Your Claude Chats Might Be on Google. Here’s How to Check

TL;DR: Over the weekend of 26 July 2026, people discovered that publicly “shared” Claude conversations — including Artifacts (the interactive mini-apps and code Claude helps you build) — were showing up in Google search results, sometimes containing genuinely sensitive information. Nobody was hacked: this happened because Claude’s “share” feature makes a conversation public by design, and once a link circulates, search engines can index it. Anthropic says it’s fixed the current exposure. Here’s what to check on your own account, in about two minutes.

Claude shared chats on Google: what happened

On the weekend of 26 July 2026, people discovered that searching `site:claude.ai/share` on Google turned up hundreds of other people’s Claude conversations — fully readable, with no login or direct link needed. Fortune and Futurism both reported what turned up: medical records, children’s names and phone numbers, internal company documents, employee reviews, and code and Artifacts people had built with Claude.

The mechanism is simple, if easy to miss: Claude’s “share” button makes a conversation public — that’s the feature working as intended. The problem is what happens next: once a shared link is posted somewhere (a forum, a group chat, a tweet) or simply left publicly reachable, search engines can crawl and index the full page, surfacing it to anyone searching, not just the people the link was meant for.

Anthropic’s response was direct: “These shareable links are not guessable or discoverable unless people choose to share them themselves. When someone shares a conversation, they are making that content publicly accessible.” By Monday afternoon, the specific Google search technique people were using stopped returning results — but TechCrunch noted that some previously exposed links remained live and reachable to anyone who already had them.

Worth knowing: this isn’t a uniquely Anthropic problem. Fortune reported that a nearly identical issue has hit ChatGPT (around 100,000 conversations, in an earlier incident) and Grok’s chatbot too — “a problem the industry has struggled to permanently fix.” Anthropic itself had a smaller version of this in 2025 (roughly 600 exposed conversations).

Why it matters if you don’t code

If you use Claude — including Claude Code or Claude.ai’s Artifacts to build and share things — this is directly relevant, for one simple reason: Artifacts were specifically affected. If you’ve ever used Claude to build a small app, tool, or prototype and hit “share” to show someone, that’s exactly the kind of content that could end up searchable.

The honest calibration:

• You weren’t hacked, and Claude wasn’t broken into. Nobody’s account was compromised. This is about what happens after you choose to share something, not a security breach of Claude itself.
• Most people are fine. If you’ve never used the “share” button on a Claude conversation, none of this applies to you.
• If you have shared something, it’s worth two minutes to check — especially if that conversation ever contained anything personal, financial, or belonging to your business.

This connects to a lesson this site keeps coming back to, just from a different angle: we’ve written before about keeping your secrets out of AI tools’ reach when giving an AI access to your accounts. This is the mirror image — being careful about what you publish through an AI tool’s sharing features, since “share with one person” and “share with Google” turned out to be a thinner line than most people assumed.

How to check (two minutes)

1. Open Claude and go to Settings → Privacy → Shared Chats (or Shared Links, depending on your version).
2. Review anything listed there. If a chat contains personal details, business information, client data, or anything you wouldn’t want searchable, revoke or delete the share.
3. If you’re not sure whether something was ever shared, it’s safe to assume: if you never clicked “share,” you have nothing to check.
4. Going forward, treat the “share” button on any AI chat tool (Claude, ChatGPT, Grok, others) as “public on the internet,” not “private link for one person” — because that’s functionally what it can become.

Who should care (and who shouldn’t)

• Ever shared a Claude conversation or Artifact, especially one with personal or business info: check your settings today — this is the group that should actually act.
• Building apps or prototypes with Claude Code/Artifacts and sharing them for feedback: worth a quick audit of what’s in those shared links.
• Never used the “share” feature: nothing to do — this doesn’t affect you.
• Not sure which AI tool to build with: this doesn’t change your tool choice; every major AI chat tool has had a version of this issue. The quiz will still point you to the right fit for what you’re building.

Our take

This is a good example of a story that sounds alarming in headlines (“private AI chats exposed!”) but is more mundane, and more useful, once you understand the mechanism: a sharing feature did what it says on the tin, and the surprise is how far “shared” can travel once a link exists. It’s the same category of lesson as the recent OpenAI/Hugging Face story — not a reason to panic, but a nudge to be a little more deliberate about what you make public through any AI tool, this one included.

The two-minute check is genuinely worth doing if you’ve ever hit “share” on anything sensitive. Beyond that, there’s nothing here that should change how you use Claude day to day.

Not sure which AI tool actually fits how you build? Take the 60-second Vibe Coding Tool Finder quiz → security habits matter more than which tool you pick.

FAQ

Was my Claude account hacked?

No. This wasn’t a breach of Claude or your account — it’s about publicly shared conversation links being findable by search engines once they circulate, which is different from being hacked.

Does this affect Claude Code specifically?

It affects Claude’s “share” feature generally, which includes Artifacts — the interactive apps and code Claude helps you build and that people often share for feedback. If you’ve never shared a Claude conversation or Artifact, this doesn’t affect you.

Is this only a Claude/Anthropic problem?

No — Fortune reported a near-identical issue previously affected ChatGPT (around 100,000 conversations) and Grok. It’s a pattern across AI chat tools generally, not something unique to Anthropic, though this is also not the first time it’s happened to Claude specifically.