Hello and welcome to Regulator, a newsletter for Verge subscribers covering the broligarchs, the influencers, and the (potentially conscious) artificial intelligence models scrambling for power in Washington. If you’re not a subscriber yet, assert your humanity against the will of the machines by signing up here.
Technology
Meet the Pentagon’s AI bro squad
Very important news: Do you want to tell me stuff and see it printed in Regulator? Well, now you can, because we have a new tip line! Send all commentary, cool information, and ~secrets~ to tina.nguyen+tips@theverge.com.
The Pentagon’s private-sector A-Team
This morning, in advance of a meeting between Defense Secretary Pete Hegseth and Anthropic CEO Dario Amodei, my colleague Hayden Field and I published a story about the Pentagon’s hardball contract renegotiations with Anthropic. The stakes are higher than it should reasonably be, with the Pentagon continuing to designate Anthropic a “supply-chain risk” if the company doesn’t comply with their demands about their acceptable use policy.
In a post-meeting readout, Axios reported that Hegseth brought several other senior Defense officials to the meeting in an attempt to show that the Pentagon was taking the dispute “seriously.” But in a post-DOGE Trump administration run by broligarchs, it’s always worthwhile to check the attendees’ bios. Some of them were normal senior officials who’d spent their careers in government and military work, but the others have somewhat unusual backgrounds:
- Pentagon CTO Emil Michael, who we reported has been spearheading negotiations with Anthropic. Michael may be familiar to longtime Verge readers and followers of Silicon Valley corporate drama as the former second-in-command at Uber when Travis Kalanick was CEO. Michael was pushed out in 2017 after an investigation found that he, and several other top executives that called themselves the “A-Team,” perpetuated a culture of sexual harassment at the company.
- For anyone curious about his history on surveillance: During a 2014 dinner with several journalists, Michael suggested that Uber hire opposition researchers to gather personal “dirt” on reporters publishing unfavorable news, suggesting that he’d wanted to target one female reporter who had recently criticized the company for its culture of misogyny. This was also around the time that Uber drew controversy for an internal tool known as “God Mode,” which employees used to track the movements of its users, including one BuzzFeed journalist who was writing about an Uber executive.
- Deputy Secretary Steve Feinberg, the founder of the private equity firm Cerberus Capital Management, which manages roughly $65 billion in assets and specializes in “distressed properties.” Feinberg, who’s widely blamed for the death of the auto manufacturer Chrysler, was also an early supporter of Donald Trump, donating to his 2016 presidential campaign and serving on the president’s intelligence advisory board in 2018. During his 2025 Senate confirmation hearing, Feinberg touted Cerberus’ investments in several companies involved in national security, saying he had “significant experience with the Pentagon as a contractor and understand[s] how it functions and is organized.”
- At the time, Democrats raised concerns that Feinberg would have conflicts of interest due to Cerberus’ numerous investments in defense companies such as DynCorp. (That year, DynCorp settled a lawsuit with the Department of Justice over allegations that it had “knowingly inflated subcontractor charges under a State Department contract to train Iraqi police forces.”)
- In 2023, while Feinberg was still at Cerberus, the firm launched Cerberus Ventures, a venture capital arm that invests in early-stage companies that address national security issues in critical infrastructure.
- Hegseth’s chief spokesperson, Sean Parnell, an Army veteran who, in 2021, attempted to run for an open Senate seat in Pennsylvania. While he won Trump’s endorsement in the heated Republican primary, he was forced to drop out in November after his ex-wife made several allegations of serious physical and psychological abuse during a custody hearing. She was afforded full legal custody. (Dr. Mehmet Oz, now serving in the Trump administration, subsequently won the nomination.)
Feinberg and Michael’s presence should draw eyeballs. Yes, they both have some amount of defense industry experience: Michael was a White House fellow during the Obama administration, and spent two years as a special assistant to Defense Secretary Robert Gates at the Pentagon, which isn’t nothing. Feinberg has clearly spent time with defense contracts. But one must fully appreciate the rapacious business mindset that private sector types love to bring into the government — especially with high-stakes negotiations such as this. Parnell’s presence, meanwhile, makes sense within the context of “being the spokesman for Pete Hegseth.”
The single-supplier shuffle
One topic Hayden and I didn’t get to explore more was the “single-supplier vulnerability” issue, but it’s turning into a crucial factor in negotiations.
In 2024, the Biden administration released a national security memorandum on the use of artificial intelligence, which laid out several directives regarding the protection of the supply chain. Among them was a directive for the Department of Defense to maintain contracts with at least two frontier AI labs that were cleared to handle classified information, in order to prevent a scenario where one compromised vendor could take down an entire IT system. But as early as the summer of 2025, I’m told, the Trump administration was trying to address that vulnerability. While they had signed separate contracts with Anthropic, Google, xAI, and OpenAI, only Anthropic’s model was cleared for classified use when Hegseth published his memo outlining his new AI policy in January.
This has placed the Pentagon in a tight situation: Even if they successfully cut out Anthropic and go through the arduous process of making every defense contractor remove Claude from their workflows, they would risk being out of compliance with the Department’s own guidelines, to say nothing of common sense. (Avoiding single-supplier vulnerability is a very basic practice in the tech industry.)
It certainly provides more context to the Pentagon’s decision last night to suddenly grant xAI’s Grok access to classified systems, even though Grok is widely considered the least capable of the available models. While The New York Times reported that Google is also close to signing a deal allowing the Pentagon to use Gemini for classified work, defense insiders view Gemini as a quality rival to Claude, while xAi’s Grok “is not considered as advanced or as reliable as Anthropic’s.” OpenAI is not close to a deal, as the company reportedly believes that it must improve ChatGPT’s safety features before deploying it on classified networks.
So let’s do the math. You have four AI models, and you’re required to work with two of them. Your choices are:
1) A company with a pretty good AI model and increasingly flexible morals
2) A company with the best AI model, but which refuses to let you use it for autonomously killing people without human input
3) A company whose AI model isn’t secure enough to deploy yet
4) A company whose AI has racist hallucinations and generates child porn, and that you don’t consider “advanced [or] reliable”
If you can’t contract with companies 2 and 3, you’re stuck with companies 1 and 4, which even Defense officials admit is not optimal from a national security perspective. “The only reason we’re still talking to these people [Anthropic] is we need them and we need them now. The problem for these guys is they are that good,” a Defense official told Axios ahead of the meeting.
The latest Clarity Act negotiations between finance and crypto last week inadvertently turned into the latest episode of recurring segment I’m now calling: “Why is Laura Loomer tweeting about obscure deep-cut tech issues as if they are MAGA loyalty tests?”
Last Thursday, a small group of powerful crypto and finance players met at the White House to continue hashing out draft language over stablecoin yields. Coinbase, which sparked these negotiations after it withdrew support from Clarity over stablecoin yields, was in attendance. Prior to the meeting, however, Loomer tweeted a classic banger that demonstrated the tactics she uses to wield influence over Trump: Cast the target as someone who once supported Trump’s enemies and is therefore disloyal.
Ironically, Coinbase has turned into one of the biggest branded boosters of the Trump administration, donating money to his pet initiatives and even having their logo splashed all over last year’s military parade.
Though Loomer tweeted a similar sentiment about Coinbase last June, it seems to have had no impact on whether Coinbase has access to Trump, and likely won’t for a while: I’m told that CEO Brian Armstrong was at Mar-a-Lago the day before Loomer tweeted, attending a World Liberty Financial event.
A wild Trumpworld character has appeared!
If you followed the saga of Logan Paul auctioning off his Pokémon card collection, you may be aware that one of those cards sold for a record-setting $16.5 million last week. But who’s that Pokémon purchaser? It’s AJ Scaramucci, the son of the one and only Anthony Scaramucci, the New York financier and former Trump ally who famously served as Trump’s White House Communications Director in 2017 for 10 days.
AJ is the founder of Solari Capital, which invested $100 million in a Bitcoin mining platform run by Eric Trump. He also now owns the Pikachu Illustrator card, one of only 39 cards in existence and in Grade 10 condition, as well as the diamond chain and carrying case that Paul wore to display the card when he appeared at WrestleMania 38. Scaramucci told reporters that he purchased the card as part of his upcoming “planetary treasure hunt,” adding that he also hoped to purchase a T. rex skull and the Declaration of Independence. (He later posted on X that he hoped to place the card in the Nintendo Museum in Kyoto and cement it as “the ‘Mona Lisa’ of the Pokemon franchise.”)
We can’t believe that a court has to tell you this, much less the Southern District of New York: If you put correspondence between you and your lawyer into a publicly available AI platform, it is no longer protected by attorney-client privilege and becomes subject to discovery!!!!
In any case, have a pleasant State of the Union watch party (if anyone does that anymore) and see you next week.
Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.
Valve has some good news and bad news about Steam Controllers. The good news: if you make a reservation for a Steam Controller, the company will now show you one of three estimates of when you’ll be able to actually order your gamepad: by September 2026, by December 2026, or sometime in 2027. The bad news: any reservations made today “indicate a 2027 date for shipping,” Valve says.
“We have no plans to stop making Steam Controller,” according to Valve. “But as we look at the current demand compared to how many we know we can make by the end of the year, we want to manage expectations as much as we can with regards to when folks can expect to receive their order.”
Valve’s very good new Steam Controller went on sale in early May, and the initial rush led some people to run into frustrating problems with trying to check out ahead of the controllers eventually going out of stock. A few days later, the company announced that it would be implementing a reservations queue for interested buyers so they could get on a waitlist. If you’re on the waitlist, when you get notified that a Steam Controller is ready for you to buy, you have 72 hours to actually make the order.
“When we launched Steam Controller last month, we quickly saw that initial demand exceeded our expectations,” Valve says. “Switching to a reservation queue has (hopefully) cut down on the headaches on the customer side, and for us it’s also been helpful as we plan ahead and try to get as many out as quickly as we are able.”
All three of Valve’s big hardware products were delayed from a planned early 2026 launch because of the component crisis, Valve still hasn’t announced when the Steam Machine PC or Steam Frame VR headset might go on sale. However, just yesterday, Valve officially launched its big SteamOS 3.8 update with support for the Steam Machine. It’s also been importing a lot of hardware into the US as of late.
NEWYou can now listen to Fox News articles!
The next time you pull up to a McDonald’s drive-thru, the voice taking your order may not be human. McDonald’s is testing a new AI-powered system called ArchIQ at five U.S. locations. The company has not said where those restaurants are located. The voice assistant, nicknamed Archy, can take drive-thru orders and has shown it can handle both English and Spanish.
For anyone who has repeated “no pickles” into a speaker box more than once, this could sound helpful. However, if you remember McDonald’s last AI drive-thru experiment, you may also wonder whether your burger order could somehow turn into a bag full of surprise McNuggets.
Sign up for my FREE CyberGuy Report
- Get my best tech tips, urgent security alerts and exclusive deals delivered straight to your inbox.
- For simple, real-world ways to spot scams early and stay protected, visit CyberGuy.com – trusted by millions who watch CyberGuy on TV daily.
Plus, you’ll get instant access to my Ultimate Scam Survival Guide free when you join.
WOULD YOU EAT AT A RESTAURANT RUN BY AI?
McDonald’s is testing an AI drive-thru system called ArchIQ at five U.S. restaurants. (Kurt “CyberGuy” Knutsson)
What is McDonald’s AI drive-thru?
ArchIQ is McDonald’s new AI system for restaurants. It can take drive-thru orders and also help with operations behind the scenes.
In a post on X, McFranchisee, an anonymous McDonald’s franchisee account, said the system is currently in five test stores and has processed more than one million transactions. The account also said about 90% of orders were completed without a human stepping in. That number sounds promising. Still, McDonald’s has not confirmed a nationwide launch date. For now, this remains a limited test.
The system also appears to connect with a bigger McDonald’s plan called “McDonald’s > NEXT.” CEO Chris Kempczinski described the strategy as a way to bring in more customers and improve restaurant productivity. The plan also includes menu changes, restaurant redesigns, technology upgrades and more focus on hospitality.
Why McDonald’s is testing AI ordering
Drive-thrus can get chaotic fast. Someone changes an order after the total appears. A child calls out from the back seat. Road noise makes the speaker hard to hear. Then the driver remembers the extra sauce after everything has already gone through. That is the type of pressure McDonald’s wants AI to handle.
If ArchIQ works well, it could help restaurants move cars through the line faster. It may also reduce mistakes during busy hours. Workers could then focus more on preparing food, handling payments and helping customers who need a real person.
ArchIQ also appears to have a management role. In the same X post, McFranchisee described Archy as a tool that could alert managers to bottlenecks or other issues before they slow down operations.
STARBUCKS USES CHATGPT TO SUGGEST DRINKS BASED ON MOOD AS EXPERT WARNS OF HIDDEN DOWNSIDES
The AI assistant, nicknamed Archy, can take drive-thru orders and may also help managers spot restaurant slowdowns. (McFranchisee)
McDonald’s tried AI drive-thru ordering before
This new test follows McDonald’s earlier AI drive-thru experiment with IBM. That program involved more than 100 restaurants. McDonald’s ended the test in 2024 after customers complained about order accuracy. Some mistakes also went viral, creating an embarrassing moment for McDonald’s and raising questions about whether the technology was ready for the drive-thru. Customers reported wrong items, strange quantities and other order mix-ups. That history is why this new test will get extra attention.
This time, McDonald’s is working with Google technology. McFranchisee also claimed every McDonald’s in the U.S. is getting Google Edge Cloud hardware in anticipation of the rollout. McDonald’s seems to believe the newer system can perform better than the last one. The real test will come when regular customers use it during real drive-thru rushes.
How McDonald’s AI drive-thru could help customers
If McDonald’s gets this right, the most obvious benefit is speed. An AI ordering system does not get tired during a long shift. It may also help more customers order in the language they prefer. That could make a busy drive-thru feel less frustrating, especially during breakfast or late-night hours.
The system may also ask clearer follow-up questions and catch missing details before the order reaches the kitchen. That would be a win for customers who want to get in, get their food and get on with the day.
The biggest problem with AI drive-thru orders
The biggest concern is accuracy. AI can still misunderstand people. That gets frustrating fast when you are trying to grab lunch between errands or get your kids fed from the back seat. A wrong order wastes time. It also puts workers in the position of fixing a mistake the machine made.
There is also the customer service side. Some people like hearing a real person at the speaker. Others may find an AI voice cold or annoying, especially if the system gets confused.
Then there is the privacy question. If an AI system takes your order, customers may wonder what gets collected, how long it is kept and who can access it. McDonald’s has not publicly explained those specifics for this current ArchIQ test.
ALEXA+ LETS YOU ORDER FOOD LIKE A REAL CONVERSATION
A drive-thru menu board stands outside a McDonald’s restaurant in Hercules, Calif., on Oct. 23, 2024, amid an E. coli outbreak linked to onions in Quarter Pounder sandwiches that has sickened dozens and killed one person across the U.S. (David Paul Morris/Bloomberg via Getty Images)
How to avoid AI drive-thru mistakes
Before you leave the drive-thru, take a moment to check the order screen. Make sure the items match what you said. Listen when the system repeats your order. Keep your receipt until you confirm the food is right.
Also, avoid sharing extra personal details at the speaker box. Your order should only require your food choices and payment.
If the AI gets confused, ask for a crew member. You do not need to keep going back and forth with a machine over fries.
What this means for you
For now, you probably will not notice a change at your local McDonald’s. The ArchIQ test appears limited to five U.S. restaurants, and the company has not said when it could expand.
Still, this gives customers a preview of where fast food may be heading. AI could soon play a bigger role in how restaurants take orders and manage the kitchen. That may speed up the line, though it could also make the experience feel less personal.
Watch the CyberGuy Live replay: Lock Down Your Phone in 30 Minutes
Your phone holds your email, passwords, photos, banking apps and personal data. In this free CyberGuy Live replay, Kurt the CyberGuy walks you step by step through simple phone security fixes you can do at your own pace. You’ll learn how to improve your privacy settings, spot the latest phone scams, use trusted security tools and walk away with a simple checklist to stay protected. Watch the replay and get our checklist here: CyberGuyLive.com
Kurt’s key takeaways
McDonald’s clearly wants AI to play a bigger role in its restaurants. From a business point of view, the idea makes sense. Shorter drive-thru lines could help franchisees and customers. Better restaurant data could also help managers fix problems faster. But I still want the human backup. Food orders can be messy because people are messy. We change our minds. We talk over each other. We forget the extra ketchup until the last second. AI may handle much of that one day. For now, I would treat it like any busy drive-thru interaction. Speak clearly. Check the order. Do not pull away until you know your food is right.
Would you trust an AI voice to take your McDonald’s order, or do you still want a real person on the other end of the speaker? Let us know by writing to us at Cyberguy.com
CLICK HERE TO DOWNLOAD THE FOX NEWS APP
Sign up for my FREE CyberGuy Report
- Get my best tech tips, urgent security alerts and exclusive deals delivered straight to your inbox.
- For simple, real-world ways to spot scams early and stay protected, visit CyberGuy.com – trusted by millions who watch CyberGuy on TV daily.
- Plus, you’ll get instant access to my Ultimate Scam Survival Guide free when you join.
Copyright 2026 CyberGuy.com. All rights reserved.
Midjourney CEO David Holz just showed off the company’s first hardware product and plans to build a San Francisco spa, which he admitted is a bit different from the “cat pictures” produced by its AI image generator. Dubbed The Midjourney Scanner, it’s an ultrasound-based full-body scanner that uses a ring of sensors to capture vertical slices of the inside of your body, looking at the composition of your muscle, fat, bone, and organs to start. Holz said ideally, you could do this once a year or every single day, as it “aims for image quality comparable to MRI in many ways.”
He mentioned that one way he’d like to use it would be to see how his body changes in response to diet and workout changes, saying, “I’m not the most measured man on Earth yet, you know, but maybe I want to have that daily [measurable information].” A set of job listings advertises the company’s goal as trying to “build and launch the world’s first full-body ultrasound CT scanner, ultimately bringing safe, fast, and high fidelity preventative scanning to billions via a magical spa experience.”
The Midjourney Scanner was developed in a partnership with ultrasound tech company Butterfly Network, which said it uses “40 Butterfly Ultrasound-on-Chip imaging modules per system.”
The scanning process starts with stepping onto a platform that drops down into the water on rails through a ring of thousands of transducers that create ultrasonic waves. It then records the ripples passing through your body to analyze them and create detailed 3D images. The scan takes about 60 seconds. Holz said about a dozen people have been scanned so far.
It starts by stepping into a shallow pool of golden light. You then begin to descend into the water. Your body passes through a ring of underwater sensors, each acting like a dolphin, using its echolocation. The sensors send ultrasonic sound waves through your body from every angle. With enough waves, and enough angles, we form an image of what’s happening inside your body.
It combines those sensors with two petaflops of processing power. But after watching the livestreamed reveal, I’m still unclear on what Midjourney’s AI image generation tech exactly has to do with the Midjourney Medical effort, beyond an alternative business for otherwise-unused AI compute.
Holz hopes to put 10 of the scanners into a Midjourney Spa location in San Francisco’s Union Square that will open before the end of 2027 and offered to scan the hands of attendees at its launch event. The Midjourney Spa will have a gym, saunas, and cold plunges to go along with the hot tub–equipped scanning rooms where visitors will get into the water to be scanned.
He did mention that various medical applications would require FDA clearances, but for now, Midjourney Medical says it’s working on “body composition maps” that don’t require the same level of clearance as diagnostic imaging. It also says the “library of scans” users create can be shared with doctors, AI health tools, or others, and that, “We take data privacy seriously — more details on our data policies will come as we get closer to launch.”
Holz suggested that eventually these scans could become better than an MRI, without radiation, powerful magnets, or other complicating factors, to get a look at what’s going on inside people’s bodies “real fast.” In response to a question, he imagined a future where the FDA had a class of devices to look at “weird” things and allowed people to “just try to get as much data as we can.”
-
San Francisco, CA9 minutes agoOakland man faces hate crime charges for Castro District attack
-
Dallas, TX12 minutes agoAt least 4 injured after vehicle drives into Dallas crowd, driver arrested
-
Miami, FL17 minutes agoMiami Central students prepare for life changing trip to Zimbabwe amid funding challenges
-
Boston, MA24 minutes agoGiannis to Boston is a possibility. Should the Knicks be worried?
-
Denver, CO27 minutes agoDenver Public Schools’ decline in enrollment continues to reshape district
-
Seattle, WA32 minutes agoWoo twirls 7 scoreless innings in bounceback outing vs. Orioles
-
San Diego, CA39 minutes agoNASCAR makes history with inaugural Naval Base Coronado race
-
Milwaukee, WI42 minutes agoStolen Oak Creek couple’s car found by viewer dumped near Milwaukee apartment complex
