Security company Malwarebytes recently pointed out that shared conversations in Anthropic's Claude can be indexed by Google Search, leading to the accidental exposure of sensitive information in user chat content. This issue was first discovered by some Reddit users, who found a large number of shared Claude conversation records in Google search results through specific search methods, including personal identity information and corporate secrets.
After investigating, Malwarebytes found the root cause to be Claude's "Share Chat" feature, which generates public access links for conversations users choose to share and sets the status to "Public." These public links are then crawled and indexed by search engines, ultimately being exposed in search results.
Malwarebytes emphasized that this is not an issue unique to Claude. The company had previously found shared chat records from Grok appearing in Google search results, and Meta AI had also experienced similar situations in the past. This means that the risk of search engine indexing is widespread among the dialogue sharing features of major AI platforms. Users may inadvertently expose private conversations without realizing it. Malwarebytes noted that it is difficult to completely avoid public web links being indexed, as there are inherent limitations in technical protections.
The Privacy Blind Spot in AI Conversation Sharing Features
The shared conversation feature was originally intended to allow users to conveniently showcase AI interaction results, such as sharing interesting answers or demonstrating workflows. However, there is a significant discrepancy between the "public" setting and users' expectations. Most users, when clicking the "share" button, are unaware that the generated link will be permanently indexed by search engines, let alone expect that the content will appear in anyone's search results. The uniqueness of AI conversations lies in the fact that users tend to input real work content and personal information within them. Once indexed, the leaked information is far more sensitive than typical social media posts.
Malwarebytes provided direct and effective protective recommendations: do not publicly share AI chat content, and avoid entering personal identification information, account passwords, and corporate secrets in AI conversations. However, this recommendation essentially requires users to change their habits rather than addressing the issue at the platform level. A more reasonable solution should include default settings that prevent shared links from being indexed, provide password protection options, and clearly inform users of privacy risks before sharing. As the amount of information carried by AI conversations increases, platform providers have a responsibility to prioritize privacy protection in product design, rather than leaving users to bear the consequences of information leaks alone.
