In late July 2026, people in a Claude community on Reddit noticed something they did not expect. A plain Google search was returning other users' shared Claude conversations. Not their own private history. Real chats that other people had shared, sitting in the open, findable by anyone.
Some of what showed up was harmless. Some of it was not. According to Fortune, the exposed conversations included sensitive personal and business information, including cryptocurrency wallet keys, personal names and addresses, work notes, and programming chats.
Nobody got hacked. No password was stolen. That is the part worth understanding, because it points to a habit every one of us can build today.
What actually happened
Claude has a "Share" button. When you use it, Anthropic's own support doc explains that the app creates a snapshot of the conversation and "anyone with the link can view the chat snapshot," including "all messages that were sent prior to sharing the chat, including any artifacts."
So a share link is a live web page. Anyone with the link can open it, the same way they would open any URL.
The trouble started with how those web pages met Google. Anthropic tried to keep the shared pages out of search. The company blocked the share paths in its robots.txt file and also set a header telling search engines not to index the pages. Search Engine Journal explained the catch: a robots.txt block stops Google from reading the page at all, which means Google never sees the "do not index" instruction. And a page Google cannot read can still appear in search results when other websites link to it. As Search Engine Journal explained, when other sites link to a blocked page, "Googlebot notes the URL without actually opening it."
Two safety measures were in place. They canceled each other out.
Anthropic's read on this is that the system did what people asked it to do. A company spokeswoman, Amie Rotherham, told The Next Web the shareable links "are not guessable or discoverable unless people choose to share them themselves," and "When someone shares a conversation, they are making that content publicly accessible."
This is the second time
Here is the detail that makes it a lesson instead of a one-off headline. It happened before.
In September 2025, Forbes reported that hundreds of Claude transcripts had surfaced in Google search. Google estimated it had indexed just under 600 conversations at that point. The 2025 statement from Anthropic reads almost word for word like the 2026 one: the company said it does not hand chat directories or sitemaps to search engines and actively blocks them from crawling its site.
Same feature. Same outcome. Nearly the same answer.
In both cases the results dropped out of Google after the reporting. But anyone who already saved the content, or who still holds a direct link, can reach it. A page that was public for a weekend does not fully return to private just because it left the search results.
What this means for a small-business owner
Most of us who use these tools are not thinking about robots.txt files. We are drafting a client email, cleaning up a spreadsheet, working through a pricing question. Then we click "Share" to send a colleague the good version, and we assume it works like forwarding an email.
It does not always work that way. On Claude, and on other AI tools with a share feature, that button can create a public web page with its own URL.
The practical rule is simple. Treat anything you Share as public. Do not paste a wallet key, a password, a Social Security number, a payroll file, or a client's private data into a chat you might later share. If it would hurt to see it on a search results page, keep it out of a shared link.
Treat anything you Share as public.
There is one piece of good news in Anthropic's own doc. If a chat contains an attached file, "the file itself is not included in the shared snapshot and remains private." The words you typed and the artifacts Claude built can go public through a share link. The uploaded file does not ride along.
The 60-second self-audit
You can check your own account right now.
- Open Claude and go to Settings.
- Click Privacy.
- Find Shared chats and click Manage next to it.
- Review the list. Anything you no longer want public, click Unshare next to it.
Anthropic's doc says unsharing "revokes access to the last snapshot you shared." That closes the public page. Do this for any old share link you have forgotten about, especially ones from projects that touched real numbers or real names.
Then keep the habit going. Before you share the next chat, read the sentence next to the button. Know what it does before you click it.
Where this fits in getting good at AI
At LaunchReady we teach a model called The 7 Levels of AI Proficiency. The first level is about awareness. It is the point where you understand what a tool actually does when you press its buttons, not what you assume it does.
This whole episode lives at that first level. The tool behaved exactly as its documentation described. The shortfall was in knowing what the documentation said. Nobody needs to be technical to close that. You need to read the label and build one small habit around it.
That is the quiet advantage of learning AI on purpose instead of by accident. You stop being surprised by the buttons.
Your next step
Take the 60 seconds. Open Settings, go to Privacy, click Manage next to Shared chats, and look at what is public. Then carry the habit into your team. The people you work with are pasting things into these tools every day, and a quick shared reminder about the Share button is worth more than any policy document.
Related reading: Level 1: The Cadet.
Sources
- Anthropic Support: Sharing and unsharing chats
- Fortune: A trove of users' seemingly private conversations with Anthropic's Claude AI chatbot showed up in Google search results
- The Next Web: Claude shared chats and artifacts indexed by Google search
- Search Engine Journal: Indexed Claude chats show why Disallow is not noindex
- Forbes: Hundreds of Anthropic chatbot transcripts showed up in Google search
Frequently Asked Questions
Was this a data breach or a hack?
No. According to Anthropic, the content became public because people used the Share feature, and Google indexed those public pages through links on other sites. No account was broken into.
Are my private, unshared Claude chats affected?
The reporting and Anthropic's statements point to shared conversations, the ones a user chose to publish with a link. Anthropic's own support doc also notes that messages sent after you share a chat "remain private by default."
If I unshare a chat, is it fully gone?
Unsharing revokes access to the snapshot you shared, which takes down the public page. It cannot recall a copy someone already saved or a link someone already holds. The action still counts, and the earlier you take it the better.
Does this only happen with Claude?
A share feature that creates a public URL is common across AI tools. The safe move is the same everywhere: assume a shared chat can be seen by anyone, and keep sensitive information out of it.
Find your AI Proficiency level
The free 7 Levels assessment places you across seven stages of AI capability. Under ten minutes. Research-backed scoring.