Researchers Discover Thousands of Publicly Exposed Muse Chat Sharing Links via DNS Database; Meta Has Fixed the Issue

Cybersecurity research team Claroc discovered through passive DNS queries that Muse's conversation sharing feature generates web addresses from user conversation titles and deploys them under the cf.metaaiusercontent.com domain, resulting in over 6,800 sharing pages being publicly searchable. Some of these pages contain personal financial information, internal corporate documents, and confidential meeting preparation materials.

The core issue with this vulnerability lies in the lack of access control in the sharing link generation mechanism. Muse automatically generates URLs based on conversation titles but does not implement any authentication or permission verification, allowing anyone who knows the link to directly access the content. This is severely inconsistent with the security expectations for AI assistants handling user privacy data.

Muse is a personal AI assistant launched by Meta, positioned similarly to ChatGPT, providing services such as conversation and information retrieval for individual users. AI assistant products typically handle a large amount of sensitive user information, including work documents and financial data, making the security of their sharing mechanisms critical. This incident exposes the neglect of privacy protection during the feature design phase of AI applications.

After discovering the issue, Meta suspended the sharing feature and stated that it is investigating improvement plans. This incident serves as a reminder to AI product developers that when launching features involving user data sharing, access control, link randomization, and expiration mechanisms must be implemented as fundamental security design elements, rather than as after-the-fact remedies.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.