TL;DR: Emotion AI—a category of tools that claim to read people’s feelings by measuring their tone and expressions—is slowly creeping into workplaces, according to a new report from the Atlantic. Your boss might not have to inform you they're using it, which means you could be secretly judged by an AI for your concentrating-too-hard face.

What happened: The list of emotion AI tools on the market is rapidly growing: There’s an app called MorphCast that analyzes meeting participants’ emotions in real time—Zoom extension included. Slack now has an integration called Aware that monitors employee chats for “sentiment and toxicity.” One AI startup says its tools can analyze the emotional states of job interview candidates. The global market for emotion AI could triple to $9 billion by 2030, according to the Atlantic.

How we got here: Emotion monitoring on the clock isn’t new (see: the entire service industry). For years, MetLife has been using an AI that listens to call center agents’ voices for energy, pace, and tone. (If the worker sounds tired, a coffee cup pops up on screen reminding them to perk up.) Earlier this year, Burger King started testing an OpenAI-powered headset called “Patty” that similarly rates drive-thru workers’ friendliness. The difference now is that more white-collar jobs are facing down such emotional surveillance.

A flawed tool: Given how famously complex human emotions are, it’s unclear how reliable AI is at reading feelings. As one expert told the Atlantic, research shows that Americans scowl when they’re angry only about one-third of the time. And even if someone scowls, it might be because they’re concentrating. Last year, the EU banned workplace emotion AI except for medical or safety reasons; the US has no federal equivalent, and only a handful of states require employers to disclose electronic monitoring of workers.

Bottom line: AI workplace surveillance is expanding before there’s strong evidence the tools actually improve work. There are few rules in place that even inform employees of how they’re being observed. And emotion AI in particular adds a whole new line item to your job description: convincing a bot you’re always cheerful.

What this means for your job: Unless you work for Meta, which is nice enough to announce it’s tracking employees, you probably won’t know exactly how you’re being watched—but if you’re on an employer-issued device, you almost certainly are. Check your employee handbook, offer letter, and IT policies. —WK

Joanna Stern’s “Great Gen AI Experiment”

If you're experimenting with AI—or want to but aren't sure where to start—award-winning tech journalist Joanna Stern spent a year doing it so you don't have to. Her new book, I Am Not A Robot: My Year Using AI To Do (Almost) Everything, releases on May 12, and it's as useful as it is funny: Stern tries everything from AI health coaches to robot massages and an intimate relationship with a chatbot, and reports back honestly on what actually works.

The following is a (slightly condensed) excerpt from the book, written as a mock research paper. —SM

THE GREAT GEN AI EXPERIMENT
PART 1: SEARCH AND INFORMATION

RESEARCH QUESTIONS: What happens when you let artificial intelligence curate 100 percent of your information intake for a year? Will you become smarter or more ignorant? And who’s really in control when AI controls what you know?

METHODOLOGY: Hi, I’m the subject. It’s me. I handed over all my web searching and information discovery tasks to AI for the year. No Google or any other search engine. Only ChatGPT, Perplexity, Google’s Gemini, Claude, and other AI tools could answer my questions or suggest what I should learn about. I even changed the default search engines in my web browsers to those AI tools.

DATA COLLECTION: At first, it felt strange to stop googling things—a bit like walking down a paused escalator. But it quickly became second nature to just get “answers” instead of a list of blue links. I replaced Google altogether for questions like these:

• RECIPES AND COOKING. “Provide a simple meatball recipe, including my secret ingredient: ketchup.”
• HOME ADVICE. “What temperature should the kids’ rooms be at?” (Answer: 70°F, and “whatever feels cozy.” Not 74°F just because it’s cold outside. We own sweatshirts. We own socks.)
• RANDOM LIFE ADMIN. “From “How often should tires be rotated?” to “How many ounces in a cup?” (Answers: Every five to seven thousand miles, and eight ounces, respectively.)
• WORK RESEARCH. “Find me a list of companies working on AI and radiology.”

CONCLUSION: I went back to Google only for maps and finding a business’ contact information. But when it came to general knowledge, AI became my default.

The real magic of AI search was the multimodal part—combining audio, images, video, and text in one query. I lost track of how many times I pointed my phone’s camera at something in the house and said out loud, “How do I fix this?”

But that doesn’t mean AI is a magical oracle. One time, I aimed ChatGPT at my garage door, which was struggling to open. Using the video feed, it confidently diagnosed a faulty sensor. Spoiler: That wasn’t the problem. There was no fault sensor. It was a broken spring, which I discovered only after calling an actual garage expert, one with a ladder and opposable thumbs.

That’s the main downside: the hallucinations. As we learned in the Non-Boring AI Glossary, large language models have a bad habit of confidently making things up, including citations, dates, entire events that never happened. They don’t actually know the truth; they just predict what words are likely to come next, which sometimes means inventing details out of thin air.

AI never made something up so egregiously that I went and did something truly unhinged or believed something completely bonkers. But then again, I was often checking its work.

FUTURE AREAS OF RESEARCH: Unlike the forthcoming Great Gen AI Experiments (one for every season, like a deranged Hallmark movie series), this one stuck.

My attempts to replace creative works—music, books, and so on—with generative AI didn’t exactly endure, but swapping out search for chatbots? That’s a habit I’ll keep well beyond this year.

From the book: I AM NOT A ROBOT by Joanna Stern. Copyright © 2026 by Joanna Stern. Reprinted courtesy of Harper, an imprint of HarperCollins Publishers.

If the Musk vs. Altman trial has shown us anything, it’s that Sam Altman and Co. love to text one another. The lone exception: Microsoft CEO Satya Nadella, who appears very firmly in the camp of phone calls over thumb-tapping. A leading theory for the discrepancy is that a generational divide is playing out—Nadella is 58, while the younger cast of characters whose texts everyone’s now reading (popcorn in hand) mostly range from their late 30s to early 40s. (Counterpoint: 54-year-old Elon Musk, who deliberately tweets out his every thought for the public to see.)

Nadella’s texting reluctance hasn’t gone unnoticed on social media, where he’s been praised for making a “pro CEO move” and being “the (only?) adult in the room.” After all, the first rule of dealing with a sensitive matter—like, say, the dramatic firing-then-rehiring of an AI firm’s CEO—is to not leave a paper trail of your internal freakout and thoughts on your replacement. It’s a lesson many tech execs have yet to learn, apparently, given the flood of embarrassing missives that have gone public over the years. (You can read a whole library of them here.) If we were lawyers, we’d advise everyone working in Silicon Valley to go Nadella mode—but then we’d be robbed of some very entertaining screenshots and memes for our own group chats. —WK

Chaos Brewing Meter: /5