Via tomsguide.com
Anthropic's shared conversation feature inadvertently let Google index sensitive user data, raising fresh questions about AI privacy and digital asset security.
If you’ve ever used Anthropic’s “share” feature on Claude to send someone a conversation snippet, there’s a chance Google was reading it too. Over the weekend of July 26-27, Reddit users discovered that shared Claude conversations were being indexed by Google’s search engine, making them discoverable to anyone with a browser and a site-specific query. Among the exposed conversations: personal legal discussions, some rather creative erotica, and, most alarmingly for this audience, at least one cryptocurrency wallet key.
What actually happened #
Here’s the thing. Claude’s default conversations remain private. The issue only affected conversations that users explicitly shared using Claude’s built-in sharing feature, which generates a public URL for any conversation a user wants to pass along.
Claude’s shared links created public snapshots of conversations, and without proper “noindex” meta tags to tell search engines to stay away, Google’s crawlers did what crawlers do. They indexed everything they could find.
Reddit users were the first to notice, performing site-specific searches that surfaced a trove of shared Claude conversations. Anthropic confirmed that the exposed links were indeed public snapshots, acknowledging the gap in their crawling safeguards.
Google has since de-indexed the relevant links, effectively removing them from search results.
This isn’t Anthropic’s first rodeo #
What makes this incident more frustrating than surprising is that Anthropic dealt with a similar problem in 2025. The company previously took steps to block web crawlers from indexing shared links after an earlier exposure incident. Somehow, those protections either lapsed or proved insufficient this time around.
Anthropic isn’t alone in this struggle. OpenAI’s ChatGPT has faced comparable privacy concerns with its own sharing features.
The crypto angle matters more than you think #
For someone whose conversation contained a cryptocurrency wallet key, the stakes are existentially different. A leaked private key doesn’t just compromise an account. It potentially grants complete, irreversible access to whatever digital assets are stored at that address. There’s no “forgot my password” flow, no customer support ticket, no fraud department to call. If someone finds your private key and drains your wallet, those funds are gone.
The fact that at least one wallet key was confirmed among the exposed conversations should serve as a blunt reminder: AI chatbots are not secure storage for sensitive financial information. If you’ve ever pasted one into an AI chat, shared or not, you’re trusting that platform’s security architecture with assets that have no recovery mechanism.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our