AI Agent API Bug Shakes X and Instagram Platforms: Users' Private Messages Temporarily Exposed Randomly

SAN FRANCISCO, KOMPAS.com - A serious cybersecurity incident shook global social media platforms **X (formerly Twitter)** and **Instagram (Meta)** this morning. A critical bug in the third-party API integration for **AI Agents** has reportedly caused random leaks of users' private messages (Direct Messages/DMs) in various regions of the world.

The first report regarding this vulnerability was published by an independent security researcher on platform X, who noticed that their autonomous artificial intelligence assistant began replying to messages using private conversation data from strangers' accounts.

Initial investigations indicate that this bug stems from a **race condition** issue in the social media platform's API gateway cache system when handling thousands of concurrent data polls from external AI Agent servers.

Technical Vulnerability Analysis

Cyber experts explained that the error occurred in the session state management architecture of the API gateway:

1. Multithreading Deadlock: Increased activity of AI Agents interacting with the API simultaneously triggered a deadlock in the session host server's memory.
2. Incorrect Session Caching: Under high load, the server incorrectly mapped user A's session token to the response of an AI Agent serving user B.
3. Metadata Exposure: In addition to the text message content, sensitive metadata such as usernames, unique account IDs, and message delivery timestamps were also temporarily exposed to external parties.