Meta Platforms Inc. recently announced the Model Capability Initiative (MCI), which records employee keystrokes, clicks, and mouse movements on work computers to train the company’s agent-based artificial intelligence (AI) models. These models automate tasks currently completed by human staff.
According to CNBCMCI tracks activity on specific sites such as Google, LinkedIn, Wikipedia, GitHub, Slack, and Threads. CNBC noted that this list is subject to change. In addition to tracking keystrokes, clicks, and mouse movements, MCI Software may also take screenshots to provide context for your interactions.
According to ReutersAccording to an internal memo, Meta’s AI agents currently fall short in areas such as “selecting from drop-down menus and using keyboard shortcuts.” Meta is moving towards having AI perform routine tasks on computers to increase efficiency, and aims to further develop these capabilities.
“Efforts like this are part of a broader push across technology companies to improve AI systems using real-world workflows and internal data. The perceived benefits are enhanced model performance and better integration into everyday work tools,” said Pooja Nihalani, former manager at Meta.
According to Reuters, members of the MetaSuperintelligence Laboratory (MSL) have promised that MCI data will not be used for performance evaluations and that sensitive data will continue to be protected.
However, many people are still concerned about the impact this initiative will have on stability, privacy, and productivity. Nihalani said that when he spoke with current Mehta employees, he heard more negative comments than positive ones.
“Human employees are essentially training AI machines to take over their jobs, which may benefit the company’s efficiency in the future, but it could potentially jeopardize the human’s job security,” said Charlene Hsu, a third-year student at Carlmont University taking Advanced Placement (AP) Computer Science A.
Alexandra Matescu is a researcher at the U.S. Labor Futures Initiative. Data Society Research Institute. She researches issues at the intersection of worker rights and technology, such as workplace surveillance.
“It’s not just the fear of being displaced by automation, but also the fact that in the United States, workers have very few data rights, and employers have some kind of total control over the data they collect from their workers,” Mateescu said.
According to Mateescu, this imbalance can have negative consequences for employees, including data sharing and salary cuts. In extreme cases, the employee may be fired.
“Workers are rarely compensated for the value created from their data, and even if companies are able to develop valuable AI products based on those workers’ data, it is unlikely to lead to an increase in that worker’s pay or even lead to their being fired or replaced,” Mateescu said.
According to forbesjust as the MCI is being implemented, Meta plans to lay off 8,000 employees, or 10% of its workforce, soon to “offset other investments,” including some AI projects.
Internal messages seen by CNBC indicated that employees were dissatisfied with MCI for another reason: privacy.
“Under the National Labor Relations Act, workers have the right to form a union. The ability to do that is a form of violation for workers who fear that their employers are collecting data about their communications,” Mateescu said.
Another effect that people believe MCI can have is a decrease in employee productivity.
“It can cause anxiety and put pressure on you to appear more productive than you actually are,” says Schuh.
Nihalani said Mehta’s culture of high performance and responsibility can increase self-awareness.
“In a highly performance-driven environment, we see a tendency to prioritize projects that demonstrate impact within a shorter time frame over work that is likely to have a longer-term, more meaningful impact. When combined with a sense of continuous evaluation and monitoring, this dynamic can further reinforce a bias toward tangible, short-term outcomes rather than deeper, more lasting outcomes,” Nihalani said.
It is important to focus on MCI in the current state of the technology industry. Meta is not alone in increasing its monitoring and integrating AI into its workflows.
“Collecting large amounts of worker data has been going on for some time, and even the part about training AI systems is not necessarily unique. Even before generative AI, there were efforts to use customer service data to train rule-based chatbots,” Matescu said.
According to Mateescu, federal law is weak and very low. Joining a labor union Some states allow companies like Meta to collect large amounts of data about their employees, which would likely be illegal under the European Union. General Data Protection Regulation Otherwise, they will face a backlash from organized labor..
Furthermore, such data collection is not limited to office jobs. The trends we’re seeing now at companies like Meta were preceded by similar policies for blue-collar employees.
“There’s a kind of trickle-down, and a lot of the most insidious forms of surveillance often start in low-wage workplaces where workers don’t have a lot of influence or protection. Now, white-collar workers are starting to see a lot of the techniques that were first used inside Amazon warehouses and within the gig economy,” Mattescu said.
