Meta logs staff keystrokes on Google, LinkedIn and Wikipedia for AI training

Meta, the social‑media conglomerate that has repeatedly positioned itself as a champion of digital innovation, has quietly instituted a program that records every keystroke and mouse click made by its employees while they browse widely used platforms such as Google, LinkedIn and Wikipedia, ostensibly to feed a proprietary artificial‑intelligence training pipeline. The internal directive, reportedly circulated among engineering and product teams in early April, specifies that data capturing will occur regardless of whether the accessed content is work‑related or personal, thereby blurring the line between corporate oversight and invasive surveillance in a manner that would make privacy advocates uneasy.

According to the rollout plan, the harvested keystrokes and click metadata will be aggregated into a centralized repository, anonymized only through superficial tokenisation, and then fed into large‑scale language models that Meta claims will improve user‑facing features such as content recommendation and automated assistance, while the company simultaneously assures staff that the initiative does not constitute a breach of confidentiality. Nonetheless, the procedural safeguards outlined in internal memos remain vague, offering no clear timeframe for data deletion, no independent audit mechanism, and no transparent reporting to employees who are ostensibly the very subjects of the monitoring.

When placed against Meta’s broader history of deploying invasive data‑collection practices under the banner of product improvement, this latest venture appears less an isolated experiment and more a predictable manifestation of a corporate culture that routinely prioritises algorithmic advancement over the fundamental expectation of workplace privacy. The episode thus reinforces the paradoxical reality in which a company that markets itself as a steward of digital rights simultaneously engineers internal mechanisms that erode those very rights, leaving regulators and employees to question whether the promised AI benefits will ever justify the cost to personal autonomy.

Published: April 23, 2026