Google search engines were briefly able to index an unknown number of public conversations conducted via Anthropic’s Claude chatbot, raising widespread concerns about users’ privacy after conversations containing highly sensitive information appeared in search results.
Users revealed that they easily found conversations that contained medical records, internal work documents, and personal data, including children’s names and phone numbers, which re-highlighted the risks associated with sharing AI conversations via public links.
The problem is due to the Conversation Sharing feature in Cloud, which allows the user to create a public link that allows others to view the conversation or what are known as “Artifacts”, which are interactive documents and applications created within the platform.
Although the service clearly warns that anyone with the link can access the content, many users believed that these links would remain limited to the people they sent them to. (Erm News)
