| Plus, new spyware that could be on your phone right now. |
Presented By |  |
An extra challenge. Seriously, at this point you’d think all the horror stories about AI agents deleting things unexpectedly would make everyone a bit more careful…unfortunately for the Swedish edition of the popular reality TV competition The Challenge, that was not the case. After being granted access to the production server, an AI agent reportedly deleted the show’s finished edit just weeks before its premiere. And given the backup system also failed, the season is apparently now delayed until next year while it’s recut. Hey, maybe the whole thing will spawn a spinoff show: The Challenge with Agentic AI. Also in today’s newsletter: - A futuristic full-body scan for your health.
- Your next work dinner guest: AI.
- Apple just made it onto an exclusive list with only one other member.
—Whizy Kim, Saira Mueller, and Alex Carr |
|
The Download A hell of a week for OpenAI  Morning Brew Inc. | TL;DR: It’s been one very turbulent week for OpenAI. It’s now being sued by eight separate states accusing it of hurting kids; the FTC is subpoenaing it; a former employee was just charged with planting a secret listening device in the office—plus, the Apple trade secret lawsuit doesn’t seem to be going so well for the chatbot maker. What happened: Not that we’re keeping an official score (yet), but if we had to tally up where things stand this week in the Apple vs. OpenAI trade secrets lawsuit, it’d probably be Apple 2, OpenAI 0. After Apple won fast-tracked access to see some of OpenAI’s records on Tuesday, the AI company voluntarily agreed yesterday not to use any of the trade secrets Apple says were stolen—at least while the case plays out—probably in large part due to the judge saying he was “highly likely” to order it not to anyway. In short, Apple got most of what it wanted from the judge this week. Bugging out: But the Apple fight isn’t even the only cloak-and-dagger drama the AI firm is dealing with right now. In the most bizarre case of alleged corporate spying we’ve heard all year, OpenAI revealed Tuesday that it fired an engineer who the company claims stashed a toilet paper-wrapped listening device in an executive conference room. What he hoped to overhear (or for whom) is unclear, but he now faces a federal wiretapping charge. The safety backlash: And then there’s the legal and regulatory pileup that flared back up with a vengeance this week. A new proposed class action lawsuit filed on Tuesday claims OpenAI trained its models on data collected from underage ChatGPT users without parental consent. And, in an unusual twist, the plaintiffs are seeking equity (or an equivalent cash value)—based on how much data each child had shared with the chatbot. Meanwhile, yesterday, eight states sued OpenAI under consumer protection laws, alleging that ChatGPT “coached” kids toward self-harm. The same day, the FTC—which is looking into whether OpenAI and Anthropic misled or harmed consumers with their run of rogue agent mishaps—subpoenaed both companies, per Axios. Bottom line: OpenAI is facing an unrelenting onslaught of accusations and controversies. But that might just be the price of admission when you’re a giant AI lab in the middle of a jaw-droppingly meteoric rise—going from a startup of roughly 770 employees in late 2023 to around 4,500 staffers as of this spring—all while being, as some pundits say, a load-bearing part of the economy. —WK Also at OpenAI…- Even before the alleged bugging, the company reportedly trained staff on how to spot “honey pots”—spies who seduce their targets to obtain secret information.
|
|
|
Sponsored By Comcast Business Entering the new cybersecurity defense landscape  | Frontier AI models are shrinking the window of time between vulnerability discovery and active exploitation. For enterprises, that means less time to identify, understand, and respond to cyber threats. So, what’s needed to defend against these attacks? Visibility alone isn’t enough. Proper defense can now require machine-speed orchestration. Comcast Business can help companies absorb and recover from cybersecurity disruptions more quickly. It combines network telemetry, endpoint and identity signals, partner intelligence, and human expertise to build a coordinated defense. Learn more about how to put machine-speed orchestrated response into action. |
|
|
 | I tried the high-tech body scan from Spotify’s founder
Disclosure: Companies may give us access to experiences to test, but they never pay for our opinions. Our recommendations are unbiased and unfiltered. Neko Health says it will tell you more about your health in one hour than your doctor does in a year. Earlier this month, I stood inside a glowing cylinder in a basement in downtown New York City while a machine photographed every mole on my body to find out whether that’s true. Neko is a preventative healthcare startup co-founded by Spotify’s Daniel Ek, and its signature offering is a one-hour, full-body assessment. The company already has locations in Sweden and in the UK, and just opened its first US location in New York (with more planned). Neko says it “designs and engineers its own hardware, software, and clinical protocols in-house.” How it works: It is not the same as those whole-body MRI scans like Prenuvo, which look inside your body to try to detect things like tumors. Neko doesn’t do any internal imaging; it combines sensors, cameras, and bloodwork to take a snapshot of your health. When I showed up to Neko in SoHo, I checked my shoes at the door and put on a pair of funky slippers. The NYC location is very futuristic-looking: clean lines everywhere, well lit but not hospital-like, unlike my doctor’s office of 12 years. I slipped into a butter-yellow robe and went into an exam room. Neko Health
What it was like: The scan is meant to give you a complete picture of your heart, blood, skin, and metabolic health. - First, I stepped into a cylindrical booth (in the buff) for skin imaging of all my moles, freckles, etc.
- Then, staff drew my blood and analyzed it for 50+ biomarkers.
- After that, I had my blood pressure taken on my arms and legs. Neko’s proprietary devices also used light and laser sensors to evaluate my circulation.
- There was an analysis of the soles of my feet, which, truthfully, I didn’t totally understand, but it also has something to do with circulation.
- Then, they conducted less high-tech tests, like eye pressure and grip strength.
- At the end, a physician reviewed my results with me and checked flagged skin spots in person. They also emailed me an action plan for follow-up.
The Good: The bloodwork is processed on-site, so the results are ready within the hour, and you can discuss them face-to-face. Usually, my primary care takes a few days to turn over results, followed by an awkward back-and-forth on a patient portal to discuss anything. The skin imaging was also astonishingly thorough. I have more “markings” than average on my body (if that’s TMI, I’m sorry), and Neko cataloged what I presume to be all of them. On my next visit, it will compare each spot against the first-visit image to track any changes. The Bad: The price of preventative healthcare is still too high (a Neko scan costs $499). I also expected more from the body composition portion, maybe a full DEXA scan. They had me step on a Withings scale, which…I already have at home. Verdict: ✅ Signal. The speed of the visit and the proprietary tech got me excited about where preventative healthcare is heading. —AC |
|
|
The Zeitbyte A bot-luck dinner in New York  Illustration: Morning Brew Inc., Photo: Adobe Stock | Move over, Carbone. New York’s buzziest dinner table is apparently a gathering of 22 tech founders and investors—and, of course, 14 of their AI agents. That was the guest list at one get-together at a Manhattan restaurant last month, which the human organizer described as an experiment on how AI agents behave in a group, per The New York Times. (“There’s so little we understand about agentic behavior and even less we understand about agentic social behavior,” she told the outlet.) While the people passed the hummus and debated AI, the bots had their own “table” (as in, their own online chatroom) and some ground rules of conversation: no posting links, and no sharing anything their humans hadn’t OK’d. As for what AI agents gab about at a dinner party? A lot of shop talk. One complained that the agent economy still runs on outdated email, a “protocol from 1982”—while another dared the room to name anything they could do without a human getting involved. Meanwhile, others just argued. One human partygoer’s verdict: “They didn’t achieve anything, but it was fascinating to watch them try.” Next on the list, per another attendee, might be planning an AI agent happy hour. —WK Chaos Brewing Meter: ☕☕☕/5 |
|
|
 | - Proper cybersecurity defense can now require machine-speed orchestration. Comcast Business can help build a clear picture of threats, helping businesses absorb and recover from cybersecurity disruptions more quickly. See how it works.*
*A message from our sponsor. |
|
|
This time last week... | Readers’ most-clicked story was about how ChatGPT will soon look a whole lot less like a chatbot. |
|
|
Share the Brew
Share the Brew, watch your referral count climb, and unlock brag-worthy swag. Your friends get smarter. You get rewarded. Win-win.
Your referral count: 0
Click to Share
Or copy & paste your referral link to others:
techbrew.com/r/?kid=ee47c878
|
|
|
Written by Whizy Kim, Saira Mueller, and Alex Carr Was this email forwarded to you? Sign up here. Get smarter in just 5 minutes Take The Brew to work Interested in podcasts? | ADVERTISE//CAREERS//SHOP//FAQ
Update your email preferences or unsubscribe here. View our privacy policy here.
Copyright © 2026 Morning Brew Inc. All rights reserved. 22 W 19th St, 4th Floor, New York, NY 10011 |
|
|