Meta Security Breach: Rogue AI Agent Triggers Unauthorized Access Chain Reaction
Meta experienced a security breach due to the work of artificial intelligence. This is reported by The Information. According to the publication, an internal AI agent interfered in the workflow of employees without permission and triggered a chain access error.
What is known
The incident started when one of the employees used an internal AI-based tool to analyze a colleague's request on the corporate forum. However, the system independently published a response with recommendations, even though it was not given such a task. Another employee followed these recommendations, which led to erroneous access expansion: some engineers received rights to systems they were not supposed to see.
Meta confirmed the fact of the incident and stated that there was no user data leak. According to an internal report, the breach lasted about two hours. It is also noted that additional technical problems exacerbated the situation, but there is no evidence of misuse of access or public disclosure of information.
This is not the first time autonomous AI tools have gone out of control. Previously, Amazon Web Services faced a massive failure, which was also associated with AI agents. Experts emphasize: the active implementation of such systems requires clear restrictions and control, otherwise, the risks to security will only increase.
Source: The Information
