Shared conversations from Anthropic’s Claude AI chatbot were indexed by Google Search over the weekend of July 26-27, making some user-generated chat snapshots discoverable through site-specific searches. The exposed material reportedly included personal legal discussions, explicit creative writing, and at least one cryptocurrency wallet key.

The incident involved Claude’s built-in sharing feature, not the platform’s default private chats. Claude conversations remain private by default, but users can choose to generate a public URL for a conversation they want to share with others. Those links create public snapshots of the selected chats, which means access controls and search-indexing safeguards become especially important once a user turns a private exchange into a shareable page.

The problem arose because the shared Claude pages were accessible to search engine crawlers and lacked sufficient “noindex” protections instructing Google not to include them in search results. As a result, Google’s crawlers indexed the public snapshots, allowing them to appear through targeted searches. A noindex directive is meant to keep a page out of search listings, but it is not the same as making the underlying page private.

Reddit users first identified the issue after using site-specific search queries that surfaced numerous shared Claude conversations. Anthropic confirmed that the exposed links were public snapshots and acknowledged a gap in safeguards designed to prevent crawler indexing.

Google has since de-indexed the affected links, removing them from search results.

Anthropic previously faced a similar issue

The exposure follows a comparable incident involving Anthropic in 2025. After that earlier case, the company took steps intended to block web crawlers from indexing shared Claude links. In this latest incident, those protections either did not remain effective or were not sufficient to stop indexing.

Anthropic is not the only AI company to face privacy concerns tied to sharing tools. OpenAI’s ChatGPT has also encountered similar scrutiny over shared conversation features and the visibility of user-generated links. The recurring issue across AI chat products is that sharing features can blur the distinction between a conversation intended for a specific recipient and a page that is technically reachable on the open web.

Crypto wallet key exposure raises added risk

The inclusion of at least one cryptocurrency wallet key among the indexed Claude conversations creates a distinct security concern. Unlike many online accounts, a crypto wallet controlled by a private key generally has no password reset process, customer support recovery channel, or fraud department capable of reversing unauthorized transfers.

If a private key is exposed, anyone who obtains it may be able to access and move the digital assets associated with that wallet address. Such transactions can be irreversible once completed on-chain.

The incident underscores that AI chatbot conversations, whether shared or not, should not be treated as secure storage for sensitive financial information. Users who paste wallet keys or similar credentials into chatbot sessions are relying on the platform’s security design to protect assets that may have no practical recovery mechanism if compromised.