Users may not be able to combine Microsoft software with competitors’ products effectively, the CMA has been told.

The UK’s competition watchdog has launched an investigation into Microsoft’s enterprise software ecosystem over interoperability concerns.

The Competition and Markets Authority (CMA) probe will consider whether Microsoft should be given a ‘strategic market status’ (SMS) designation reserved for companies that dominate the UK market in a particular digital activity.

Apple and Google received this status last year over their dominance of mobile platforms.

The investigation will also assess whether Microsoft’s product bundling, limits in interoperability for users, and default settings prevent customers from switching to competing services. This will include probing how Microsoft’s AI competitors are able to integrate with its business software.

Hundreds of thousands of UK organisations across public and private sectors use Microsoft’s enterprise software daily, spanning the Windows operating system, Microsoft 365 suite and Copilot AI.

The CMA finds that Microsoft has more than 15m commercial UK-based users across its ecosystem, making it a key provider of productivity tools in the country.

The competition watchdog said it has received information that users may not be able to combine Microsoft software with competitors’ products effectively.

In its probe, it will examine Microsoft’s wide range of products, including productivity software, personal computer and server operating systems, database management systems and security software.

The investigation will take up to nine months to complete. The authority has invited organisations based in the UK and around the world, including rival tech companies and business software customers, to share their experiences.

This is the fourth SMS investigation opened by the CMA since the UK’s digital markets competition regime came into force in January 2025. The law gives the CMA additional powers to propose remedies and improve market competition in the country. Early last year, it opened a probe into Apple and Google over their mobile ecosystems,

The designation would also allow the CMA to potentially intervene on concerns from a separate investigation into Microsoft’s software licensing in the cloud market.

“Business software is a cornerstone of how the UK economy functions, from small businesses to major public services and infrastructure. Hundreds of thousands of customers in the UK rely on Microsoft’s systems, which is why it’s so important to ensure these services are delivering good outcomes,” said Sarah Cardell, the CMA’s CEO.

“Our aim is to understand how these markets are developing, Microsoft’s position within them and to consider what, if any, targeted action may be needed to ensure UK organisations can benefit from choice, innovation and competitive prices.”

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

For AI systems to keep improving in knowledge work, they need either a reliable mechanism for autonomous self-improvement or human evaluators capable of catching errors and generating high-quality feedback. The industry has invested enormously in the first. It’s giving almost no thought to what’s happening to the second.

I’d argue that we need to treat the human evaluation problem with just as much rigor and investment as we put into building the model capabilities themselves. New grad hiring at major tech companies has dropped by half since 2019. Document review, first-pass research, data cleaning, code review: Models handle these now. The economists tracking this call it displacement. The companies doing it call it efficiency. Neither are focusing on the future problem.

Why self-improvement has limits in knowledge work

The obvious pushback is reinforcement learning (RL). AlphaZero learned Go, chess, and Shogi at superhuman levels without human data and generated novel strategies in the process. Move 37 in the 2016 match against Lee Sedol, a move professionals said they would never have played, didn’t come from human annotation. It emerged from AI self-play. 

What enables this is the stability of the environment. Move 37 is a novel move within the fixed state space of Go. The rules are complete, unambiguous, and permanent. More importantly, the reward signal is perfect: Win or lose, and immediate, with no room for interpretation. The system always knows whether a move was good because the game eventually ends with a clear result.

Knowledge work doesn’t have either of those properties. The rules in any professional domain are dynamic and continuously rewritten by the humans operating in them. New laws get passed. New financial instruments are invented. A legal strategy that worked in 2022 may fail in a jurisdiction that has since changed its interpretation. Whether a medical diagnosis was right may not be known for years. Without a stable environment and an unambiguous reward signal, you cannot close the loop. You need humans in the evaluation chain to continue teaching the model.

The formation problem

The AI systems being built today were trained on the expertise of people who went through exactly that formation. The difference now is that entry-level jobs that develop such expertise were automated first. Which means the next generation of potential experts is not accumulating the kind of judgment that makes a human evaluator worth having in the loop.

History has examples of knowledge dying. Roman concrete. Gothic construction techniques. Mathematical traditions that took centuries to recover. But in every historical case, the cause was external: Plague, conquest, the collapse of the institutions that hosted the knowledge. What’s different here is that no external force is required. Fields could atrophy not from catastrophe but from a thousand individually rational economic decisions, each one sensible in isolation. That’s a new mechanism, and we don’t have much practice recognizing it while it’s happening.

When entire fields go quiet

At its logical limit, this isn’t just a pipeline problem. It’s a demand collapse for the expertise itself.

Consider advanced mathematics. It doesn’t atrophy because we stop training mathematicians. It atrophies because organizations stop needing mathematicians for their day-to-day work, the economic incentive to become one disappears, the population of people who can do frontier mathematical reasoning shrinks, and the field’s capacity to generate novel insight quietly collapses. The same logic applies to coding. Our question is not “will AI write code” but “if AI writes all production code, who develops the deep architectural intuition that produces genuinely novel systems design?” 

There is a critical difference between a field being automated and a field being understood. We can automate a huge amount of structural engineering today, but the abstract knowledge of why certain approaches work lives in the heads of people who spent years doing it wrong first. If you eliminate the practice, you don’t just lose the practitioners. You lose the capacity to know what you’ve lost.

Advanced mathematics, theoretical computer science, deep legal reasoning, complex systems architecture: When the last person who deeply understands a subfield of algebra retires and no one replaces them because the funding dried up and the career path disappeared, that knowledge isn’t likely to be rediscovered any time soon. 

It’s gone. And nobody notices because the models trained on their work still perform well on benchmarks for another decade. I think of this as a hollowing out: The surface capability remains (models can still produce outputs that look expert) while the underlying human capacity to validate, extend, or correct that expertise quietly disappears.

Why rubrics don’t fully substitute

The current approach is rubric-based evaluation. Constitutional AI, reinforcement learning from AI feedback (RLAIF), and structured criteria that let models score models are serious techniques that meaningfully reduce dependence on human evaluators. I’m not dismissing them.

Their limitation is this: A rubric can only capture what the person who wrote it knew to measure. Optimize hard against it and you get a model that’s very good at satisfying the rubric. That’s not the same thing as a model that’s actually right.

Rubrics scale the explicit, articulable part of judgment. The deeper part, the instinct, the felt sense that something is off, doesn’t fit in a rubric. You can’t write it down because you need to experience it first before you know what to write.

What this means in practice

This isn’t an argument for slowing development. The capability gains are real. And it’s possible that researchers will find ways to close the evaluation loop without human judgment. Maybe synthetic data pipelines get good enough. Maybe models develop reliable self-correction mechanisms we can’t yet imagine.

But we don’t have those today. And in the meantime, we’re dismantling the human infrastructure that currently fills the gap, not as a deliberate decision but as a byproduct of a thousand rational ones. The responsible version of this transition isn’t to assume the problem will solve itself. It’s to treat the evaluation gap as an open research problem with the same urgency we bring to capability gains.

The thing AI most needs from humans is the thing we’re least focused on preserving. Whether that’s permanently true or temporarily true, the cost of ignoring it is the same.

Ahmad Al-Dahle is CTO of Airbnb.

Welcome to the VentureBeat community!

Our guest posting program is where technical experts share insights and provide neutral, non-vested deep dives on AI, data infrastructure, cybersecurity and other cutting-edge technologies shaping the future of enterprise.

Read more from our guest post program — and check out our guidelines if you’re interested in contributing an article of your own!

Ahead of the return to District Court for the continued appeals process in the Apple-Epic legal fight, the two companies have made a schedule to discuss a change in App Store fees for outbound links.

The May 6 decision by the Supreme Court to decline Apple’s request for a stay on a mandate to meet Epic Games to negotiate a new commission rate means more legal action in the coming months. In preparation, the two sides have agreed on a few ground rules for that meeting.

In a joint filing to the U.S. District Court for the Northern District of California on May 15, there is a multi-point schedule of actions and activities for the two companies.

The list starts with Apple being given 45 days to file with the court a “proffer,” namely a good-faith offer of evidence, testimony, documents, or some form of evidence to the court. This proffer will propose commissions for any “linked-out purchases,” along with supportive evidence, in a document that’s up to 30 pages long.

Up to ten days after the proffer, Apple will hand Epic all non-privileged documents about the decision-making process to create the proffer’s proposal. This includes fee proposals and a privilege log.

Within five days of that action, Apple will meet with Epic to discuss the privilege log and whether Epic needs any extra material to evaluate the Apple proposal. Epic can then designate up to 10% of the documents on that privilege log for further review by a third party.

Within 60 days from the latter of Apple’s proffer filing or Apple’s completion of document production, Epic will then file its own response to Apple’s proffer to the court. This too can be up to 30 pages in length, and if Epic makes an objection, it must include evidence supporting that move.

Once that has been filed, Apple has another 30 days to file its reply, though with a 15-page limit. After all that, the court can then hold a status conference or decide any further proceedings.

Based on all parties taking the last possible moments to complete each stage in the schedule, it could be 150 days or five months before the court schedules another meeting.

How Apple got here

Apple and Epic have endured a six-year legal battle that started with “Fortnite” allowing players to make in-app purchases using a third-party payment processor. Epic also made demands, including allowing alternate app storefronts in iOS and a change in the commission structure.

For the most part, Apple succeeded against Epic, but the court found against Apple’s anti-steering measures, which prevented developers from directing users to other ways to pay. Apple was ordered to make changes, and it claimed it did.

Epic responded, arguing Apple didn’t follow the spirit of the law, convincing a court in April 2025 to its viewpoint. The court viewed Apple’s moves as a “gross miscalculation” of acceptability.

Across the country, thousands of students are waving goodbye to high school and preparing to head off to college. If you know one such student, and want to gift them a gadget to send them off in style, you’ve come to the right place. I’ve rounded up what I think are the best high school graduation gifts you can buy, at a range of different budgets.

All of these gifts will come in handy in college, whether for work, play or general student life. I’ve included noise-cancelling headphones to block out distractions while studying, robust Bluetooth speakers for when it’s time to bring the party, and budget-friendly coffee makers for an instant boost in time for morning seminars.

The GameSir G8+ MFi is the controller that the G8 Galileo probably should have been from the start. It takes the same full-size grips, Hall effect thumbsticks and triggers, and programmable back buttons that made the original one of the best mobile grips going, then layers on MFi certification for iPhone and iPad Mini, swappable ABXY button caps, dual vibration motors, and crucially, improved case compatibility.

Best of all? It’s still the same $79.99 / £79.99 / AU$129.99 asking price.

Short battery life frustrates many smartwatch owners more than anything else. Prices for decent models climb higher every year. CMF by Nothing Watch 3 Pro, priced at $69 (was $99), changes that equation completely. A 47-millimeter case sits comfortably on most wrists at just 51 grams. Liquid silicone straps attach easily with standard 22-millimeter pins so swapping colors takes seconds. Brushed details around the rotating crown add a subtle finish that feels more expensive than the asking price suggests.

The screen size increases to a pocket-sized 1.43 inches, with a square 466 by 466 pixel AMOLED panel that produces sharp text and silky smooth swipes because the refresh rate remains at a steady 60 Hz. Even in harsh sunshine, the maximum brightness of 670 nits allows you to keep track of the time. And if you want to save battery life, the always-on display remains visible – just don’t expect deep blacks; that’s the price you pay for this function.

Sale

CMF by Nothing Watch 3 Pro Smart Watch for Men Women, Bluetooth Call AI Energy Score, 1.43” Heart Rate…

• Ultra-Accurate Health & Fitness Tracker: The CMF Watch 3 Pro smart watches enhance your fitness journey with advanced heart rate tracking (4-channel…
• Your AI Life Companion: The CMF Smartwatch enables crystal-clear Bluetooth calls with AI noise reduction, a built-in mic, and speaker, supporting…
• Style for Every Occasion: Elegant Brushed Metal to Rugged Outdoor Designs feature a vibrant 1.43” AMOLED Display with sharp resolution, bigger…

The battery itself has 350mAh capacity and could last up to 13 days on a single charge. If you’re a heavy user, however, and spend a lot of time tracking your heart rate or GPS route, you’ll get at least 10 days out of it, even if you’re constantly checking your watch. And if you elect to leave the always-on display on, you should be able to go about 4-5 days between top-ups, which is plenty of time between charges in my opinion. You can put the watch on the charger for a short 100 minutes, just enough time to have a coffee, and then you’re ready to go.

When it comes to health tracking, Nothing has worked harder to get the basics right: continuous heart rate monitoring with updates every few minutes, as well as the ability to check blood oxygen levels and stress levels. It’ll even tell you what stage of sleep you’re in, dividing it into neat little charts of time and quality each morning. For the females, there’s a built-in period tracker that will tell you when ovulation is due. There are also other sports modes available, over 130 in all, so whether you’re a serious athlete or just want to go for a brief jog, you should be able to find something that suits you. The dual-band GPS works great, and it even identifies when you start or end an exercise, so you don’t have to fiddle with manual start/stop times. All of that data is then synced to the Nothing X app, where you can see how you’re doing over time. I mean, if you’re a real runner, you’ll want a little more precision, but for the typical user, it’s plenty to keep you on track.

The Bluetooth 5.3 onboard allows you to take and make calls directly from the watch, as long as you have a built-in microphone and speaker. You may also receive notifications from your phone apps on the watch, which you can reject or respond to without having to reach for your phone. Additionally, you can manage your music directly from the watch, which should be a big advantage for all you runners out there.

The watch is water resistant to the tune of IP68, so a good splash in the rain or a fast hand wash should be fine, but, as with other similar watches, you should avoid deep water sports, since swimming laps are out of the question. As an extra plus, the lack of an NFC chip allows you to utilize your phone to make contactless payments.

Announced almost a year ago, the golden Trump phone will finally begin shipping this week to people who preordered it, according to Trump Mobile.

“Those who pre-ordered the T1 Phone will be receiving an update email. Phones start shipping this week!!!,” Trump Mobile said in an X (Twitter) post on Wednesday. A Community Note added to the post suggests that the long-delayed phone is a modified version of the T-Mobile REVVL 7 Pro 5G, which was first released in early 2024.

The Trump phone, officially known as the Trump Mobile T1, was supposed to ship last August, but ran into production issues, the company said.

“The technology business is more difficult than some may realize, as parts must be tested for quality assurances,” Trump Mobile CEO Pat O’Brien said in a statement to CNET. “We have experienced delays during a variety of steps in getting the T1 to completion, but those delays were worth it in our minds as we are delivering an amazing product.”

CNET Director of Content Patrick Holland had put down a $100 deposit to preorder the T1 phone in June 2025 so CNET could review it, but he has yet to hear from the company about a shipping date. However, on Wednesday morning, when he logged in to his Trump Mobile account, he was notified that his payment method was outdated and needed to be updated.

Holland says his preorder was “confirmed but not processed,” and he sees a red bubble that reads, “Awaiting Sim Assignment.”

Advertisement

Trump Mobile launched in June 2025 with a $47.45-a-month mobile phone plan, and announced that a Trump phone would be made in the US and launch in August 2025. But when it became obvious that domestic large-scale smartphone manufacturing would not be possible, Trump Mobile dropped the “made in the US” claim. 

Now, the site simply says it is “designed with American values in mind,” an “American-proud design” and “shaped by American innovation.”

Last month, a redesigned Trump Mobile website finally showcased a new-look T1 phone, its third redesign. It followed Trump Mobile executives showing off a handset they said was a near-production version of the T1 phone to The Verge in early February. 

The website claims it will have a 6.78-inch AMOLED screen; a 50-megapixel, front-facing camera; a camera bump featuring a 50-megapixel main lens, an 8-megapixel wide lens and a 50-megapixel (2X Tele) lens; a 5,000-mAh battery; a fingerprint sensor and AI face unlock; a Snapdragon mobile platform; and an Android operating system.

The site is still accepting $100 deposits on the Trump phone, which it continues to advertise as having a promotional price of $499. Trump Mobile also sells refurbished Apple and Samsung phones, ranging from $369 to $629. 

While Trump Mobile has not said where the T1 phone is manufactured, executives told The Verge that it’s being made in a “favored nation” with “final assembly” in Florida. It’s unclear what qualifies another nation as “favored” to handle most of the assembly of the T1 phone. Meanwhile, President Trump has returned from China after a two-day summit with President Xi Jinping to discuss a potential trade deal.