Claude AI chats leaked on Google search, exposing crypto wallet keys. Reddit users found shared conversations indexed by Google. Anthropic confirmed the gap. If you pasted a private key into a shared chat, assume it was public.
Anthropic's shared conversation feature let Google index sensitive user data over the weekend of July 26-27, including at least one cryptocurrency wallet key. Reddit users discovered that shared Claude conversations were appearing in Google search results, making them accessible to anyone with a browser and a site-specific query.
The issue affected only conversations that users explicitly shared using Claude's built-in sharing feature, which generates a public URL. Those public snapshots lacked "noindex" meta tags, so Google's crawlers indexed them. Anthropic confirmed the exposure and acknowledged the gap in crawling safeguards. Google has since de-indexed the relevant links.
For crypto holders, the stakes are existential. A leaked private key grants irreversible access to digital assets at that address. There is no password reset, no customer support, no fraud recovery. If someone drains a wallet, the funds are gone. The confirmed presence of at least one wallet key among the exposed conversations reinforces a blunt lesson: AI chatbots are not secure storage for sensitive financial information.
Anthropic faced a similar problem in 2025, when it took steps to block crawlers after an earlier exposure. Those protections apparently lapsed or proved insufficient. OpenAI's ChatGPT has faced comparable privacy concerns with its own sharing features.
Anyone who has pasted a private key, seed phrase, or wallet address into a shared Claude conversation should assume it was publicly accessible. The risk is not limited to Anthropic. The convenience of AI chat features can conflict with the security requirements of self-custodied crypto assets.
No specific wallet addresses or amounts have been confirmed as drained. The exposure window was at least several hours, possibly longer, before Google de-indexed the links. Users who shared conversations containing sensitive data should treat those assets as compromised and move them to new addresses.
Anthropic has not disclosed how many conversations were exposed or whether it will implement automatic "noindex" headers on all shared links going forward. The company did not immediately respond to a request for comment on additional safeguards.
The incident follows a pattern of AI chat platforms treating shared conversation links as public by default, relying on users to understand the implications. For crypto owners, the margin for error is zero.
Drafted by a large language model from the source reporting linked above, then screened by automated publishing checks. It is not read by a journalist before publication. Some articles cite our Alpha Score. Verify prices and figures against the original source. Educational coverage, not personalized advice.