Google's latest feature drop might be a big win for summer travel
As with flights, you can now track prices and set alerts for hotels
Google Maps can also now pull potential places to visit from screenshots
If you’re a fan of Google Flights, especially for the price tracking data and how the current prices you’re seeing rank against other days, you’re in for a treat. As part of a drop of features fit for upcoming summer travel, Google aims to do for hotels what it’s done for flights.
And yes, it’s as good as it sounds. Now, when you search for hotels on Google, you’ll have the option to ask the search giant to track prices. Essentially, you turn on the feature and then get an alert if there is a price drop.
Similar to flights, you can be a bit descriptive, setting a price range or a 'don’t bother me if it doesn’t fall' here. It will even factor in a star rating if you have one selected and the general area where you were searching for a hotel.
(Image credit: Google)
Google is rolling out this new hotel price tracking feature globally on desktop and mobile. Once it’s available, you’ll find it right in search, complementing the historical knowledge of hotel pricing history.
This hotel-focused feature is launching alongside some other new functionality from Google, all billed under getting ready for summer travel. The ability to set up price alerts for hotels is undoubtedly the most user-friendly feature and could have the most significant impact. It could potentially help you save on a stay.
Another new feature that could help you better prepare for a trip is screenshot support within Google Maps. If you enable it, Google Maps will look through photos and deliver a list of places you've screenshotted.
So, if you've been screenshotting TikToks about the best places to eat in New York City or maybe a list of the best ice cream spots in Boston, you won't need to dig through all of them to find every place mentioned.
Instead, with some AI help, Google Maps will look through your screenshots, find those spots, and list them well in a handy list for you. It'll live in the app in a list titled "Screenshots," and this feature is entirely optional.
(Image credit: Google)
This feature could prove helpful, but considering that screenshots aren’t just used for travel or remembering specific spots, this could also be a bit of a privacy concern.
It is opt-in only and not on by default, but it is rolling out now to mobile devices with U.S. English on iOS first, with Android following shortly.
Microsoft has reportedly canceled data center leases amid shifting infrastructure priorities
TD Cowen analysts suggest lease deferrals reflect oversupply vs demand
Microsoft doesn't deny the report but says it's investing $80 billion in infrastructure
We recently reported Microsoft had cancelled leases with at least two private data center operators in the US, totaling “a couple hundred megawatts,” and how the company was also not converting “so-called statements of qualifications into leases,” according to claims from TD Cowen analysts.
Shortly after that news broke, Microsoft pulled out of a $12 billion deal with CoreWeave (the "WeWork of AI"), passing on buying more data center capacity from the AI hyperscaler. That option was snapped up by OpenAI, but as it counts Microsoft as its biggest backer, it was essentially paying CoreWeave with Microsoft money!
Microsoft appears to be taking a more tactical approach to AI spending, a move that is echoed in a new Bloomberg report quoting TD Cowen analysts saying Microsoft has walked away from additional data center projects in the US and Europe.
On track to spend $80 billion
Bloomberg writes, “Microsoft’s retrenchment in the last six months included lease cancellations and deferrals, the TD Cowen analysts said in their latest research note. Alphabet Inc.’s Google had stepped in to grab some leases Microsoft abandoned in Europe, the analysts wrote, while Meta Platforms Inc. had scooped up some of the freed capacity in Europe.”
Responding to the article, Microsoft pointed out that it was still on track to spend about $80 billion investing in growing infrastructure projects.
“Thanks to the significant investments we have made up to this point, we are well positioned to meet our current and increasing customer demand,” a Microsoft spokesperson said in a statement.
“While we may strategically pace or adjust our infrastructure in some areas, we will continue to grow strongly in all regions. This allows us to invest and allocate resources to growth areas for our future.”
Bloomberg adds TD Cowen analysts Michael Elias, Cooper Belanger, and Gregory Williams said, “We continue to believe the lease cancellations and deferrals of capacity points to data center oversupply relative to its current demand forecast.”
AI has dominated headlines, product strategies, and investment for the past two years, but as businesses reflect heading into 2025, an uncomfortable question lingers: where is AI’s financial impact, really?
Amidst the AI hype climate, businesses have been eager to invest in emerging technologies that promise the world. So eager, in fact, that the market is now saturated with hastily-developed products designed more to showcase adoption than to deliver measurable impact.
While AI tools have existed for some time, the rise of generative AI — starting with the release of ChatGPT just over two years ago — has captured broader attention and rekindled a frenzy of innovation, akin to the dot-com boom of the late 1990s. Generative AI's accessibility is lowering barriers to entry, sparking both a rush to investment and concern across industries.
When applied strategically, it’s clear that AI can revolutionize user experiences at places like websites, where the potential for customer experience enhancement is unparalleled. But to keep up with lofty AI predictions and heavy investor demands, many businesses today are investing first and looking for ways to measure return on that investment later — an approach that’s led to over-promised and under-delivered initiatives, followed by disappointment among customers and teams alike.
AI has the potential to deliver transformative outcomes when businesses align it with strategic goals, such as improving website functionality and user satisfaction. Rather than integrating AI first and looking for a problem to solve later, it’s time to return to the tried and true formula for innovation: find a problem, figure out how to solve it.
The AI Paradox: ROI vs. FOMO
ROI must be the central factor in AI investment decision-making.
While the number of senior business leaders investing $10 million or more in AI is set to double next year, a Gartner report found that at least 30% of AI projects will be abandoned by the end of 2025. These circumstances — high costs and low success rates — make prioritizing business needs and ROI critical.
AI is an expensive, time consuming endeavor — so in order for an AI product to be worth it, especially in consumer-facing applications, it must add real value to customers. Rushing to bring high-potential technology to market can often hinder, rather than enhance, user experience — particularly in the case of websites, where users increasingly demand seamless interaction.
The early surge of businesses racing to adopt AI chatbots is a prime example. In the push to get the latest feature onto their websites, a critical question was often overlooked: will this actually improve our customer service? Despite their high-potential, chatbots were introduced widely before the technology was developed to the point of adding proper value, often resulting in frustrating user experiences and failing to provide accessible support.
Consider Watsonville Chevy’s viral failure, where a chatbot offered to sell a customer a brand new Chevy vehicle for only $1. Rather than helping customers buy cars, the under-developed technology — despite its high-potential — caused an embarrassing headache for the dealership. More than the technology itself, this failure underscores the critical importance of businesses putting appropriate guardrails in place. Effective AI implementation requires not only understanding the potential of off-the-shelf solutions, but ensuring they are adapted to the specific needs and limitations of the business environment.
In another recent counterproductive AI use case, Spotify whiffed on its perennially popular Wrapped feature by going all-in on AI. While removing features like top genres, the music streaming giant opted to add experiences like an AI-generated podcast. Listeners, predictably, were critical, highlighting the importance of using AI to enhance user experience rather than diminishing features that made products popular in the first place.
Businesses must ask themselves: is this AI use case truly adding to the customer experience? Investments must prioritize functionality and customer needs over hype. By focusing on thoughtful, ROI-driven AI adoption, businesses can avoid costly mistakes and improve outcomes.
Solving Problems, Driving Results
But of course, not every AI investment is destined to fail or hinder customer experience. There are many examples of how AI brings clear value-adds when implemented strategically, especially in ecommerce and content.
In a crowded online environment, frustrated users are a sales killer — customers have access to limitless products and content online, so when their search fails to draw results, they leave.
Netflix, Google and Amazon have all dominated their respective verticals for several years in no small part because of their use of natural language processing. In 2017, more than 80% of the TV shows users watched on Netflix were discovered through its recommendation system. In 2012, 35% of purchases on Amazon came from product recommendations. Google has utilized AI since 2015 to process and provide more relevant search results.
All of these use cases have had a tangible impact on customer experiences — an impact that has long differentiated them from competitors with less resources.
More recently, the rise of generative AI is amplifying this trend — as the current AI frenzy has led to not only the development of cutting-edge technology by major AI players, but also made it more accessible for businesses of all sizes. These advancements enable even mid-market players to leverage tools once exclusive to industry giants, creating new opportunities for differentiation and growth.
In short, these technologies are no longer limited to just big tech like FAANG — AI democratization is unlocking more cost effective tools for SMBs and mid-market companies to optimize their websites. The potential of AI is immense when businesses choose the right product, and legacy examples in e-commerce and user experience, like Netflix’s recommendation engine or Amazon’s personalized shopping, offer a blueprint for websites leveraging AI to create transformative, ROI-driven outcomes. AI-driven smart search and recommendation technologies already exist; businesses just need to evaluate problems and responsibly implement solutions to unlock AI's transformative potential.
Starting with the pain point — such as low conversion rates — and answering with AI empowers better business outcomes. Implementing AI technologies can help bridge the gap between mid-market companies and enterprises, drive higher conversion rates and justifiable ROI for the cost of AI.
Thoughtful AI integration has the potential to revolutionize industries, and in some cases, already has. These use cases underscore AI’s potential to add real value, making a tangible impact on conversion rates and customer experience.
This article was produced as part of TechRadarPro's Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here:https://www.techradar.com/news/submit-your-story-to-techradar-pro
from Latest from TechRadar US in News,opinion https://ift.tt/TCwEm9a
The Poco M7 Pro is a budget smartphone that offers a big display, a big battery, and a big relief to your pocket. It is equipped with a MediaTek Dimensity 7025 Ultra chipset, a 50-megapixel primary rear camera, a 6.67-inch AMOLED display with 120Hz refresh rate, and a 5,110mAh battery. All of this is available at a starting price of Rs. 14,999, but does that make it a...
MagStor's Thunderbolt 5 LTO drive builds on its 2020 Thunderbolt 3 model
TB5 certainly adds speed but what's the real-world benefit for tape?
There's no word on pricing, but it's unlikely to be cheap
MagStor introduced the world’s first Thunderbolt 3 LTO tape drive back in 2020, blending traditional tape-based storage with modern connectivity, and now, the company has announced the world’s first Thunderbolt 5 LTO tape drive.
The company describes its latest product as the next step in offering flexible, high-speed backup and archival solutions for professionals working with large volumes of data.
Tape storage continues to be a standard for long-term archival needs due to its durability and capacity, and the Thunderbolt 5 LTO drive is designed for use in data-heavy environments such as media production and enterprise IT. By integrating Thunderbolt 5, MagStor hopes to offer a faster, more streamlined connection between tape hardware and modern computing systems.
Increased speed
The new drive works with both macOS and Windows and while Thunderbolt 5 offers higher bandwidth than previous versions, tape speeds remain limited by the format itself.
Although MagStor hasn’t provided many technical specifications, it’s a given the new product will support LTO-9 tapes (18TB native / 45TB compressed capacity), as its predecessor does.
There's no confirmation yet on compatibility with next-generation LTO-10 tapes, expected to arrive in the second quarter of 2025, which offer up to 36TB native and 90TB compressed capacity, but it would be a missed opportunity if that support isn’t included.
Thunderbolt 5 achieves data transfer rates of up to 80Gbps (10GB/s) bi-directionally in standard mode, and up to 120Gbps in one direction when using Bandwidth Boost mode.
LTO-10 is expected to deliver read speeds of around 472MB/s, which is a step up from LTO-8 at 360MB/s and LTO-9 at 400MB/s.
MagStor says the Thunderbolt 5 LTO drive will be released by the end of 2025. Pricing has not yet been announced, but it's unlikely to be cheap.
The company’s LTO-9 Thunderbolt 3 drive retails for $6,299 and whether the added speed of Thunderbolt 5 will justify the inevitable price hike remains to be seen.
“At MagStor, we are committed to pushing the boundaries of what’s possible in data storage,” said Tim Gerhard, VP of Product at MagStor. “After revolutionizing the market with the first-ever Thunderbolt 3 LTO drive, we’re excited to raise the bar again with Thunderbolt 5, ensuring our customers have access to the most powerful and flexible storage solutions available.”
Look, I love Windows, I do, I really do. It's one of those things that I just can't live without at this point. I've tried MacOS, I've tried Linux, I've even dabbled in the world of Android and Chromebooks during my time, and yet, none of it compares to Windows; it just doesn't.
There's a certain amount of familiarity, of indoctrination into that Microsoft cult that's rife in me. I grew up using Windows 98, and onwards, it was what I gamed on, what I studied on, what I made lifelong friends on—you name it. 98, 2000, XP, Vista, 7, 8, 10, and finally we're here at Windows 11, at least until Microsoft inevitably tells us that its "final" operating system isn't its final operating system.
The thing is though, it really is a bag of spanners at times, and I've kinda developed this habit of going thermonuclear on my own machine at quite regular intervals over my lifetime.
Mostly by flattening and reinstalling Windows onto my PC every other month or so. Why? I'm glad you asked.
The need for an occasional refresh
Well, the thing is, although Windows gives you a lot of freedom and has broad compatibility with more programs than any other operating system out there, it does make it somewhat susceptible to bugs. Lots of them.
These can be inflicted by Microsoft directly through Windows Updates or drivers accidentally corrupting files or programs, or well, any number and manner of avenues.
The worst culprit, of course, is the classic "upgrade from the previous Windows version to this version." Just don't; it's never worth it.
Windows is great, but no operating system is designed to run perfectly forever.(Image credit: Microsoft)
See, registry files corrupt, file directories get mislabelled, and inevitably you'll end up with programs you forget about sitting in the background sucking up critical resources. It's just a bit crap like that, and ironically, although I do have a massive disdain towards macOS, I can't deny its closed-off ecosystem does avoid a lot of these pitfalls.
Whenever anyone asks me about a system bug or help with troubleshooting, my first and often instant reaction is to suggest just flattening the machine entirely and reinstalling a fresh version of Windows on top.
That's why I advocate tying a full-fat Windows license to your Microsoft account so you can easily reinstall and activate Windows 11 on your machine on a dime.
An arduous task
It does require some getting used to this salting-the-earth kind of strategy, but the benefits are just too great to ignore.
The first thing I recommend is splitting up your storage solution. In every build I've ever done, I've almost always recommended a two-storage drive system. The first and fastest of the two should be used as your main OS drive, and the second, usually slower, cheaper, and larger, being your media/games/back-up drive. Any valuable documents, assets, or big downloads live here.
What that allows you to do is keep all your games and important files on your D: drive, and then, whenever that re-install time comes a-calling, allow you to quickly flatten and re-install Windows on your C: drive.
If you've got slow internet or just can't be bothered to re-download everything, it is a huge time-saver doing it this way. You can get away with partitions, but it's far easier to accidentally delete the wrong one on your next Windows install.
Laptop, desktop; it doesn't matter, just give your hardware an OS break now and then.(Image credit: Sergey Kisselev / Behance.net / Microsoft)
It also helps really reduce program and document clutter and encourages good back-up practice too. If you know you're going to flatten a machine every 2-3 months, then the likelihood is you'll keep all of your important files and documents safely stored in the cloud, or off-site, backed up with solid authentication procedures as well.
You'll end up with a minimal desktop that's stupidly rapid, clean, up-to-date, and as error-free as Microsoft can muster. If you're building a new PC or transferring an old one to updated hardware, save yourself the hassle and just back up and move your most important files, download a fresh USB Windows Installer, and get cracking. I promise you it's worth it.
A new lease on (virtual) life
With that, and good internet education and practice, plus a solid VPN, you can then dump aftermarket antivirus as well and rely on good ol' Windows Defender. It's one of the best antivirus programs out there, and lacks the resource vampirism many third-party solutions have.
Worst-case scenario, you get tricked into opening a dodgy email or land on an odd website, and your machine gets whacked with some crypto-scam; just flatten it. Job done. Although again, I'd highly recommend just being a bit more internet savvy first.
The only thing I'd say if you do go this route, be careful on the device you do it on and prep accordingly. Some motherboards won't support ethernet or wireless connectivity without drivers too.
Grab your USB stick, get the Windows Installer setup on it, and then stick a folder in it called DRIVERS. Head to your motherboard's product page, grab the relevant drivers, then once you're finally on the desktop, you should be able to install all your chipsets and drivers and get that internet connectivity back, no sweat.
If you do get stuck on the "need to connect to the internet" Windows 11 install page, hit Shift + F10, click the command window, type OOBE\BYPASSNRO, and hit enter. The installer will reboot, and you'll now have the option to tell Microsoft you "don't have the internet" and continue with the installation regardless.
So yeah, PSA complete. I got 99 problems, and most of them are Microsoft-related. At least for about 20 minutes anyway.
from Latest from TechRadar US in News,opinion https://ift.tt/fGyj3XU
Welcome to our live coverage of World Backup Day, which starts early on Sunday, the 30th (midnight GMT, but there is already 1300 in Auckland, New Zealand). We will update this live blog a few times today, with our real-time coverage of the day starting at 0830 UK time with regular updates throughout Monday, March 31st.
Making sure your data is safe and protected has never been so important as we continue our way in the wild world of digital transformation, but with so much to do these days, it can sometimes go forgotten. Don't worry though, we've rounded up
advice from our experts to make sure your data is protected.
horror tales from our team and elsewhere to remind you of what can happen
exclusive deals from our backup partners to keep your data safe
backup content from our extensive archive
data backup stories around the world
And much more!
Save 20%
Acronis Cyber Protect offers an integrated secure backup and rapid recovery solution for your business for data protection with cybersecurity features. AI-based threat detection is one of the standout features that helps users stay protected from the latest malware, ransomware, and cryptojacking attempts.View Deal
Save 75%
Whether you’re a business safeguarding critical data or a home user looking to protect precious memories, Carbonite delivers peace of mind with top-notch security and simplicity. There are three plans to choose from and they cover automatic cloud backup, ransomware protection, unlimited cloud storage, easy recovery, cross-platform support, and encrypted data protection.View Deal
from Latest from TechRadar US in News,opinion https://ift.tt/FydGlD2
Chinese company DapuStor builds high capacity ultra-fast enterprise SSDs
Its 1-DWPD Roealsen6 R6101 7.68TB SSD seriously impressed in a new review
An "engineering masterpiece," the SSD delivered record read speeds
DapuStor is a Chinese start-up specializing in the development and manufacturing of enterprise-grade SSDs - and although you’ve possibly never heard of it, it makes very large - and very fast - storage products.
At the start of 2025, TweakTown tested DapuStor’s J5060 61.44TB SSD against a number of enterprise SSDs, including Solidigm's same size beast, and came away impressed with the drive’s superior read performance, declaring it to be “the most efficient SSD of its capacity point currently in circulation”. A new 122.88TB version of that SSD has been spotted online, and we look forward to seeing how it compares.
Before that, however, TweakTown managed to get its hands on another DapuStor product, the Roealsen6 R6101 7.68TB Enterprise SSD. It may not be anywhere near as big as the J5060, but – spoiler alert – it is incredibly fast.
An engineering masterpiece
Introducing the new drive, TweakTown says, “The new SSD is built on DapuStor's in-house developed DP800 controller and firmware. The new series features a PCIe 5.0 interface and 3D eTLC NAND Flash. Supporting the NVMe 2.0 protocol, it delivers twice the performance of PCIe 4.0 SSDs.”
The 1-DWPD (1 Drive Write Per Day) SSD proved to offer blistering performance in the tests that TweakTown put it through, leading the site to declare the Roealsen6 R6101 7.68TB PCIe Gen5 x4 U.2 SSD to be an “engineering masterpiece”, scoring it 100% for performance, quality, features and overall.
In summing up his findings, TweakTown’s Senior Hardware Editor Jon Coulter gushed, “The drive delivers a record breaking 3.62 million 4K random read IOPS at QD512. This is a whopping 10% more than anything we've encountered previously. Additionally, its 14,600 MB/s sequential read throughput is right up there with the best of them, as is its over 11,000 MB/s sequential write throughput.”
That level of throughput alone would be impressive, but it’s only part of the story.
“Then there is its mixed workload prowess where our 1-DWPD test subject delivers more than anything in its class at queue depths of up to 64. Its mixed workload performance is so good that it can hang in with 3-DWPD SSDs at queue depths of up to 16. And finally, as perfectly illustrated by our preconditioning charts, the R6101 7.68TB delivers QOS that is as good as we've ever seen,” Coulter concluded.
You might also like
These are the fastest SSDs that you can buy right now
It's a form factor that naturally draws quite a bit of ire depending on which side of the great PC gaming form-factor debate you sit. For those that love it, ITX and SFF machines are the stuff of legends. Intricately complex builds, ungodly power, and a real David-vs-Goliath kinda gaming rig.
For the full-tower fans, they're pointless, overly hot, and lack the full connectivity that you'd find in a more robust, balanced chassis and form factor.
For me, I 100% live life in that first category. In fact, I wouldn't have a career in hardware journalism if it weren't for building a machine inside of Bitfenix's now legendary Prodigy ITX gaming chassis from way back when in 2013 and showing it off to PC Format's editor back in the day.
It's not all been glamorous, of course. Motherboards have often been lacking, case design has been pretty terrible at times, and there are all manner of cooling and cabling problems that have needed to be overcome. But I honestly feel like we're in a position now that, technologically at least, ITX is in a place where for the vast majority of power users, there's no major difference between it and a big boy build.
The challenges of building in compact cases
That's the big thing too: a lot of the problems that ITX initially faced, even less than a decade ago, stemmed from how we managed hardware.
For instance, ATX builds once used to be able to house multiple graphics cards running in SLI or Crossfire, but over time that was whittled down to just two cards, until finally, with Nvidia's 30 series, support was removed entirely.
There's no major difference between running an RTX 5090 in an ITX build than there is in a full E-ATX setup, and in fact, you'll notice that pretty much every motherboard now only comes with a single (usually reinforced by some "armor" or gimmick) PCIe x16 slot up top for the graphics card. But just one.
(Image credit: Future / John Loeffler)
Then there's the M.2 conundrum, and again, ITX used to be massively disadvantaged here as well.
Yet similarly, as NAND density has increased and cost decreased, along with some clever and quite intuitive raised PCBs and M.2 slot designs, it's quite easy to find ITX motherboards with two or even three M.2 slots.
Combine that with one of the best SSD you can get your hands on for your OS drive and a nice chunky backup SSD for your secondary storage, and once again, you're already well-equipped to compete with larger mid-tower cases for the vast majority of people.
I could go on, but the fact is that ITX today, from cases to coolers to the hardware we use, is really nowhere near as limited as it used to be. Even processor performance, with auto-turbo galore, isn't exactly held back anymore. But the question remains: Why bother? What's the point? Why do I care so much?
Power in small packages
It matters because ITX represents that defining principle of modern-day technology of the last 30 years. In my lifetime, I've seen tech shrink and become more powerful time and time again. It might be because I'm a short(ish) fellow, but there's something quite admirable I find about a tiny but mighty build.
That Moore's Law-esque power creep, or the sleeper build housing ungodly performance in a form factor that could fit in an entertainment center in a living room. It's like rocking up to a drag race with a seemingly clapped-out VW Caddy Mk1, only to know you've got a 500-horsepower engine under the hood. It has that vibe, and I love it.
(Image credit: Geekom)
Over the years, I've built countless PCs and published hundreds of build logs, pursuing all manner of hardware launches and gimmicky headlines designed to entice the reader in pushing the limit of what's possible with off-the-shelf hardware. And time and time again, the builds that stick with me aren't the crazy $10,000 dream machines or the full-fledged RTX 5090 E-ATX monsters; no, it's the ITX ones.
Don't get me wrong, I'm screaming at them each and every time I build them as I struggle with cable management or liquid-cooling runs, but deep down, I love that form factor more than anything else. And if you haven't given it a shot and are looking for a PC building challenge, and a rig that not only delivers on the performance but also takes up a fraction of the space, I highly recommend giving it a go.
Rubin Ultra GPUs previewed at Nvidia GTC 2025 with Kyber rack mockups
Each NVL576 rack may include 576 GPUs across four internal pods
Projected power draw reaches 600kW with performance targets of 15 EFLOPS
At Nvidia GTC 2025, the company gave a preview of what its future data center hardware could look like, showcasing mockups of its Rubin Ultra GPUs housed in the Kyber-based NVL576 racks.
These systems are expected to launch in the second half of 2027, and while that’s still some way off, Nvidia is already laying the groundwork for what it describes as the next phase of AI infrastructure.
A single NVL576 rack, according to Jensen Huang, co-founder, president, and CEO of Nvidia, could draw up to 600kW. That's five times more than the 120kW used by current Blackwell B200 racks, suggesting a steep rise in power per rack going forward.
Powering the future
Tom’s Hardware reports, "Each Rubin Ultra rack will consist of four 'pods,' each of which will deliver more computational power than an entire Rubin NVL144 rack. Each pod will house 18 blades, and each blade will support up to eight Rubin Ultra GPUs - along with two Vera CPUs, presumably, though that wasn't explicitly stated. That's 176 GPUs per pod, and 576 per rack."
The Kyber rack infrastructure will support these systems, along with upgraded NVLink modules which will have three next-generation NVLink connections each, compared to just two found in existing 1U rack-mount units.
The first Rubin NVL144 systems, launching in 2026, will rely on existing Grace Blackwell infrastructure. Rubin Ultra arrives in 2027 with far more density.
Tom’s Hardware says that the NVL576 racks are planned to deliver “up to 15 EFLOPS of FP4” in 2027, compared to 3.6 EFLOPS from next year's NVL144 racks.
During the GTC 2025 keynote, Jensen Huang said future racks could eventually require full megawatts of power, meaning 600kW may only be a stepping stone.
As power climbs toward the megawatt range, questions are inevitably growing about how future data centers will be powered.
Nuclear energy is one obvious answer - The likes of Amazon, Meta, and Google are part of a consortium that has pledged to triple nuclear output by 2050 (Microsoft and Oracle are notably missing for the moment) and mobile micro nuclear plants are expected to arrive in the 2030s.
Bill Gates has been making the rounds lately to promote his new memoir, Source Code, and sharing his vision of rapid and massive transformation of our lives over the next decade by artificial intelligence. He pitched a world where "Intelligence will be completely free" in an interview with Harvard professor Arthur Brooks, with ubiquitous and universally available AI tutors and doctors that outmatch most educators and medical practitioners.
He went even further in a recent appearance on The Tonight Show. When Jimmy Fallon somewhat nervously asked him if we'll still need humans, Gates quipped, "Not for most things."
Look, I admire Gates’ enthusiasm. He is clearly invested in a future where AI extends opportunity to underserved populations and pushes the limit of what people are capable of, which is great. It's a little absurd to believe that AI will replace most human roles within a decade, though, and doubly so for positions built on human-to-human interaction. That's more than optimism; it drifts into the kind of AI hallucination that limits the adoption of AI in precisely those fields.
AI Limits
For one thing, current AI models aren't completely ready for what he describes. Yes, large language models behind tools like ChatGPT and Gemini are impressive when it comes to mimicking conversation, writing code, and even imitating human painters. But the illusion of competence hides a laundry list of unresolved issues. AI still makes mistakes, sometimes hilarious ones, but it's not so funny when you fail a test or get misdiagnosed.
Anyone who’s spent more than ten minutes with a chatbot has probably watched it veer off into at least some nonsense, whether confidently inventing facts or suggesting you eat rocks. These aren’t just glitches. They’re systemic quirks that stem from the way these models work, using statistical pattern recognition without real understanding.
Even the companies building this stuff are quietly worrying they’re running out of quality training data. Once you’ve consumed the entire publicly available internet, you hit diminishing returns. It’s like trying to get smarter by rereading the same old textbooks; you might sharpen some things, but you won’t have new insights. Without breakthroughs in how we train and structure AI models, we may be closer to an awkward plateau than the exponential curve that Gates's future would require.
Human touch
Even if AI gets way better, it still won’t be human. That’s not just sentimental—it’s functional. So many jobs that Gates suggests AI could “solve” rely on things no machine has: a childhood, a body, a lifetime of subtle emotional calibrations.
Yes, AI is getting better at reading and employing emotional nuance, but I remain skeptical that it could match the above-average human equivalent of a teacher or doctor, let alone the best of them. Could an AI earn a teenager’s trust when they’d rather be literally anywhere else but in a lesson or sit with a patient in pain and make them feel heard? Maybe, but not in ten years.
What makes for competent logistics planning, customer service, human resource management, and so many other roles is the ability to balance human needs, motivations, and unpredictability. AI can help in all of these fields. It already does. It can write reports, crunch numbers, and flag anomalies. In some cases, it can outperform humans. But replacing the entire role suggests that just because an AI can paint in the style of Van Gogh, it could’ve also survived his mental illness, navigated 19th-century Paris, and invented post-impressionism. It’s not just about output—it’s about the messy, lived-in process behind it.
A deep reservoir of subtle, emotional intelligence is baked into any human career involving other humans. Gates seems to think this can be simulated convincingly enough to make no difference. I’m not so sure.
AI suspicions
This brings me to my last point: even if AI could match or beat human performance in nearly everything, it doesn’t mean people will want that. Let’s not forget that we’re a species with many members who enjoy small talk with baristas even when there’s a self-checkout option. Most people value other humans for more than just the mechanical aspect of their profession, especially in areas like medicine, education, and caregiving.
On The Tonight Show, Gates joked that no one wants to watch robots play baseball, and he’s right. But he stops short of realizing that many people won’t want only robots to teach their kids how to play. Not because the robots aren't technically competent, but because we still prefer the flawed but relatable experience of other humans.
Sure, I'd love a hyper-precise machine with a well-trained AI to perform micro-surgery on me, but there had better be a human surgeon overseeing its work and keeping an eye on me beyond the machine's focus.
AI futures
Gates’s suggestions aren't bad ideas if applied correctly. He’s absolutely right that AI can help extend access to critical services in places that don’t currently have enough teachers or doctors. The part of his vision where AI becomes a helpful assistant for everyone, filling gaps and enhancing what humans already do, is something I'd love to see come true. That said, the leap from “AI can help” to “AI will do everything” is a dangerous oversimplification of both technology and humanity.
So yes, AI is going to change the world. It already has. But not in the clean, utopian, humans-on-vacation way that Gates imagines. It’s going to be messier. Slower. Full of unexpected detours and stubbornly human resistance. People often like their teachers and trust their doctors. They might let AI help, but they won’t give up that human touch without a fight.
Microsoft pulled out of a $12bn deal with CoreWeave, citing delays
OpenAI took over the contract, backed by Microsoft’s own investment funds
AI sector remains a closed loop driven by a few dominant players
CoreWeave is eyeing a huge (potentially $2.5 billion) IPO in the coming weeks, but it has also had a few unflattering news stories to contend with recently.
Jeffrey Emanuel, whose viral essay described Nvidia as overpriced and led to it losing $600 billion in a single day, has described CoreWeave as a turkey and called it the “WeWork of AI”.
More recently, Microsoft chose to walk away from a nearly $12 billion option to buy more data-center capacity from the AI hyperscaler.
OpenAI to the rescue
The Financial Times (FT) reported sources familiar with the matter saying Microsoft had withdrawn from some of its agreements “over delivery issues and missed deadlines” which shook the tech giant’s confidence in CoreWeave.
The FT added that despite this, Microsoft still had "a number of ongoing contracts with CoreWeave and it remained an important partner.”
Microsoft is CoreWeave’s biggest customer, and the AI hyperscaler refuted the FT's story, saying “All of our contractual relationships continue as planned – nothing has been cancelled, and no one has walked away from their commitments.”
Shortly after that news broke, it was reported that OpenAI would be taking up Microsoft's nearly $12 billion option instead, helping CoreWeave avoid a potentially embarrassing setback so near to its closely watched IPO.
Rohan Goswami at Semafor made a couple of interesting observations on the news, noting, “This isn’t a sign that Microsoft is pulling back on AI - “We’re good for our $80 billion,” Satya Nadella said on CNBC - but an indication that the company is being more tactical about exactly when and where it spends. At the same time, OpenAI’s biggest backer is Microsoft, meaning that OpenAI is paying CoreWeave with money that is largely Microsoft’s to begin with.”
He described this as the rub, saying, “The AI economy is currently a closed loop and will stay that way until a broader swath of economic actors like big and medium-sized companies start spending real dollars on AI software and services. Until then, nearly all the money is coming from a few companies - chiefly Nvidia and Microsoft - which themselves depend on the goodwill of their public shareholders to keep underwriting it all.”
Apple has confirmed that its Worldwide Developer Conference (WWDC) will take place from June 9 through June 13. The annual symposium of all things Apple software should give us our very first look at Apple’s latest software for the best Apple Watches in 2025, watchOS 12.
Apple has provided annual updates to its wearable software alongside iOS and macOS since the inaugural debut of the Apple Watch in 2015. Many of the features on the Apple Watch that we take for granted in 2025 were only added after the earliest models launched, including small features like Activity Sharing and even the App Store.
With WWDC 2025 on the horizon, watchOS 12 is almost certainly very close at hand. Rumors are scant as to what Apple might add to the software this year, with the biggest Apple Watch headlines for 2025 currently encompassing hardware upgrades to the Apple Watch Series 11 and Ultra 3 rather than software tweaks to the whole lineup.
But that doesn’t mean we can’t make some educated guesses and start to piece together what we know so far. Inevitably, watchOS 12 leaks will probably emerge closer to the time, and we’ll get more rumors between now and the launch. Here’s what we know so far.
watchOS 12: Cut to the chase
What is it? Apple's next wearable software for Apple Watch
When is it out? Expect betas from June, with an official release in September 2025
Who will get it? Likely users of the Apple Watch Series 7 and later, plus Ultra users, and maybe Apple Watch SE 2 and Series 6 owners
watchOS 12: Potential release and unveiling date
(Image credit: Future)
Unless Apple breaks 10 years of watchOS release schedules, the company should unveil watchOS 12 at the WWDC 2025 keynote on June 9. Apple always showcases its latest and greatest software at the opening address on the morning of the first day, so we should see watchOS 12 alongside iOS 19, macOS 16, and iPadOS 19.
So, we have a good idea of when the new software will be unveiled. But what about the release date?
The Apple watchOS 12 beta should go through a series of staggered releases, if Apple continues form with its software rollouts.
Usually, it’ll release a developer beta immediately following WWDC 2025, often on the same day as the keynote announcement. This means we could see it in the wild as early as June 9. For the last two years, Apple has made the developer beta available to all of its customers, so you won't need to be a paying account member to sign up.
Then, a public beta will follow, usually one month later. As the name suggests, this beta is also available to everyone. It tracks about a month behind the developer betas, so it doesn’t quite have as many new features but tends to be more stable as a result.
Finally, after months of testing over the summer, Apple will launch the software publicly for everyone to download. In 2024, this happened on September 16 alongside the launch of iOS 18, right between the pre-order date for the iPhone 16 and Apple Watch Series 10 on September 13 and the official launch on September 20. As such, I’d expect the full release of watchOS 12 to coincide with the launch of the new Apple Watch Series 11, Apple Watch Ultra 3, and Apple Watch SE 3, as well as the iPhone 17.
Apple needs to push the software live before launching the new best Apple Watch models, and it also needs iOS 19 to go live at the same time, because you can’t have an Apple Watch without an iPhone. Generally, then, you can expect watchOS 12 to debut in mid-September.
watchOS 12 expected compatibility: Which Apple Watch models will get the new software?
(Image credit: Future)
While we won't know about compatibility for certain until Apple unveils watchOS 12, we can make a very educated guess on which Apple Watch models will get the new software.
That's because Apple's chip progression and features usually move in lockstep. Let me explain: Each year, Apple adds more features to its watchOS software. More advanced features require more power and put more strain on battery life, which means that, inevitably, Apple drops support for a model or two each year. Last year, Apple watchOS 11 ended support for the Series 4, Series 5, and first-generation SE.
You can expect guaranteed support for watchOS 12 on the Apple Watch Ultra 2 and original Ultra, as well as the existing Series 10, Series 9, Series 8, and probably Series 7. The Series 6 might be on the chopping block, owing to its older S6 chip.
The SE 2 has an S8 chip and is currently still on sale, so I'd expect support for that to persist for a year or two more.
Finally, and most obviously, we can expect watchOS 12 to be compatible with all three new Apple Watch models expected for 2025. Namely, the Ultra 3, Series 11, and SE 3.
If Apple does drop support for a model this year with watchOS 12, the Series 6 seems like the most obvious candidate because of its age.
watchOS 12: Rumored design, new features, and what we want to see
(Image credit: Future)
While we've heard that iOS 19 is set to usher in a major redesign and there are big plans for macOS and iPad OS, too, rumors about any new features or design changes for watchOS 12 are scant so far.
If Apple is planning a major overhaul for its other software platforms, that could indicate one of two things. Either, Apple will overhaul watchOS 12 to reflect the revised look and design of its other software, or watchOS 12 will be a very muted update because it's spending so much time on its other software.
Obviously, we'd prefer if the former was the case, but there are a couple of other bits and pieces we'd like to see, as well as some potential updates we can guess at.
If it's exclusive to the new models, a mention is unlikely at WWDC, as Apple will reserve the announcement for the September iPhone event instead.
Apple AI health coach
(Image credit: Future)
Apple has reportedly been working on an Apple AI Health Coach app for at least two years. Bloomberg's Mark Gurman has previously reported the company will revamp its health app, including a new AI-based coaching service. Gurman reported in January that Apple is "increasing its ambitions in the health realm" and that work continues on the app.
Any upgrade in this department is likely to include a significant Apple Watch component, so a reveal of this progress at WWDC 2025 for watchOS 12 isn't out of the question.
Apple Intelligence
Apple Intelligence remains a glaring absence in watchOS. For better or worse, Apple has intentionally left its AI features out of its wearables lineup, reserving them for iPhone, iPad, and Mac.
There's some sense to this. Apple Intelligence is generally power-hungry, and would impact Apple Watch battery life and possibly performance. However, I think there are a few obvious integrations that we could see in the next watchOS iteration:
Genmoji - Genmoji lets you create custom emojis on iPhone that you can send to your friends and family, and it could be a fun way to improve communication on Apple Watch.
Notification Summaries - An Apple Watch that could give you AI-powered summaries of important notifications could be an excellent upgrade, especially on a device with a small screen designed to be glanced at quickly throughout the day.
Priority notifications - Apple Watch notifications can be overwhelming; using Apple's priority notifications could significantly improve how many notifications get pushed to your Apple Watch.
Smart Reply - Apple Intelligence can write replies to messages. With the Apple Watch's tiny screen and fiddly keyboard, this one could be a lifesaver.
Siri upgrade - Either through ChatGPT or Apple's own AI improvements, Siri needs a desperate upgrade, and it would be great to get the ChatGPT integration in Siri on Apple Watch while we wait for the 2026 upgrade.
watchOS: 2024's most significant updates
There's a good chance that watchOS 12 includes some tweaks and iterations of existing features, as well as new ones. Headline features from last year's release include the new Vitals app, which gives you health metrics measured during your sleep.
There was also the new Training Load feature, customization in the Activity Rings and Fitness app, the new Smart Stack, improved pregnancy support, and new safety tools, including Check In. All of these newer features are prime candidates for further upgrades and changes, and my prediction is that we'll get an upgrade for the Vitals app.
It's almost time to bid farewell to March (how are we a third of the way through 2025 already!?). But, before April *ahem* springs into action, there are a number of eagerly-anticipated new movies and TV shows to enjoy on the world's best streaming services.
So, whether you're looking forward to watching Mufasa: The Lion King, The Studio, or something else at home, we've got you covered. Whatever you stream this weekend, we're sure you'll enjoy it. – Tom Power, senior entertainment reporter
Mufasa: The Lion King (Disney+)
The latest Lion King live-action adaptation has arrived on Disney+ if you're looking for some escapism to the jungle this weekend. This one divided the critics with a split 56% Rotten Tomatoes score, which is actually higher than its 2019 predecessor that only scored 51%. So fans were more enamored with this one, which focuses on the story of Mufasa as he meets a sympathetic lion named Taka, who is heir to a royal bloodline.
We're all familiar with the animated classic, and Mufasa: The Lion King takes us to where it all began, swapping between timelines to tell the story of Simba's father and the events leading up to the beloved original movie. While it's not good enough for our best Disney+ movies round-up, you might still have fun with it anyway, especially if you're looking to entertain the whole family. – Lucy Buglass, senior entertainment writer
Based on the novel of the same name by Lori Nelson, this new Netflix movie follows Alex Rose (Sofia Carson) as she embarks on a quest to complete her childhood bucket list following the death of her mother. As she revisits her youth, she’s taken on a surprising journey of self-discovery, romance and family secrets.
Netflix is home to many rom-com flops like A Family Affair and Find Me Falling, but The Life List promises to "make you both laugh and cry", which is usually a winning formula for some of the best Netflix movies. I’m definitely getting the tissues ready.– Grace Morris, entertainment writer
With Severance season 2 firmly in the rear view mirror, Apple needs another outrageously brilliant TV Original to fill the void. While Surface season 2 and Dope Thief are worth watching, it's The Studio that I'm recommending as your next must-see show on Apple TV+.
A clever cringe comedy series that doesn't hold back in its satirical take on Hollywood and the entertainment industry more broadly, The Studio is an absolute riot. Seth Rogen stars as Matt Remick, the newly-appointed CEO of fictional company Continental Studios whose attempts to save the floundering business leads to all manner of slapstick moments, belly-aching laughs, and not-so-subtle digs at celebrityism, the LA party scene, and the churn of the Hollywood machine.
With top-tier performances from its supporting cast, including Kathryn Hahn and Catherine O'Hara, plus a whole host of cameo appearances – Martin Scorsese and Netflix CEO Ted Sarandos are in it, for goodness sake! – The Studio is absolutely unmissable. Easily one of the best Apple TV+ shows of all-time. – TP
Who else has been excitedly waiting to see Timothée Chalamet's take on the iconic role of singer-songwriter Bob Dylan? Despite not winning a single Oscar after being nominated for eight at the 97th Academy Awards, A Complete Unknown remains one of the most critically-acclaimed box office releases in the past year and will no doubt be one of the best Hulu movies once more audiences get a chance to stream it this weekend.
Set in the early 60s, the docudrama follows Dylan's rise to fame in New York under the mentorship of Pete Seeger (Edward Norton) and offers a new dramatization of the musician's life, including his relationship with fellow singer-songwriter Joan Baez (Monica Barbaro). Chalamet sings all the classic songs in the film himself, so expect a fresh take on favorites like Mr Tambourine Man and Like A Rolling Stone. – Amelia Schwanke, senior entertainment editor
After the Nicole Kidman-led Netflix series The Perfect Couple disappointed me recently, I’m really hoping Holland will be a return to form for the iconic actress. The plot is intriguing enough, but with a 30% Rotten Tomatoes score from the critics, there’s a real chance I’ll be let down again when I watch it this weekend. Having said that, critics' scores don’t guarantee the audience will hate it too, so we’ll have to make our own minds up.
There’s a great cast here, at least, with Kidman joined by Matthew Macfadyen, Naomi Watts, and Bryan Cranston, so the movie is certainly not lacking when it comes to talent. Holland is one to try if you’re hoping for a mystery-thriller to get stuck into, it’s now streaming on Prime Video. Maybe you’ll like it more than you think, there’s only one way to find out! - LB
Luca Guadagnino's latest film sees the Italian director back in his element. Since the theatrical release of Queer in 2024, many can't stop talking about it, including TechRadar's Rowan Davies who said it was his favorite movie of 2025 so far, and I'm not surprised following the success of Challengers and Call Me By Your Name. The director has a way with romantic dramas.
Adapted from a novel by William S. Burroughs of the same name, Queer is set in 1950s Mexico City and tells the story of an expat's (Daniel Craig) relationship with a young student (Drew Starkey). Critics haven't been overly taken by the new movie, which means it won't be one of the best Max movies, but that doesn't mean that audiences won't enjoy streaming another great A24 movie. – AS
MobLand is another TV crime drama created by the king of the gangster genre Guy Ritchie and features a star-studded cast of Tom Hardy, Helen Mirren and Pierce Brosnan. In this new Paramount+ show, two London crime families battle for power that threatens to end empires and ruin lives. Caught in the crossfire is Harry Da Souza (Hardy), a street-smart ‘fixer’ who is called upon by the Harrigans to protect their family.
This potential best Paramount+ show will be the perfect alternative while I wait for season 2 of Ritchie’s hit Netflix series The Gentlemen and I can’t wait to see his signature violence and unique characters back on the small screen in another gangland saga. – GM
HP ZBook Ultra G1a comes with a massive 126GB RAM and 2TB SSD
Preorders are available now, priced between $3,797 and $4,049 at US retailers
ZBook Ultra G1a is built for AI, data analysis, and video production
HP has revealed the ZBook Ultra G1a, a mobile workstation that features a 16-core AMD Ryzen AI Max+ Pro 395 processor with a 3GHz base clock and 5.1GHz boost speed.
The new device will handle professionals' AI applications, large datasets, and high-resolution content, with integrated AMD Radeon 8060S graphics ensuring smooth performance for intensive tasks like video editing, while providing real-time power for LLM execution, rendering, and data analysis.
Just like the HP EliteBook X G1a, it's equipped with 128GB of LPDDR5X RAM and a 2TB PCIe NVMe large SSD, delivering one of the fastest computing experiences for demanding AI workloads and high-performance computing.
Performance-driven workstation with AI capabilities
The business laptop measures just 0.7 inches thick and weighing 1.5 kg, while its 14-inch OLED touchscreen offers a 2880 x 1800 resolution and a 120Hz refresh rate, delivering sharp visuals with deep contrast.
There are two Thunderbolt-enabled USB-C ports, an additional USB-C 3.2 Gen 2 port, a USB-A 3.2 Gen 2 port, HDMI 2.1, and a 3.5mm audio jack, plus Wi-Fi 7 and Bluetooth 5.4.
It's powered by a 74.5WHr four-cell battery for extended on-the-go usage and includes a 140W USB-C power adapter for fast recharging.
The HP ZBook Ultra G1a is now available for preorder across multiple US retailers, though the exact cost varies.
While B&H Photo Video lists it as high as $4,049.00, Directdial offers it for as low as $3,797 - either way, it's for those who don't mind paying a premium for top-tier performance.
'Signalgate' continues to rumble on, with even more of the Signal messages that were shared between Trump officials – and, inadvertently, The Atlantic – revealed today. But how exactly did this group chat debacle happen, and what does it say about Signal? We've answered all of this and more in our one-stop explainer about the app that's in the eye of a political storm.
With robust privacy credentials, Signal has long set the standard for secure messaging. But in the wake of those revelations that US government officials inadvertently added a journalist to a group chat where confidential military plans were discussed, there are fresh (and mostly unfair) question marks over how reliable the app’s protections really are.
Signal’s reputation has made it popular with journalists, politicians and privacy advocates. Used correctly on an Android or iOS device, it can absolutely make your conversations more secure. Even so, it’s not a tool designed for sharing classified information.
So why have senior US politicians been using it for sensitive national security communications? And how safe were those messages? Here’s what you need to know about Signal – and how to use it properly for private messaging.
What is Signal?
Signal is a secure messaging app with end-to-end encryption
It’s operated by the Signal Foundation, a non-profit organization
Open source code makes the platform more resilient
Signal is a messaging app that's available for iOS and Android devices. Like the best messaging apps, it supports cross-platform text, voice and video chats. What sets Signal apart is its robust privacy features: it’s regarded as the benchmark for secure mobile communications. Which makes Signalgate all the more ironic, even though it has little to do with the app's technical security.
End-to-end encryption ensures messages can only be read by the sender and receiver, while open-source code ensures that there are fewer vulnerabilities for hackers to exploit.
(Image credit: Shutterstock / BigTuneOnline)
Signal was created in 2012 by Moxie Marlinspike. It’s now run by the Signal Foundation, a non-profit organization founded in 2018 by Marlinspike and WhatsApp co-founder Brian Acton. The Foundation relies on donations rather than ad revenue to fund its services. That means users can enjoy an ad-free and tracker-free experience, knowing that their data won’t be sold to third parties.
In our in-depth Signal review, we noted that the app offers “fewer bells and whistles than the more popular (and less secure) messaging apps”. Instead, its developers have focused on creating a secure, minimalist communications tool.
It’s this approach which has made it popular with everyone from whistleblowers and activists to journalists and privacy advocates, who favor its more secure architecture.
How safe and secure is Signal?
Used correctly, Signal is the most secure messaging app
It’s only as secure as the devices sending and receiving messages
User error can compromise the privacy of group chats
Used correctly, Signal has the most robust privacy credentials of any major messaging app. It has the most layers of security at both the front-end and back-end. Messages themselves are deeply protected against hacking, while the app offers a toolkit to ensure communications are only seen by who they’re intended for.
So how did a journalist from The Atlantic end up on a group chat with US government officials, including Vice-President JD Vance and White House chief of staff Susie Wiles? Signal CEO Meredith Whittaker maintains that the app is “the gold standard” in private communication. Here’s the challenge: as strong as Signal’s security features are, they are reliant on the end user understanding how they work.
This incident was not really a failing of Signal. The journalist in question did not infiltrate the group chat through a backdoor. Instead, a member of that chat – which included 18 people – unintentionally but actively added the reporter to the group, who was then privy to sensitive messages discussing air strikes in Yemen.
(Image credit: Shutterstock / Shutterstockies)
Any group chat is only as secure as its members. Even with disappearing messages enabled, there is a window in which anyone in that group can read them. By mistakenly adding the journalist to the group, the US official became responsible for compromising its integrity.
Some voices have criticized the fact that the app allowed this, but the option to add a contact to a chat is a core function of group communications. The fault here doesn’t lie with Signal – the reality is that, as secure as the app is, it’s not an appropriate platform for sharing highly confidential state information.
What’s more, messages sent on Signal are only as secure as the device that is receiving them. If a smartphone is compromised or left unlocked, all of the Signal messages on that device can be read. There is also nothing to prevent someone simply reading messages over your shoulder.
Matthew Mittelsteadt, a technology policy research fellow for the Cato Institute, said as much in a statement emailed to CNN. “Messages may be secure when they are in transit between phones, but once they reach the recipient, security can indeed fail.”
(Image credit: Shutterstock / Boumen Japet)
This is why the use of Signal by top US officials fell far short of government security protocols. Data expert Caro Robson, quoted by the BBC, said communications like these would usually take place on “a very secure government system that is operated and owned by the government using very high levels of encryption.”
While officials from the Trump administration have claimed that none of the information shared was classified, a memo from the Defense Department circulated in 2023 and obtained by NPR banned the use of mobile apps for "controlled unclassified information”. Since the leak, the Pentagon has issued an advisory prohibiting the use of Signal even for “unclassified information”.
National Security Adviser Mike Waltz has acknowledged the failing. Speaking to Fox News, he described it as “embarrassing” and took “full responsibility”.
How does Signal work?
Signal uses open-source, end-to-end encryption to secure messages
The Signal Foundation doesn’t monetize user data or sell ads
User features are designed for security, including personal PINs
Signal offers greater security in three key ways. The first is through end-to-end encryption, which means messages are scrambled in transit, then decoded when delivered to the intended device. No-one else can read these messages, not even Signal.
While other messaging apps also offer end-to-end encryption, Signal’s is more secure because it’s open source. Not only does this make the platform more transparent, but it also allows absolutely anyone to examine the code for potential vulnerabilities. This community scrutiny makes it more likely that problems are found and fixed before hackers can exploit them.
(Image credit: Signal)
Then there are the Signal Foundation’s principles. The app itself collects less user data than other services, with message history stored on user devices rather than Signal’s servers. Of the very limited information that is collected by Signal, none of its is monetized: as a non-profit, the Foundation relies on donations rather than advertising revenue. That also means users won’t be tracked or encounter ads on the platform.
Finally, there are the safety features integral to the user experience. These include a personal PIN to secure your profile and the option to hide your phone number. Every one-to-one Signal chat also has a safety number, which can be used to verify that you’re communicating with the right person. In addition, you can’t be added to a group chat without giving your express approval.
How to get started with Signal
The Signal app is available for iOS and Android devices
Setup requires a phone number to receive a verification call or text
Privacy features include personal PIN numbers and disappearing messages
Getting started with Signal is pretty easy – the app is free to download from the App Store for iOS devices and the Google Play Store for Android smartphones. To create an account, you need a phone number which will be verified by phone call or text message. Once you’ve set up your account, your number will be hidden from other Signal users by default (see below).
The app's interface and basic functions will be familiar to anyone who has used a messaging app such as WhatsApp, Messenger or Telegram. Tap the pencil icon to start a one-on-one or group chat. Within a chat, you can share messages, photos and voice notes. You can also tap the phone or camera icons to start voice or video calls.
If you’re keen to secure your messages, there are a few more advanced features to explore and configure. A Signal PIN can be used to recover your profile and settings on a different device. It’s configured by going to Signal Settings, tapping Account then selecting Change your pin.
Signal also encourages users to verify Safety Numbers. These are generated for every one-on-one chat to confirm that you’re sending messages to the right person. To view a Safety Number, open a chat, tap the header and select View Safety Number. To verify it, you would ideally compare numbers with the recipient in person. Otherwise, you can share it using a trusted channel.
(Image credit: Signal)
Disappearing messages add an additional layer of privacy. After a set period of time, the contents of messages are no longer visible, whether they’ve been read or not. You can set a default timer by going to Signal Settings > Privacy > Default timer for new chats. You can also configure timers for specific chats. Just go to chat settings and select Disappearing messages.
Signal also offers features to keep your communications activity hidden. Screen Security stops a preview of Signal appearing when you switch apps. You can enable it by heading to Settings > Privacy and selecting Screen Security on Android or Hide Screen in App Switcher on iOS.
In addition, you can hide Signal calls from your device’s call log. This is enabled by default, but to double check, head to Signal Settings > Privacy and look for Show Calls in Recents.
Finally, to manage the visibility of your phone number, go to Signal Settings > Privacy > Phone Number and tap 'Who can find me by my number'. To set up a unique username that you can use instead of your number, go to Signal Settings > Profile.
Even with all of the above features enabled, remember that your Signal communications are only as secure as your smartphone itself. To avoid a leak of information like the US government suffered, be sure to activate your device’s full set of security features, keep it locked with a passcode when not in use and don’t access sensitive messages in public. For more tips on how to keep your phone safe, read our dedicated feature here.
