You are probably being watched at work. Not by a person with a clipboard, but by software that logs your keystrokes, samples your screen, reads your calendar, and turns all of it into a score. In 2026 that software increasingly runs on AI, and here is the part that matters: it is not measuring your work. It is measuring your activity. Those are not the same thing, and the gap between them is where the whole fight lives.
Most workers have no idea how much is collected, how little they are owed in the way of notice, and how thin the evidence is that any of it improves anything. This is what the tools actually do, what the law actually says, and what you can actually do about it.
What They Are Actually Watching
Start with the stack, because “monitoring” undersells it. A modern employee monitoring suite can capture, in varying combinations:
- Activity and keystrokes. Typing cadence, mouse movement, and how many seconds each minute you were “active.”
- Screen capture. Many tools take a screenshot on an interval, often every few minutes, sometimes keyed to activity.
- App and URL logs. Which apps are open, which websites are visited, and for how long.
- Email and chat scanning. Keyword and sentiment analysis over internal communications.
- Location. GPS on company phones and badges, especially for field and delivery work.
- Biometrics. Webcam gaze and head-pose tracking, and in some cases face analysis.
- Emotion AI. Attempts to infer stress, engagement, or mood from voice, face, or typing patterns.
Then a scoring layer rolls all of it into a number. Adoption is not fringe. An ExpressVPN survey of 1,500 United States employers found 74 percent use online tracking tools, 75 percent monitor physical workplaces, and 61 percent use AI-powered analytics. A 2025 Gartner survey put monitoring of remote workers at over 70 percent of large employers, up from roughly 30 percent in 2019. The global market for employee monitoring technology is expected to pass four billion dollars in 2026, and that growth is not being driven by companies that already trust their people.
Myth 1: It Is Mostly About Security
Security is the polite word for it, and it is sometimes true. But the product being sold is not a firewall. It is productivity scoring, and productivity scores now feed employment decisions.
The clearest example is a lawsuit filed in July 2026 by 26 current and former Meta employees. They asked a federal court to halt layoffs, alleging that the company relied on AI-assisted systems that generated productivity scores from keystroke data, screen activity, emails, browsing history, and workplace communications. The complaint argues those scores disadvantaged workers on medical leave, workers who requested accommodations, and employees caring for family members, people whose measured activity naturally dips for reasons that have nothing to do with how well they do their jobs. The suit also alleges Meta failed to audit the systems for bias as required under newer California and New York City AI employment rules. Meta denies that AI made the layoff decisions, and none of the claims have been proven in court.
Set aside the specific case and look at the mechanism. Any tool that scores visible activity will mark down anyone whose best work does not look busy.
Myth 2: They Have to Tell You
In most of the United States, they do not. Federal law sets a low baseline. The Electronic Communications Privacy Act, passed in 1986, includes a business-use exception that lets employers monitor communications on company equipment, and there is no federal requirement to disclose it.
State law is where the rules actually bite, and they are a patchwork. Four states require written notice before monitoring electronic communications: Connecticut, Delaware, New York, and Colorado, where the notice duty covers AI-driven monitoring tools. A separate set of states has biometric privacy laws that reach employers: Illinois, Texas, and Washington. If you work anywhere else in the country, secret monitoring on company equipment is generally legal.
Here is the practical translation. If you have never received a written monitoring notice, and you work in one of those notice states, that is worth a question to HR. If you work elsewhere, you should assume the device is instrumented and act accordingly.
Myth 3: It Measures My Work
It measures activity. Keystrokes, clicks, active-window seconds, screenshots. None of that is output. A support agent who resolves a hard ticket in four minutes and then thinks for six minutes looks less productive than one who scrolls for ten minutes while touching the keyboard. A developer reading a spec carefully looks idle. A designer staring at a layout looks idle. Deep work looks like nothing on an activity monitor, because the monitor cannot see inside the work.
Once a metric becomes the target, people optimize the metric. Wells Fargo fired more than a dozen employees in 2024 after an investigation found they had used tools to simulate keyboard activity, making idle machines look busy. You can read that as a story about cheating. It is more useful to read it as a story about a measurement that stopped measuring the thing it claimed to.
Myth 4: It Makes People More Productive
The evidence for the productivity claim is weak, and the evidence for harm is not. One 2026 survey found that 75 percent of employees whose workplaces use surveillance software said the monitoring increased stress rather than productivity. Another comparison put 68 percent of managers saying monitoring improves work against 72 percent of employees saying it has no positive impact. Monitored employees also report more stress and tension than unmonitored colleagues, by a wide margin.
That is not an argument against all oversight. It is an argument that activity surveillance is a measurement that fails on its own terms: it raises anxiety, invites gaming, and leaves you with a number that correlates poorly with whether the work got done.
Myth 5: The Law Protects Me
In the United States, protections are thin and uneven, and they vary by state as described above. Two other levers are worth knowing. The National Labor Relations Act still protects concerted activity, which is why unionized and non-union workplaces alike have pushed back when monitoring chills organizing. And New York City’s Local Law 144, the first law to force audits of automated employment decision tools, covers hiring systems, not the monitoring that happens after you are hired.
The European Union is a different story, at least on paper. The EU AI Act bans emotion recognition in the workplace outright, treating it as an unacceptable risk. It classifies AI used in employment and worker management as high risk, which triggers human oversight and transparency duties, with the substantive obligations landing in August 2026. Penalties for high-risk breaches run up to 15 million euros or 3 percent of global turnover, whichever is higher.
One caveat you should hold onto: a proposed “Digital AI Omnibus” would delay the standalone high-risk deadline for employment tools from August 2026 to December 2027, and it is still moving through the process. If you read a confident claim that EU workplace AI rules either “start now” or “start in 2027,” check the date on the article. This one is genuinely in flux.
What You Can Actually Do
You cannot opt out of a device that is instrumented before you open it. You can shrink the surface.
- Assume the work device is fully monitored. Treat every tab, message, and keystroke on it as potentially visible. Do personal banking and private messages on your own hardware.
- Keep personal accounts off work machines. Credential capture from monitoring tools is a real failure mode, not a hypothetical one.
- Find out what is collected. Ask HR in writing what tools run, what data they capture, and how long it is retained. In Connecticut, Delaware, New York, and Colorado you are owed notice; in the EU you have transparency and oversight rights.
- Do not fake activity. Mouse jigglers and activity simulators are detectable and end careers. The metric is not worth gaming.
- Push for outcome-based measurement. If your team is judged on activity, propose goals that describe results. Managers respond better to “here is how we measure whether the work happened” than to “stop watching us.”
- Use collective channels if you have them. Works councils and unions have formal rights in Europe, and concerted activity is protected in the United States.
None of that defeats a black box you cannot audit. It reduces your exposure and puts the burden back where it belongs, on the employer to justify what it collects.
The Short Version
- Employee monitoring is now the norm, and a growing share of it is AI-powered scoring.
- It measures activity, not work, and the two diverge most for the people doing the hardest thinking or dealing with real life.
- Federal US law does not require notice; only Connecticut, Delaware, New York, and Colorado do, plus biometric rules in Illinois, Texas, and Washington.
- The EU AI Act bans workplace emotion recognition and treats employment AI as high risk, but the employment deadline is being renegotiated.
- The evidence that monitoring improves output is weak; the evidence it raises stress is not.
FAQ
Can my employer read my personal messages on a work computer?
Often yes. Many monitoring tools capture keystrokes and screen content including personal tabs, and anything typed on the device can be retained. Keep personal communication off work hardware.
Is AI monitoring legal in the United States?
Usually, on company devices, under the federal business-use exception. A handful of states add notice or biometric requirements. Check your state before assuming you are protected.
Does monitoring include my webcam?
Some tools use the webcam for gaze and presence checks, and some use face analysis. Where that crosses into emotion recognition, the EU AI Act bans it at work; in the US it is largely governed by state biometric law.
Can I be fired for faking activity?
Yes. Wells Fargo fired more than a dozen employees in 2024 over simulated keyboard activity. It also hands your employer a clean justification, which is the last thing you want.
The Bigger Picture
Surveillance at work is not new. The factory floor, the call center headset, and the time clock all came first. What changed is the scale, the inference, and the feeling. Software can now watch everything, guess at how you feel, and hand a manager a score that looks objective. It rarely is. It is a model, trained on somebody’s idea of what a good worker looks like, applied to people who mostly never consented and rarely get told.
That is why this belongs in the same conversation as algorithmic hiring. The screen at the front door and the scoreboard inside the building are built by the same industry, sold with the same promises, and fail in the same way: by mistaking a measurable proxy for the thing that actually matters.
Already filtered out before you started?
If a model decided your application never reached a human, that is a different problem with its own playbook. Here is how AI hiring screens actually work, and how to format a resume so the software reads it correctly.
If you want the formatting counterpart to that guide, we keep a copyable ATS-safe resume template here: how to format a resume for AI screeners.
About The Author
