A Meta AI service that discovers sensitive information when sharing online conversations has shown a lack of protective behavior regarding its users' privacy. Researchers have identified 6,800 URL links used to share conversations with others, often containing sensitive information. This problem is due to a specific software design that makes URLs easily accessible, compromising user security.

The sensitive information found includes notes prepared by candidates for a high-quality research appointment, financial analysis internal reports of a marketing firm, loan requests from business owners, bank transaction lists, and employee vacation schedules of an real estate selling company. These sensitive information was discovered through the detection of 6,800 URL links created by users to share conversations with others. Meta's response to this problem was to replace the service with a new model of AI assistance called 'Muse'.

Mark Zuckerberg, CEO of Meta, stated: 'This service was built from the beginning to protect privacy and security.' However, researchers noticed that the Muse URL system was inadequately protected in terms of privacy. Moreover, the management of URLs was entrusted to a team of coders, which could easily identify and manipulate sensitive information. The problem became so serious that Meta decided to stop the service of sharing online conversations and inspect it more closely.

Meta's responsible parties promised to improve the efficiency of URL systems and create a more effective system against the infiltration of sensitive information. The consequences of the problem were complicated. Sensitive data can be used by anyone who has their URL, compromising user privacy.

This problem has led to a series of incidents and privacy violations, pushing many users to change AI assistance services. The incident of Meta Muse demonstrated the need for a stricter approach to data privacy and AI assistance systems. Meta's responsible parties have committed to improving the security of URL systems and protecting the privacy of its users.