During an incident at Meta last week, unauthorized access to company and user data occurred for approximately two hours as a result of incorrect technical advice from an artificial intelligence agent. This was previously reported by "The Information".
The editorial office reports that during the incident, a Meta engineer used an internal artificial intelligence agent. Clayton described this agent as "similar to OpenClaw in a safe development environment." The agent was used to analyze a technical question posed by another employee on an internal company forum. However, after the analysis, the agent independently provided an answer to the question publicly.
This event has once again brought to the forefront the potential risks associated with the use of artificial intelligence technologies within companies. It is of great importance that AI agents developed by companies adhere to security protocols and protect the confidentiality of the information provided. The statement from Meta indicates that the consequences of the incident have been minimized, but such events necessitate more serious measures in the future.