finance
Employers Use Monitoring Data to Train AI Agents in Workplace
Companies are increasingly monitoring employee activities on work devices to gather data for training AI agents. Meta has implemented a tool to track actions like keystrokes to improve AI systems, while other firms monitor attendance and AI usage. Experts note this represents an evolution in workplace surveillance practices.
Apr 23, 5:57 AM·2m read1 source
Vassia Atanassova - Spiritia / Wikimedia (CC BY-SA 3.0)Audio version
Tap play to generate a narrated version.
Developing·Limited corroboration so far. This page will refresh as more sources emerge.
More companies are monitoring employee activities on work devices, including laptops and phones, to ensure productivity and compliance. This monitoring now extends to using the data to train AI agents that can automate tasks. The US Government Accountability Office reported in September that employer surveillance has increased due to remote work and available tools.
Employers such as AT&T use technology to track office attendance. JPMorgan monitors software engineers' AI usage through dashboards, according to a Business Insider report from Thursday. These practices aim to collect data on how employees perform tasks.
invest in AI agents to automate tasks and make decisions, requiring high-quality data from real work scenarios. Worker data, including emails and messages, provides insights into decision-making processes. This data is specific to company operations and valuable for building effective AI agents.
Meta deployed an internal tool that tracks employee activities, such as keystrokes and mouse movements, to train AI systems. The company stated in an internal memo that the goal is to understand task completion for AI replication or assistance. A Meta spokesperson said safeguards protect sensitive content and the data is used only for this purpose.
on worker data is more straightforward for roles like software development, where work involves defined steps on computers. For other roles, the data may not fully capture job complexities. Companies often store more data than they use due to costs and security risks.
Some Meta employees expressed concerns about the extent of monitoring. Dan Schawbel, managing partner at Workplace Intelligence, described this as an evolution from measuring work to learning how to replace it. He noted that companies capture institutional knowledge in real time.
