who knows the URL can see the content, and search crawlers can add public pages to their index. According to TechCrunch (report dated July 27, 2026), Google results may have included both conversations with the model and the Artifacts built inside those chats. The difference between a "link for a friend" and a public web page disappears here: once the URL is reachable by Googlebot, the page can be indexed and shown in search results for a relevant query.

Artifacts are documents, code snippets and mini-apps generated in Claude, which the model displays in a separate panel next to the dialogue. If an Artifact contained work data, drafts or personal information, its appearance in a search index means the material became discoverable by keywords, not just via a direct link.

What to do about chat links

A public link to an AI chat is effectively a publication: the content is available to anyone who has the URL and, potentially, to search engines that crawl that address. According to TechCrunch's July 27, 2026 report, it was the sharing mechanism itself, not a hack, that caused seemingly private conversations to end up in open search.

The practical takeaway is simple: before sharing a Claude conversation via link, check it for sensitive data, and don't assume that "only a select few know the link."

What this means

The Claude episode is another reminder that "share via link" in AI services is closer to open publication than to a private message. For Anthropic, it is a reputational warning: privacy of conversations is one of the key arguments for paid AI assistants, and any hint of a leak through a standard feature hurts user trust.

Frequently asked questions

What is the "share chat" feature in Claude?

It is a Claude feature that creates a link to a conversation or project. According to TechCrunch, anyone with the URL can open and view the content via such a link — it requires no separate password.

Why did shared Claude chats end up in Google search?

According to TechCrunch's report of July 27, 2026, public links to chats and Artifacts created via "share chat" made it into Google's search index. As a result, the content could be found not only by the person who was given the link, but by any search user.