An unexpected data breach has seen countless Claude conversations and Artifacts made publicly accessible via Google searches. Reddit users first flagged the issue last weekend after discovering that certain search operators could retrieve shared chats, revealing everything from health records to personal employee reviews.

The problem stems from Claude’s ‘share chat’ feature, which allows anyone with a link to view private conversations. While Anthropic claims responsibility lies with users posting links publicly, it highlights the potential for sensitive information to be exposed accidentally or maliciously.

Before the issue was addressed, reports surfaced of detailed medical reports and company documents being inadvertently shared. Even more concerning, some chats contained explicit content, sparking debates about AI-generated material’s limits and adherence to guidelines.

Anthropic asserts it does not share chat directories with search engines, but the incident raises questions about how effectively these measures prevent public exposure. This isn’t a one-off; similar issues were noted last year when hundreds of Claude chats appeared in Google searches before disappearing.

To avoid such mishaps, users are advised to review their shared content settings regularly. As AI becomes more prevalent, issues like this serve as stark reminders of the need for robust privacy controls and transparency measures.