You open your laptop and a memo has already beaten you to the punch. I felt that small, cold drop when I first read the notice: Meta will install software that records every mouse movement and keystroke. The company says it’s to teach its AI agents how to do your job.
I’m going to walk you through what that memo actually means, what Meta says it will do, and what you should be watching for if you work there or anywhere else that borrows the idea.
Monday: A short memo landed in inboxes — then silence
The memo, reviewed by Reuters, named a tool: Model Capability Initiative. Meta plans to record screens, mouse movements and keystrokes to teach its agents how to complete tasks across multiple apps. You’re asked to keep doing your job while the software watches.
Meta rebranded an internal effort — now called Agent Transformation Accelerator — and folded this tracking into its AI for Work push. On paper the promise is benign: training data, not performance metrics. I don’t find that comforting when the company is reportedly preparing a 10% workforce reduction, a move covered by USA Today.
Tuesday: Employees told their daily work will train models
Teams heard that helping models “get better simply by doing their daily work” won’t affect performance reviews. Reuters reported the exact phrasing. You should treat that assurance like a pledge from management during an acquisition: comforting until incentives shift.
The surveillance software is a microscope on daily habits. Meta’s agents still fumble simple UI tasks — dropdowns, keyboard shortcuts — and the company wants real human traces to fix those gaps. The goal is agents that can act across systems, and that requires granular human data.
Will employers use employee data to train AI?
Yes — and Meta’s move gives the clearest evidence yet. Other firms, from Microsoft to Google, collect usage data to refine tools, but Meta is formalizing continuous recording of individual workflows. That changes the moral economy of work: training your replacement becomes a line item in corporate AI strategy.
Wednesday: The power dynamics in the room shifted
Managers can promise the data won’t be used for reviews, but promises are fungible when budgets and headcounts tighten. I’ve seen technologies pitched as productivity aids become the very lever for downsizing. You end up teaching a system that may out-perform you on a spreadsheet of roles.
Meta’s own internal chatbots — including one fashioned around Mark Zuckerberg for employee use — are already learning social behaviors. The new collection program accelerates that learning and feeds Meta’s broader AI stack, which competes with OpenAI and in-house agents at Google and Microsoft for enterprise footholds.
Can companies legally monitor keystrokes and screens?
Legality varies by jurisdiction. In many U.S. states employers can monitor workplace devices, but notice and contract terms matter. In the EU, privacy rules are stricter and could require clearer consent or data minimization. If you’re an employee, check company policies and local labor protections, and consider who retains the raw recordings.
Thursday: Practical risks, and what you can do next
People will have different reactions: some will cooperate; others will resist. You get no extra pay for training models and you lose privacy while adding uncompensated labor to your day. That tension is real and practical.
Start with the paperwork. Read the memo, the benefits plan, and any updates to acceptable-use policies. Ask questions about retention, access, and deletion. If you’re represented by a union, raise it there. If not, talk to colleagues — silence benefits the company more than it helps you.
Meta says the data will improve agents and not be used for performance evaluations. Reuters and USA Today have reported the announcement and the context; you can read their pieces for primary sourcing. I’ll be tracking whether other big tech firms follow Meta’s lead or pull back under regulatory pressure.
For now, you’re being asked to teach the systems that might replace you, and management’s reassurances feel thin when layoffs are on the table. The question isn’t just technical: it’s ethical and personal, and the answers will shape how we work — and who benefits from that work — over the next decade. Are you ready to hand over the recipe for your job?