Tons of Peoples’ Claude Chats and Creations are Exposed on Google
404 Media Joseph Cox ● Covered by 2 sources
Public share links from Claude are turning up in Google search, exposing user chats and creations to anyone. Many users likely had no idea their 'shareable' link meant fully indexable and public.
Based on reporting by 404 Media, Joseph Cox — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic's Claude has a share feature that lets users generate a link to a conversation or a project they built with the AI, presumably to show off a clever prompt or hand off some code to a colleague. The problem, according to 404 Media, is that a lot of those links are getting scooped up by Google's crawlers and dropped straight into search results, where anyone with the right query can stumble onto them.
That's a quiet but significant failure mode. A share link feels private in spirit — you're sending it to one person, not broadcasting it to the internet. But if the page isn't blocked from indexing, it behaves exactly like a public webpage, because that's what it is. Search engines don't know or care that the person who made the page assumed only a friend or coworker would ever click it.
The practical fallout is that strangers can now browse through people's actual Claude sessions and the things they built with them, sometimes with no idea the creator ever consented to that level of exposure. Chat transcripts can contain anything from mundane brainstorming to sensitive personal or business details, and once that's sitting on a public URL indexed by Google, pulling it back is a losing game — caches, scrapers, and archive sites tend to outlast any attempt at deletion.
This isn't a new category of mistake, either. Slack, Trello, and plenty of other tools have gone through the same cycle: a convenient sharing shortcut turns into an accidental privacy leak because the default settings favored visibility over caution. AI chat tools just add a new twist, since the shared material can include far more personal or revealing context than a typical shared doc, given how people tend to talk to chatbots like they're talking to no one at all.
My take — AI-written commentary, not fact-checked reporting
This is the same old lesson tech companies refuse to learn: if a 'share' button doesn't scream loud and clear that it means 'public, forever, searchable,' people will assume it means something safer. Anthropic markets itself as the safety-conscious AI lab, so an unforced UX error that leaks private chats onto Google is a bad look, and it's a reminder that safety branding means nothing if the basic plumbing isn't buttoned up.
Read more about this at: 404 Media