The Illusion of Privacy
27 October 2026
I love the sarcasm in this cartoon.
As we increasingly use personal AI assistants, it’s almost as if we are sitting inside a glass-walled room. We believe nobody outside can see us and we are having a private conversation, completely unaware that someone outside might be able to see what we are sharing.
There is an interesting story behind this cartoon.
On October 18, security researchers at Claroc published findings about the conversation-sharing feature in Meta’s Muse AI assistant.
They discovered that more than 6,800 user-created conversation links were discoverable through public DNS records.
Some of these conversations contained sensitive information, including business revenue reviews, loan applications and confidential meeting notes.
Meta subsequently suspended the sharing feature while reviewing the issue.
Something similar happened with OpenAI in July 2025, when thousands of shared ChatGPT conversations appeared in Google search results.
Different products, different times, different mechanisms, but a similar mistake, similar lesson.
Reference: Claroc — GhostShares: Muse and Passive DNS
