Hundreds of conversations with Anthropic’s Claude chatbot were reportedly available to view online, creating fresh concerns about how people use AI tools for sensitive work and personal tasks. The issue centers on chats that were publicly accessible, rather than kept private as many users may have assumed.
Examples highlighted in the report suggest that some of the exposed material may have included confidential or unpublished information. In one case from April, a user asked Claude to help draft an unpublished blog post about cloud security that included details tied to a corporate project. Another conversation from last month was also described as publicly available, though the trimmed report does not provide its full contents.
The discovery adds to broader worries about privacy and data handling in generative AI systems. People often paste drafts, internal notes, planning documents, or business context into chatbots to speed up work, but incidents like this show how easily that information can become exposed if sharing settings or publication features are misunderstood.
For Anthropic, the report is likely to intensify scrutiny over how clearly AI platforms explain public versus private chat options. As businesses and individuals rely more heavily on assistants like Claude, even a limited number of publicly reachable conversations can raise serious questions about user awareness, platform design, and the risks of entering sensitive material into AI tools.
