Last weekend, security researchers made a troubling discovery about the AI chatbot Claude. Hundreds of private conversations had become searchable on Google. Users who had shared their chats using Claude's built-in feature were affected. The shared links, which were intended for specific people, had been indexed by search engines. This meant that anyone could find and read these supposedly private conversations.
The exposed chats reportedly contained highly sensitive information. Researchers found personal details, business documents, health records, and even cryptocurrency keys. Some conversations included children's schoolwork that had been submitted for feedback. Had Anthropic, the company behind Claude, added proper technical safeguards earlier, this exposure could have been prevented. The root cause was a missing piece of code that should have blocked search engines.
Anthropic responded by saying the system was working as intended. The company stated that it does not share chat directories with search engines. However, critics argue that users were not adequately warned about the risks. Many people assumed that sharing a link meant only the recipient could view it. This incident is not unprecedented, as similar problems have occurred with ChatGPT and Grok.
By Monday, Google had begun removing the indexed conversations from its search results. Users who are concerned about their privacy can check their settings in Claude. They should navigate to Settings, then Privacy, and review their Shared Chats. This situation serves as a crucial reminder about digital vulnerability. Whatever we share online, even through AI tools, might not remain as private as we expect.
