Your Shared Claude Chats Might Be on Google

Your Shared Claude Chats Might Be on Google — Here's How to Check

Key Takeaway: Over the weekend, people on Reddit and X discovered that shared Claude conversations and Artifacts — the apps, documents, and tools you build inside Claude — were showing up in regular Google search results. Some of what got found included medical reports, kids' names and phone numbers, internal company documents, and crypto wallet keys. Anthropic says the sharing system worked as designed and the exposed chats had been posted somewhere public, like a forum or social post. Either way, it's worth five minutes of your time to check your own account.

If you've read our Local AI for Bloggers post about keeping client work off cloud AI servers, this is exactly the kind of scenario that post was warning about — and it applies just as much to shared links as it does to what you type into a chat.

What Actually Happened

Claude gives every chat a shareable link if you choose to make one — a normal, useful feature for sending a conversation to a colleague or friend. The problem: hundreds of those links ended up searchable on Google and Bing, meaning anyone could stumble onto them through a simple search, not just the person you meant to share them with.

Reporters who reviewed the exposed content found a real mix — programming chats, work notes, book reviews, and also genuinely sensitive material: a detailed medical report with a real patient's information, internal-only company documents, and personal details like names and phone numbers of children. One conversation that generated explicit content, against Anthropic's own content policy, was also found.

Proof Block: Go to your Claude account settings, open the Privacy tab, and look at your list of Shared Chats and Shared Artifacts right now. Screenshot what's there. If you don't recognize something on that list, or you shared it a long time ago and forgot about it, that's exactly the kind of link this incident is about.

Why This Happened (In Plain Terms)

Warning: Anthropic's explanation is that these links are "not guessable" and only end up in search results after someone posts them somewhere public, like a forum thread or social media post — not because Anthropic handed a sitemap of shared chats to Google. That may be true technically, but the practical result is the same either way: a shared link isn't automatically the same thing as a link only your intended recipient will ever see.

This also isn't a first for the industry. A similar thing happened with ChatGPT conversations in 2025, and Meta had a related incident with its own AI chatbot the same year. If anything, this is turning into a pattern worth remembering across every AI tool, not just Claude: sharing features that feel private can behave more like a public webpage once a link exists.

How to Check and Protect Yourself Right Now

Quick Win: This takes less time than reading the rest of this article.

1. Open Claude and go to your account settings.
2. Find the Privacy section.
3. Look under Shared Chats and Shared Artifacts.
4. Review every link on that list. If you don't need it public anymore, or don't remember creating it, click Unshare.

Do this even if you're fairly sure you've never shared anything sensitive. It costs nothing, and it's the only way to actually confirm your own exposure instead of assuming you're fine.

What to Do Differently Going Forward

• Treat every "share" button on any AI tool as if it might become public, not just shareable with the specific person you have in mind. That's the safest working assumption until platforms prove otherwise.
• Never put real client names, medical details, financial account numbers, or anyone's personal contact information into a chat you might ever share, even temporarily. If you need AI help with something containing real sensitive data, either strip that data out first or keep the conversation private.
• If you handle client work regularly, this is a good moment to revisit whether a local AI setup makes sense for the sensitive pieces of your workflow specifically, even if you keep using cloud tools like Claude for everything else. Our Top 5 Local AI Tools post covers what that setup actually looks like.

Is This a Reason to Stop Using Claude?

Not really, on its own. This is specifically about the sharing feature, not a breach of private, unshared conversations. If you've never used Claude's share button, this incident doesn't affect you directly. That said, it's a fair reminder that trust in any AI platform's privacy protections should come with regular, active checking — not a one-time decision you make and then forget about.

FAQ

Q1. Were unshared, private Claude conversations exposed too? 

Reports specifically describe this affecting chats and Artifacts that had been shared through Claude's share feature, not private conversations that were never shared at all.

Q2. How do I know if one of my old shared chats is still public? 

Check your Claude account's Privacy settings under Shared Chats and Shared Artifacts. Anything listed there that you don't actively want public should be unshared.

Q3. Did this happen to other AI chatbots too? 

Yes — a similar issue affected ChatGPT conversations in 2025, and Meta's AI chatbot had a related public-exposure incident the same year. This isn't unique to Claude.

Q4. Has Anthropic fixed the issue? 

As of the most recent reporting, the specific search technique people used to find these chats no longer surfaces them, though the broader question of how shareable links get indexed by search engines remains something to stay cautious about going forward.

Pro Tip: Set a recurring reminder every couple of months to check the Shared Chats and Shared Artifacts list on any AI tool you regularly use this way. Sharing links tend to pile up quietly over time, and a quick periodic cleanup is a lot less stressful than discovering an old one publicly indexed months later.