Meta will use employees' mouse and keyboard data to train its AI
Meta is starting to record how employees work on their computers in the U.S. to improve its AI models. This move raises concerns about privacy and job security.

Meta is rolling out a new tool called Model Capability Initiative (MCI) that will monitor how employees use their computers. The goal is to collect mouse movements, clicks, and keystrokes to train their AI models.
Internal documents reveal that the aim is for the models to perform tasks they currently struggle with, like navigating complex interfaces or selecting options from dropdown menus. Employees will continue their daily work, unknowingly helping train the AI to become more accurate and efficient.
This initiative is part of a broader shift at Meta to compete with AI giants like OpenAI, Google, and Anthropic. The company has renamed this effort as the Agent Transformation Accelerator, emphasizing that humans will mainly supervise AI systems that handle most tasks.
In an internal memo, Andrew Bosworth explained that employees will mainly review and guide AI agents, which will do most of the work. Meta also confirmed that data collected through MCI will be used solely for training models, although details about what information is excluded remain unclear, raising privacy concerns.
Meta is also pushing its staff to use AI agents for programming and other tasks, even if it slows down their work initially. Ultimately, the company seems focused on replacing many workers with digital agents trained on their own data, aiming to cut costs and streamline operations.
Keep reading online — scan the code
https://go.tricuatro.com/Tz3f8
© tricuatro.com
Article topics
Related articles

TikTok tests AI tool to protect creators from deepfakes
The video platform is rolling out an opt-in feature allowing users to verify their identity and report unauthorized AI-generated likenesses.

Apple Sues OpenAI Over Alleged Hardware Trade Secret Theft
The Cupertino giant accuses former key employees, including Tang Tan, of stealing confidential hardware information for OpenAI's benefit and its expansion into consumer hardware.

Anthropic Tackles AI's Toughest Questions for a Responsible Future
The company launches an initiative to understand public concerns and transparently address the challenges of artificial intelligence development.
Latest news
View all
Sweet Interstellar Discovery: Erythrulose Detected in the Milky Way
Scientists from the Spanish Astrobiology Center have for the first time identified a true sugar, erythrulose, in a gas and dust cloud near our galaxy's center. This finding suggests complex molecules can form before stars and planets are born.

Paramount and Warner Bros. Merger Halted by Judge Over Antitrust Concerns
A coalition of 12 states successfully petitioned Judge Araceli Martinez-Olguin to temporarily block the deal, citing fears of higher prices and reduced content.

MIT Chip Converts Invisible Infrared Light into Crucial Data
MIT researchers developed an innovative chip detecting methane and propane using mid-infrared light, with potential for millions of pixels and optical computing.
Comments (0)
No comments yet. Be the first!
Only registered readers can comment.