Meta CEO Mark Zuckerberg Admits to Monitoring Employee Operations to Train Llama
Meta CEO Mark Zuckerberg stated at a company-wide meeting that the company is recording employees' keyboard, mouse, and screen activities to train the Llama model's writing and programming capabilities. He claimed that Meta employees have an average intelligence significantly higher than outsourced personnel, and that AI learns by observing employees' actual work.
When employees questioned why they were not informed in advance, Zuckerberg responded, "Communicating all the details does not align with your strategic interests."
Twenty days after the recording was leaked, Meta sent layoff emails to approximately 8,000 employees worldwide while reallocating 7,000 employees to a newly established AI department responsible for developing AI counterparts for employees.
Source: Public Information
ABAB AI Insight
Meta has previously accelerated Llama iterations through large-scale internal data, and this public acknowledgment of monitoring behavior continues its path of converting employee production activities into AI training assets. The company has prioritized using high-quality internal data to enhance model capabilities in various projects.
From a capital perspective, Meta is directly converting labor costs into model advantages through monitoring and AI counterpart development. Although this move has sparked privacy and trust controversies in the short term, it significantly reduces the cost of acquiring external data and enhances overall operational efficiency through the deployment of internal counterparts, creating a closed-loop value transformation from employee behavior to company AI capabilities.
Similar to other tech giants that utilize internal data to train models, Meta is currently in an expansion phase transitioning from labor-intensive operations to a comprehensive replacement with AI counterparts. This incident exposes the structural tension between data privacy governance and the pursuit of efficiency.
Essentially, this represents technological substitution and capital concentration: monitoring employee behavior directly replaces traditional outsourcing and manual labeling, accelerating the concentration of AI training capital from dispersed labor to the Meta Llama platform, reshaping the company's internal productivity structure, privacy boundaries, and AI capability iteration model.
ABAB News · Cognitive Law
The higher the employee's capability, the greater the leverage for AI observation and learning.
The more covert the monitoring, the more strategic efficiency is prioritized over transparency.
The faster the counterpart development, the more thoroughly human capital is transformed into AI assets.