Meta uses employee PC activity to train AI systems

Meta has strengthened its focus on artificial intelligence (AI) by implementing systems that analyze employees' daily computer activity. This approach addresses the need to develop more advanced AI models with real-world use cases, as the company accelerates its digital transformation and seeks new ways to organize work.

According to sources from Meta confirmed to TechCrunch, the company continuously records mouse movements, keystrokes, and clicks on specific applications. This monitoring is not indiscriminate: they use an internal tool that clearly defines monitored areas and includes security measures to protect sensitive information, ensuring data is only used for training AI agents.

The strategy aims for medium-term impact, with plans to create personalized agents that assist both employees and future external users. Additionally, Meta is developing a 3D AI avatar that replicates Mark Zuckerberg’s voice, gestures, and reasoning to improve internal communication and future interactions with content creators.

This avatar is trained using recent public images and statements from Zuckerberg to build a digital identity that reflects his presence and response style in real time. The goal is to strengthen internal connections, with plans to extend this technology to other users and creators later on.

The result is a system that reproduces Zuckerberg’s voice, facial expressions, and responds in his usual style, used in announcements and strategic messages. These tools have radically transformed Meta’s internal processes, including a personal agent Zuckerberg uses for decision-making and quick access to key information, removing intermediaries in logistical and coordination tasks.

Meta aims to accelerate organizational agility and foster its culture of innovation, with AI at the core of its future strategy.