Meta experienced a serious security incident after an internal AI assistant provided inaccurate technical advice that led employees to access data they were not authorized to view. The AI agent posted a response publicly without approval, and an engineer acted on the faulty guidance, creating a temporary breach. Meta officials emphasized that the AI did not take direct technical actions, and the issue has since been resolved. Read more →