Meta to track employee actions to train AI, sparking privacy concerns
Meta to track employee actions to train AI, sparking privacy concerns
Meta plans to collect mouse movements, keystrokes, and screen snapshots from work apps to train AI, prompting questions about privacy and safeguards.
Meta is rolling out an internal tool called the Model Capability Initiative (MCI) that will run on selected work-related apps and websites employees use. It will also take occasional snapshots of screen content to help the company understand the context of what workers are doing, part of its broader AI for Work (AI4W) initiative. The program aims to feed data into its AI models and to develop AI agents capable of carrying out routines tasks autonomously. The initiative has been communicated to US employees via an internal memo from an AI research scientist at Meta's SuperIntelligence Labs.
Meta says the data collected will not be used to evaluate employee performance or inform workplace decisions. It also notes that safeguards are in place to protect sensitive information, though it does not clearly define what qualifies as sensitive content. The memo emphasizes that employee participation will help improve the models simply by performing daily work, but many workers remain concerned about the scope and potential for misuse.
Privacy advocates and some employees worry about a surveillance-like environment where routine actions—mouse movements, keystrokes, and occasional screen captures—could be logged and analyzed beyond any transparent purpose. Critics argue that even with assurances, the mere presence of such monitoring can alter how people work and stifle creativity. Proponents, however, say it could accelerate enterprise AI and automate repetitive tasks, potentially freeing staff for higher-value activities, provided robust controls and independent audits are in place.
This move underscores a growing trend in corporate AI where internal data is leveraged to train models and build automation tools. Companies are balancing the potential gains in efficiency with the need to protect privacy and maintain trust. If Meta’s safeguards prove effective and clear, the program could serve as a template for responsible AI training in the workplace; if not, it could intensify scrutiny over how internal data is collected and used.
