TechRadar News

• New report claims OpenAI has detected evidence of distillation by DeepSeek
• Move represents a potential breach of intellectual property
• Whitehouse AI czar weighs-in on the subject

According to a new article by the Financial Times, OpenAI claims to have evidence that DeepSeek, the Chinese startup that has thrown the US tech market into financial turmoil, used the company's proprietary models to train its own open-source LLM, called R1. This would represent a potential breach of intellectual property, as it goes against the OpenAI terms of service agreement.

In the article the FT writes that a source at OpenAI claims it has evidence of “distillation” occurring, which is a technique used by developers to leapfrog on the work done by larger models to achieve similar results at a much lower cost.

The OpenAI terms of service clearly state that users cannot copy any of its service or “use output to develop models that compete with OpenAI.” David Sacks, the Whitehouse crypto and AI “czar” said in an interview on Fox that there is “substantial evidence” of distillation occurring from DeepSeek.

OpenAI statement

Speaking to TechRadar an OpenAI spokesperson said: "We know PRC based companies – and others – are constantly trying to distill the models of leading US AI companies. As the leading builder of AI, we engage in countermeasures to protect our IP, including a careful process for which frontier capabilities to include in released models, and believe as we go forward that it is critically important that we are working closely with the U.S. government to best protect the most capable models from efforts by adversaries and competitors to take US technology.”

OpenAI tells us it has observed and investigated attempts to distill its models and has responded through banning the accounts in question and revoking access.

Security concerns

Meanwhile security concerns still seem to be dogging DeepSeek, particularly around the security of user data, exactly what data is being collected, and where it is storing it.

If you or your company has issues with data being stored in China, Perplexity, the AI search engine, is now offering its Pro users access to DeepSeek using data that is only stored on servers in the US.

New registrations for DeepSeek are still temporarily paused, “due to large-scale malicious attacks on DeepSeek's services”. For the latest news on this big breaking story, see the our DeepSeek live blog.

You may also like

• Is DeepSeek AI safe? Or is it just a data minefield waiting to blow up?
• DeepSeek has already dethroned ChatGPT – is it coming for Midjourney and DALL-E next?
• DeepSeek is the new AI chatbot that has the world talking – I pitted it against ChatGPT to see which is best

• Academic researchers found two new speculative execution flaws
• The pair are affecting M2 and M3 processors
• Apple has acknowledged the flaws, and said it would fix it

Apple devices powered with the M2/A15 and M3/A17 chips are vulnerable to side-channel flaws which could put user data at risk of being stolen, experts have warned.

Cybersecurity researchers from the Georgia Institute of Technology and Ruhr University Bochum, who recently published two separate papers, detailing the two vulnerabilities called FLOP and SLAP.

These flaws, however, don’t affect power consumption patterns during cryptographic operations, but rather speculative execution, similar to what the dreaded Spectre and Meltdown vulnerabilities were. Speculative execution is a technique used by processors to improve performance. It involves the CPU guessing the likely path of a program (like which instruction will be executed next) and starting to execute it before the actual decision is made. If the guess is correct, it speeds up processing; if not, the incorrect results are discarded.

Practical application

Explaining their findings to BleepingComputer, the researchers said mispredictions can lead to chips performing computations with the wrong data.

"Starting with the M3/A17 generation, they attempt to predict the data value that will be returned from memory. However, mispredictions in these mechanisms can result in arbitrary computations being performed on out-of-bounds data or wrong data values," they said.

Usually, when academic researchers find computer bugs, they are mostly theoretical, or otherwise extremely difficult to pull off in a real-life scenario. For these, however, the researchers explained how a threat actor could create a malicious website, containing JavaScript code, and use it to pull personally identifiable information from the victims.

They shared their findings with Apple (in late March for SLAP, and in early September for FLOP), who acknowledged their findings and confirmed it would be working on a fix. However, it seems that the Cupertino behemoth won’t be rushing, since it doesn’t think the bugs are that big of a deal

"We want to thank the researchers for their collaboration as this proof of concept advances our understanding of these types of threats," Apple told BleepingComputer.

"Based on our analysis, we do not believe this issue poses an immediate risk to our users."

Those interested in technical details can read the in-depth analysis here. These are the same researchers that discovered the iLeakage vulnerability a year and a half ago, BleepingComputer reminds. That one, too, was a side-channel flaw.

You might also like

• Google and Microsoft find new strain of Spectre and Meltdown
• Here's a list of the best firewalls around today
• These are the best endpoint security tools right now

• iPhone SE 4 no longer tipped to get a Dynamic Island
• Instead it will more closely match the iPhone 14's design
• It will, however, get an Action button per recent leaks

A new model in Apple’s budget-friendly iPhone SE line, the iPhone SE 4, is expected to land later this year, and leaks are already teasing key design details – and the latest reports suggest that the SE 4 will miss out on the Dynamic Island, but could get the Action Button.

Starting with the bad news, rumors had suggested that the iPhone SE 4 would feature the Dynamic Island, which users of all iPhone 15 and 16 handsets, and the iPhone 14 Pro and Pro Max, have been enjoying (or not, as the case is for our editor-at-large Lance Ulanoff).

However, that now doesn’t seem set to be the case. We reported a couple of days ago that a new leaked video had teased the return of the notch on the upcoming SE, and now display analyst Ross Young has claimed on social media that the notch is indeed returning.

The Dynamic Island reports came from another usually reliable leaker in Evan Blass, and these more trustworthy tipsters typically don’t contradict each other – especially not in the run-up to an expected product launch. However a third leaker – who goes by Majin Bu online – has explained that the confusion comes from Apple testing an SE 4 with a Dynamic Island during development before ultimately settling on the notch.

Push the button

Bu has also been sharing images and video said to show the iPhone SE 4, and in his most recent post on X he claims the phone will get an upgrade from a 4.7-inch screen to a 6.06-inch 60Hz OLED display (so with the notch it should look a lot like the iPhone 14), and the Action Button.

Having debuted on the iPhone 15 Pro and Pro Max before making its way to all the iPhone 16 models, the Action Button is a programmable button which replaces the silence mode switch the featured on older models.

The iPhone 15 Pro Action Button (Image credit: Apple)

You can still use it to silence your phone in a hurry, but if you leave your iPhone’s ringer off anyway you can change the button to instead let you access other tools like your camera or torch – and using shortcuts you can even make the button do multiple different things based on how and when you use it.

As with all leaks we should take these notch and Action Button reports with a pinch of salt. We’ve already seen the leaked design of the iPhone SE 4 change, so there’s always a chance that it’ll change again before Apple makes the device official.

That said, based on everything we've seen and heard, the iPhone SE 4 is shaping up to be a solid smartphone with the same A18 chip found in the iPhone 16 and 8GB of RAM – both of which suggest Apple Intelligence support.

You might also like

• I think the iPhone SE 4 could break new ground for Apple
• Read our full iPhone SE (2022) review
• The iPhone SE 4 is reportedly almost ready

• Logitech Spot monitors an office environment for better flow
• Poor air quality could be costing your workers' productivity
• The sensors are also built into the $7,000 Rally Board 65

Logitech has unveiled a brand-new office monitoring tool it hopes will help businesses manage their collaborative workspace better.

The new Logitech Spot tracks occupancy and environmental changes through CO2 levels, air quality, and temperature to provide real-time data for office admins.

Spot is a battery-powered device that uses a peel-and-stick base to stick to walls. It pairs via Bluetooth to CollabOS devices, like Tap Scheduler, or via Long Range Wide Area Network (LoRaWAN) gateways.

Office-tracking sensor

It’s all designed with the hope of making the office a more comfortable space for workers to spend their time and better employee experience – actionable suggestions allow admins to improve the workplace while making reductions to their energy costs.

The company says the real reason workers are getting weary in the office might not be because they’re tired, but rather than inefficient use of the space available is causing air quality issues.

“In bustling workplaces, employees move in and out of scheduled or ad-hoc meetings in spaces that continuously use energy and recirculate air,” Logitech says, citing Harvard and Berkeley studies showing high CO2 levels and pollutants like dust, perfume, and aerosols can reduce cognitive function.

“Logitech Spot solves the mystery of the number of people using spaces, the quality of the air they’re breathing, and environmental factors that can impact energy consumption–arming businesses with the information they need to address these unseen barriers," noted Logitech for Business COO and GM Prakash Arunkundrum.

In the name of interoperability, Spot is also integrated with Microsoft Places and Teams, as well as other platforms like Samsung Smart Things Pro and Zoom Workplace.

Logitech has also lifted the wraps off its Rally Board 65, a new 65-inch portable all-in-one video conferencing solution that integrates Spot’s sensors within its system.

Spot will be available from the second half of 2025, and its APIs can be unlocked via the $199/year Essential or $399/year Select service plans (priced per room). The Rally Board 65 will be available from May 2025, priced at $6,999.

You might also like

• Thousands of UK workers might now get a four day work week
• We’ve listed the best productivity tools and online collaboration tools
• Want a hybrid role? Check out the best job sites

• Accusations have been made around RTX 5090 review samples
• Some folks think GPUs labeled as ‘Press Build’ could be faster than retail products
• Nvidia has assured us this isn’t the case

Nvidia has denied that the RTX 5090 graphics cards sent out to press for review were in any way different to the flagship boards that’ll be sold by retailers, following online accusations that the review samples were somehow ‘binned.’

As Tom’s Hardware reports, there are markings on some review models of the RTX 5090 which say ‘Press Build’ (as seen at TechPowerUp, for example), something remarked upon by Andreas Schilling (editor at German tech site Hardwareluxx) on X.

Ian Cutress (of TechTechPotato) replied to Schilling’s comment about whether there was anything ‘special’ about these graphics cards by jokily altering the words on the RTX 5090 chip to read ‘Press Binned’ as you can see below (if you expand the post).

FTFY pic.twitter.com/i9OtTpDYBCJanuary 24, 2025

Binning refers not to throwing graphics chips in the trashcan, of course, but to the process of allocating silicon to different models of GPU (in ‘bins’ or groups). Every chip in a particular group that pertains to a certain GPU model conforms to the base level spec (silicon that doesn’t is usually repurposed in a lower bin), but some chips are actually of a slightly higher quality, and can be pushed to faster than base clock speeds.

This is why you get different levels of overclock with a certain GPU, or indeed a CPU, with more headroom in a higher-quality chip (even though it’s in the same bin) – which is referred to as ‘winning the silicon lottery’ (you got a good overclocker, in short).

So, without getting too deep into the weeds on this, the insinuation made by some folks online is that Nvidia has ensured that those better-performing chips from the top-end of the bin, as it were, have been used in review samples to get the best results for the RTX 5090.

Tom’s directly asked Nvidia if these press build RTX 5090 Founders Edition cards were “higher-performing iterations of retail counterparts” and a rep from Team Green replied to say: “Some early GeForce RTX GPUs include a top mark related to intended use [the ‘Press Build’ stamp]. Their functionality and performance are identical to retail GPUs.”

In short, the marking means nothing save for indicating that these graphics cards are models put aside for sending out to the press.

(Image credit: Future / John Loeffler) Analysis: An unusual clarification, but clearly Nvidia felt it necessary

This is an odd one, and it’s unusual that Nvidia would respond to such chatter, but it’s clear Team Green felt the need to set the record straight here.

We have to take Nvidia’s assertion at face value, but of course, the difference that any such binning might make wouldn’t be that huge anyway. But even if it was, that’d be a silly route for Nvidia to take, as reviews of third-party RTX 5090 boards compared to the Founders Edition would surely look odd if there was an obvious discrepancy. Yes, Nvidia’s own boards may be well-designed and cooled, but you can be sure third-party models with more premium price tags will be, as well.

I believe that Nvidia is being up front here, and any markings are just for admin or legal purposes (guarding against reselling review boards for example). Indeed, if there was any binning going on with RTX 5090 reviews, it would surely make more sense for Nvidia not to identify those chips as ‘press samples’ so explicitly.

There are more pressing concerns on the RTX 5090, for sure, including whether you’ll actually be able to buy one on launch day (tomorrow, January 30) – as that isn’t the case with MSI’s online store, we’ve just discovered. While you will be able to pre-order an MSI RTX 5090 model tomorrow, it won’t be shipped until February 6, we’re told – and Nvidia is warning on potential stock shortages of the RTX 5090 and RTX 5080. This is a topic that’s been buzzing on the rumor mill for some time now, with a whole lot of pessimistic noises being made, unfortunately.

You might also like...

• Nvidia's RTX 5080 might be only slightly faster than the RTX 4080 Super without DLSS, if this leaked review is genuine
• The end of the Nvidia RTX 4000 series is nigh – with the RTX 4070 expected to sell out in weeks
• Nvidia releases stats that prove DLSS and Frame Generation are here to stay - sorry, angry gamers

• Severance season 2 has dropped an intriguing new teaser for episode 3
• It refers to the mysterious baby goats from season 1
• Episode 3 arrives on Apple TV Plus on Friday, January 31

Severance is the master of suspense. It's really making good use of that Apple TV Plus weekly release by relentlessly teasing fans, and its most recent drop has suggested we'll see more of the goats in Friday's episode. I'm hoping that one of the best Apple TV Plus shows finally answers some questions about why Lumon is raising these animals, I'd even settle for crumbs right now. I need to know what's going on!

Truthfully, I have never been so obsessed with goats since that week when I unexpectedly became interested in the video game Goat Simulator. Now, I'm writing about five of the best Severance season 2 goat theories I've seen. You've got to hand it to this series, it knows how to reel you in.

The clip doesn't give us much, admittedly, but it's worth watching if you want to start coming up with your own theories ahead of episode 3's release on January 31.

What's happened in Severance season 2 so far?

We're only two episodes into Severance season 2 and it feels like so much has happened. If you want some in-depth recaps we've put together a Severance season 2 episode 1 recap and a Severance season 2 episode 2 recap to bring you up to speed. I expected nothing less from this show, but they've opened up on a high and left us with plenty of questions, hopefully paving the way for the rest of the season.

Many fans were shocked when the Emmy-winning title sequence had changed, but it's somehow even better than its predecessor and leans very heavily into body horror, something I couldn't be more pleased about. If you're interested in that, check out the five major clues the Severance title sequence gives us.

There's plenty more to come from Severance and I can't wait to see how deep the rabbit hole goes.

You might also like

• 'He's got a lot of work to do': Severance actor Tramell Tillman teases what lies in store for Seth Milchick in season 2 of the hit Apple TV show
• 4 shows similar to Severance with over 80% on Rotten Tomatoes that you should stream
• Severance creator Dan Erickson reveals the funniest fan theory he's heard about the popular Apple TV Plus show: 'I think that's ridiculous'

• The latest generation of Supercharger stalls can now deliver up to 325kW
• But the Cybertruck is the only Tesla vehicle capable of those speeds
• Full V4 Supercharger stations will soon be capable of 500kW

Tesla has announced that all V4 Supercharger stations in North America are now capable of delivering charging speeds of up to 325kW, bringing the network more in line with high-powered chargers from the likes of BP, ChargePoint and Electrify America.

The increase of 75kW will reduce the amount of time waiting around to charge, but the only model currently capable of accepting such a rate of electrons is the Cybertruck, leaving the vast majority of Tesla owners having to make do with 250kW.

On top of this, there is still lots of confusion surrounding the capabilities of Tesla’s Supercharger network in general, which still includes some of the earlier V2 stalls, as well as a mix of V4 stalls (the physical thing you plug into) that are linked to older V3 cabinets (the important part that handles the delivery of electricity).

That means Model 3, Model S and Model Y owners, along with not being able to make the most of the recent upgrade in charging speeds, experience varying rates depending on the technology available at a destination.

Despite this, Tesla says that it is already rolling out the latest generation V4 Supercharger stalls with accompanying cabinets that it claims are capable of delivering a whopping 500kW, despite there not being a single EV publicly on sale today that can draw that sort of power.

Right now, 800V electrical architecture and today’s battery technology is capable of speeds of 350kW to 450kW in the most extreme cases, with the likes of the Lotus Eletre only managing those headline figures when paired with a specific Lotus charging stall, of which there are very few.

Analysis: Tesla needs to play catch up

(Image credit: Tesla)

The news of faster Tesla Supercharger stalls is going to be welcomed by EV owners, but ironically, not by the wider Tesla community itself.

Most of today's Tesla line-up can only charge at a rate of 250kW, which is rapidly being eclipsed by rival manufacturers that have recently made the switch to the North American Charging Standard (NACS) – or the plug that Tesla uses, to you and I.

Granted, Musk's latest move is helping to future-proof the Supercharger network, which is often considered among the best charging networks in the world, but it is also gradually becoming more relevant to owners of rival EVs.

Even the recently updated Model Y still has to make do with the old 400V electrical architecture, limiting the speed at which it can charge, despite tweaks to software helping to reduce charging times slightly over the previous generation.

With Cybertruck sales on the wane and Elon Musk firing most of the team that looks after the Supercharger network, the V4 Supercharger stall announcement is likely not something to get too excited about, seeing as the roll-out is predicted to be slow and Musk has made no mention of fitting the Cybertruck’s 800V system to other models just yet.

You might also like

• Tesla's new Cybercab and futuristic 20-seat Robovan show it's lost its grip on reality
• A recent Android Auto update has messed with Google Maps, and users aren’t happy about it
• Hackers expose serious Subaru security flaws that allow them to remotely start cars

About a week after its release, Italy became the first country to formally request a probe into how DeepSeek's chatbot handles users' personal data.

"Possible risk to the data of millions of people in Italy," wrote the Italian data watchdog authority, known as the Garante, on Tuesday, January 28, 2025, in an official announcement.

Developed by a little-known Chinese startup, DeepSeek has seen almost 3 million downloads since its launch, reaching an estimated total user base of between 5-6 million users worldwide, according to SEO.AI data. Yet, experts in and out of Europe have warned of the potential privacy implications.

The companies that provide the Chinese ChatGPT rival – Hangzhou DeepSeek Artificial Intelligence and Beijing DeepSeek Artificial Intelligence – now have 20 days to respond to Italy's complaint.

Against GDPR rules

The Garante's complaint follows a report filed by Euroconsumers – a coalition of five national consumer organizations that also include its Italian member, Altrocunsumo.

After looking at how the DeepSeek app and chat services collect and handle users' data as per its own privacy policy, Euroconusmers found "multiple violations of European and national data protection regulations," the report reads.

These include transfers of European data to China without safeguards as required by GDPR rules, unclear information on how user data is used for online profiling, non-transparent information over data retention periods, and the lack of details on age verification or handling data of minors, among other things.

The Garante is also asking the companies behind DeepSeek AI to provide the clarifications needed for the services to legally operate in the country.

Italy's data watchdog seeks more information on the type of data used to train the artificial intelligence system. If personal data is collected through web scraping activities, DeepSeek would also need to clarify "how users registered and those not registered to the service have been or are informed about the processing of their data," wrote the Garante.

"Technological efficiency, however important, cannot prevail over people’s fundamental rights. It is essential to ensure that the development of AI occurs in full compliance with European regulation, which represents a pillar in the protection of fundamental values ​​and rights," Marco Scialdone, Head of Litigation at Euroconsumers, said to Eurofocus, commenting on the complaint.

This is not the first time that Italy has called out the privacy violations of AI chatbots.

In April 2023, the Garante temporarily banned the use of ChatGPT across the country as the AI mammoth was found responsible for improperly collecting and storing Italian data and, consequently, breaching GDPR rules. At that time, the ChatGPT ban provoked a surge in VPN downloads as Italians looked for ways to retain access to the popular AI chatbot.

It's too soon to tell whether DeepSeek will share the same fate, but we nonetheless advise you to approach the AI novelty with caution if you care about your online privacy.

• GTA 5's Trevor wants to return for GTA 6 so he can die at the beginning of the game
• Steve Ogg thinks it would be "fun" if Trevor appeared and was killed to "pass the torch" on to the next playable character
• Although, as far as Ogg is aware, Trevor won't be returning for GTA 6

Steve Ogg, the actor who plays Trevor Philips in Grand Theft Auto 5, wants his character to die in Grand Theft Auto 6.

In a recent interview with ScreenRant, Ogg touched on the possibility of Trevor returning to the popular franchise and even offered his idea of what it could look like.

"It would be fun if Trevor appeared in it just to be killed at the beginning," Ogg said, a pitch which seems to be a callback to how Trevor killed GTA 4's Johnny Kleibitz at the start of GTA 5.

However, as far as Ogg is aware, his iconic character isn't returning in GTA 6.

"I think that would be cool, because it also acknowledges the fans of like, 'Hey, thank you.' Pass the torch, stomp Trevor's head in, and sort of put an end to that and allow a new generation to take over," he added.

GTA 6 was officially announced in December 2023 alongside a reveal trailer, which confirmed long-time rumors that the game will feature two playable protagonists, Lucia and Jason.

The game is set to launch on PS5, Xbox Series X, and Xbox Series S. There's no word on a potential PC version just yet, but it is scheduled for a Fall 2025 release, according to Take-Two.

You might also like...

• Best PS5 games 2025 - top PlayStation 5 titles to play right now
• Don't panic! Baldur's Gate 3 players on PS5 who were affected by the accidental release of Patch 8 can fix their game using this easy solution
• Hideaki Nishino is now the sole CEO of Sony Interactive Entertainment, Hermen Hulst remains CEO of PlayStation Studios

• A suitcase-style turntable with speakers and Zig-Zag branding
• USB in plus Bluetooth, RCA and headphone outs
• $99 (around £79 or AU$159, if you can find one)

There's a long tradition of brands doing weird things for publicity, and for every Elon Musk stunt that makes you want to push every Tesla into a lake there's a brand that gets it right. And I think Zig-Zag is in the latter category, because it's come up with a rather inspired idea.

If you're not familiar with Zig-Zag, it's a very famous brand of rolling papers – so famous that their product inspired the cover of Dr Dre's 1992 album, The Chronic. And to remind you of that fact (and of course to ensure that you don't forget about Dre), the company has decided to make an – ahem – smokin' record player just in time for the Grammys…

(Image credit: Zig-Zag) The Zig-Zag Record Player: key features

The Zig-Zag Record Player is, as you'd expect, a record player with Zig-Zag branding. It's a suitcase design with the classic logo on top, and it has built-in stereo speakers so you can have music wherever you go. In addition to the turntable there's also a USB port so you can also play digital files without lugging all your vinyl around.

It's a three-speed belt-driven turntable – ideal for all your, er, rap 78s – and it has RCA outputs and a headphone jack too. There's Bluetooth streaming to headphones or powered speakers, and there's even a handle so you can carry it around.

According to Chisto.com, "Nothing beats the sound of spinning a vinyl whilst you’re chilling at dwelling. If you’re restricted on house however maxing out on chill vibes (particularly throughout Dry January), then you definitely want this... It’s the last word improve out of your standard transportable speaker that units the temper in any room with out the cumbersome setup of your typical document participant." Now, I don't know what many of those words mean, in that order, but I think they liked it…

The Zig-Zag Record Player is available now for $99, which is around £79 or AU$159, if you can find it.

You might also like

• The best turntables you can buy
• 3 great turntables for vinyl beginners
• Korg's new portable turntables make me want to rock the block

• Sophos report finds firms are worried GenAI flaws could harm their cybersecurity
• 99% claim AI is essential when choosing a provider
• Human-first approach appears to be key, Sophos says

The rise of artificial intelligence is coming at the expense of increased cybersecurity threats, and businesses are struggling to adapt, new research has claimed.

A report from Sophos revealed nine in 10 (89%) IT leaders worry that flaws in generative AI system could harm their businesses’ cybersecurity strategies.

Despite this, almost all (99%) IT leaders now consider AI capabilities essential when selecting a cybersecurity provider in the perfect example of fighting fire with fire.

AI’s role in cybersecurity

Artificial intelligence has given threat actors new powers, turning unskilled attackers into more sophisticated code creators, while making it harder for analysts to trace the origin of threats.

One in five respondents hoped AI will help them improve protection from cyberthreats, with 14% hoping for reduced employee burnout.

It all comes at a cost, though, with four in five believing that new AI tools embedded into their cybersecurity solutions will increase the cost of tools. Still, 87% believe the savings will outweigh the initial costs.

“We have not actually taught the machines to think; we have simply provided them the context to speed up the processing of large quantities of data,” Sophos Global Field CTO Chester Wisniewski said, adding firms should “trust but verify” GenAI tools.

An overwhelming majority (98%) of the companies surveyed now have some form of AI embedded into their cybersecurity infrastructure, but 84% are worried about the pressure to reduce workforces due to an over-reliance on the tech.

Wisniewski added: “The potential of these tools to accelerate security workloads is amazing, but it still requires the context and comprehension of their human overseers for this benefit to be realized.”

Looking ahead, Sophos calls for IT leaders to evaluate AI vendors for things like the quality and source of their training data, to establish measurable outcomes they hope to achieve from AI, and to adopt a human-first approach.

You might also like

• We’ve listed the best AI tools to help save you time
• AI could play a huge role in protecting companies from cyberattacks
• Boost protection with the best endpoint security software and the best antivirus software

Last week, DeepSeek, a Chinese AI startup, announced its R1 reasoning model, a breakthrough that changes the world of AI. DeepSeek’s model matches or even surpasses the quality of leading LLMs, but at a fraction of the training and running cost. This essentially commoditized what was once the exclusive domain of tech giants. Along with the model, DeepSeek released a detailed playbook describing how others can replicate their success, dramatically lowering the barriers to entry. 

Apple Just Saved Billions

For the last few years, Apple has been criticized for lagging behind in the “LLM wars.” Analysts and journalists pointed out that while companies like OpenAI, Google, and Meta poured billions into building state-of-the-art large language models (LLMs), Apple seemed to be on the sidelines. In fact, some industry insiders went as far as to call Apple’s approach “complacent.” Yet, as it turns out, it just paid off in a way no one expected.

This breakthrough means Apple can now develop competitive AI models without the multi-billion-dollar investments previously required. By staying out of the early stages of the AI arms race, Apple may have saved billions.

Goodbye, OpenAI Partnership

Apple’s perceived weakness in AI pushed it to form a partnership with OpenAI, a move that raised eyebrows across the tech world. Critics questioned the necessity of this alliance, with some calling it a “panic move” to compensate for Apple’s lack of in-house AI capabilities. Elon Musk, for instance, openly criticized the partnership, calling it “creepy spyware” and warning against potential privacy risks.

Now, Apple has a way out. With DeepSeek’s methodologies in hand, Apple no longer needs to rely on OpenAI for cutting-edge AI. They can cheaply develop their own models tailored to their ecosystem without external dependencies. This not only gives Apple the freedom to leave the partnership if they choose but also provides significant leverage in renegotiating terms.

Efficient compute is a key advantage for Apple

Over the past year, it’s become clear that LLMs alone won’t be the ultimate differentiator in AI. Instead, the real value lies in data, user context, UI, and distribution. Apple already excels in these areas, with a massive ecosystem of devices and services that seamlessly integrate into users’ lives. But now, thanks to DeepSeek’s breakthrough, Apple has an even greater advantage. 

DeepSeek’s R1 model is designed for efficient inference, meaning it can run on standard hardware rather than requiring expensive, specialized GPUs. This aligns perfectly with Apple’s strengths. Apple’s devices—from iPhones to Macs—are already optimized for high performance and energy efficiency. With this new capability, Apple can deploy powerful AI models directly on their devices, ensuring user privacy and reducing reliance on cloud-based solutions. 

There have long been rumors that Apple was studying how to run LLMs locally, and now that vision can become a reality. DeepSeek’s innovations make it possible for Apple to deliver state-of-the-art AI experiences without compromising on efficiency or security. For Apple’s users, this means smarter devices that protect their data while delivering unmatched performance. 

Markets React

DeepSeek’s announcement had an impact across the tech industry. Nvidia, a company synonymous with AI hardware, saw its stock plummet by 17%, wiping $589 billion off its market value. Investors reacted to the realization that LLMs no longer require the kind of specialized hardware that Nvidia’s business depends on.

Apple, on the other hand, remained unaffected. Its stock actually grew. Whether it’s caused by investors just moving their investments within a tech sector or whether investors also see that Apple might benefit more than any other company is a question for them. Whatever the reason it is, we live in exciting times and the times become more and more exciting for Apple as well.

Checkout our comprehensive list of the best AI tools.

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

• 51% of British businesses consider data sovereignty “crucial”, report claims
• Nearly equal measures see it as a burden or as an asset, OVHcloud finds
• Data access, handling, storage, and portability are all benefits

New research has revealed the growing need for data sovereignty, influenced by ongoing geopolitical tensions and market competition investigations.

A report from cloud computing giant OVHcloud found half (51%) of British businesses now consider data sovereignty as a crucial part of their data management strategies.

Moreover, three in four (77%) IT decision-makers now say that data sovereignty is more important than it was three years ago.

Data sovereignty is at the heart of building customer trust

Exactly why businesses are so keen on data sovereignty varies, though – two in five (40%) see it as a compliance issue, but more than one-third (36%) said that it involves data access, handling, and storage. A further 28% said sovereignty is related to data portability.

Data sovereignty has gained popularity in recent years following America’s battle with China and investigations into the dominance of American hyperscalers over global cloud markets - in 2024, AWS confirmed it would be supporting European data sovereignty, choosing Germany as its first location.

“Having a robust data sovereignty strategy means that you know who controls your data, where it is, who has access to it, and which regulations it’s subject to," noted OVHcloud multi-cloud evangelist Matt Tebay.

Of the 500 IT decision-makers surveyed, 41% said it’s just something they need to comply with, however, 42% added data sovereignty is important to their customers, and therefore an asset to their business.

“Organisations are increasingly realizing that it’s not simply a cost and regulatory burden, but can build better customer trust, enhance governance, and provide a layer of security and transparency to customers," Tebay added.

“Clearly, the intersection of customer trust, technology, regulation, and commercial considerations can make for a challenging journey, but it’s an important one and like all journeys, the first step is the most important.”

You might also like

• Choose from the best cloud storage solutions to keep your data accessible from anywhere
• Oracle CEO thinks there will be a sovereign cloud for every country soon
• We’ve curated a list of the best servers for small businesses

• OpenAI unveils ChatGPT Gov, built for the US government
• The AI chatbot can help with summarization, translation, and more
• ChatGPT Gov will be able to hand "non-public, sensitive information"

OpenAI has released a new version of ChatGPT, built bespoke for the US government.

ChatGPT ‘Gov’ can be fed "non-public, sensitive information” on secure self-hosted cloud computing platforms, allowing agencies to comply with cybersecurity requirements.

Government workers will be able to create and share custom GPT models for specialised tasks, with the Gov version also deploying an admin console for CIOs and IT teams.

Chat with the Gov

OpenAI claims over 90,000 users from upwards of 3,500 federal, state, and local government agencies have sent over 18 million messages on Chat GPT since 2024 as part of their day-to-day work across AI tools training, basic coding, translation services, and much more.

Visually, the service is similar to ChatGPT enterprise, with the main difference being that ChatGPT Gov can be hosted on internal Microsoft Azure commercial cloud, or Azure Government community cloud infrastructure.

As always, there are legitimate concerns around the security of using AI within government, particularly when entering sensitive information, to which OpenAI refers users to its Usage Policy, which ChatGPT Gov is also subject to.

Demonstrations made to the press by OpenAI solutions engineer Aaron Wilkowitz showed ChatGPT Gov create a five-week plan for a Trump administration employee, which was then analysed by the GPT after the addition of numerous notes. Wilkowitz also demonstrated its translation abilities by having it draft a memo of job plan and then translating it into multiple languages.

ChatGPT Enterprise is currently being subject to Federal Risk and Authorization Management Program (FedRAMP) testing before it can be used on the types of sensitive data it could be fed if adopted. OpenAI CPO Kevin Weil said (via CNBC) that while there wasn’t a timeline in place for ChatGPT Enterprise’s FedRAMP approval, OpenAI is working with the new administration to cut through some of the red tape.

“I know President Trump is also looking at how we can potentially streamline that, because it’s one way of getting more modern software tooling into the government and helping the government run more efficiently. So we’re very excited about that,” Weil said.

OpenAI’s CEO Sam Altman reportedly donated $1 million to President Trump’s inauguration fund, alongside OpenAI’s partner company Microsoft, which provides many of the systems, software, and cloud services that the government needs to function.

You might also like

• These are the best AI writers around right now
• Take a look at our guide to the best cloud storage
• “This is a wake-up call" - the DeepSeek disruption: 10 experts weigh in

• You can now use DeepSeek R1 inside Perplexity Pro
• The model is available with Perplexity on the web and iOS
• You get uncensored results and US-stored data

It's been quite the week for DeepSeek, the Chinese-made open-source AI model that's shot to the top of the app store charts while simultaneously raising privacy concerns. Now the DeepSeek R1 model is available within the Perplexity AI search engine, on US servers and with no censorship.

Announcing the news, Perplexity CEO Aravind Srinivas (via Search Engine Journal) described it as a "phenomenal experience", while also acknowledging that there are limits on query volume – limits Perplexity is working to increase.

And uncensored! https://t.co/vGUG0w1zlo pic.twitter.com/ceGYpLON50January 28, 2025

Perplexity is able to add DeepSeek R1 because it's open-source and freely available to anyone who wants to use it, and the move gets around two key concerns: first that China might censor the results DeepSeek returns, and second that the model would send user data back to the Chinese government.

Srinivas promises that Perplexity has tweaked R1 to keep results uncensored, and that all user interactions are kept secure on US-based servers. The CEO even used the classic test case, a search about Tiananmen Square, to show the AI is uncensored inside Perplexity.

You'll have to go Pro

You can get at DeepSeek R1 from the Perplexity iOS and web apps (Image credit: Future)

To be able to access the DeepSeek R1 model, you'll need to have signed up for a Perplexity Pro account – that'll set you back $20 (about £16 / AU$32) a month or $200 (about £160 / AU$320) a year. A subscription also gives you access to other models, like OpenAI o1.

For the time being at least, you're also going to have to use Perplexity on the web or through the iOS app – the feature hasn't arrived on Android yet. If you do use Perplexity on Android, you can try out the new Perplexity Assistant feature instead.

The arrival and instant success of DeepSeek has seriously disrupted the AI landscape, with US companies such as Nvidia and OpenAI politely praising their new rival while also scrambling to compete with the new R1 model – which can match the best models from the likes of ChatGPT for free, while using less processing power and less energy.

There's no doubt lots more to come from DeepSeek, and from its rivals (and partners). You can find plenty of experts weighing in on what's next, and the best place to keep up to date with the latest developments is at our DeepSeek live blog.

You might also like

• DeepSeek could come for Midjourney and DALL-E next
• All the latest DeepSeek news from our live blog
• DeepSeek is so popular it had to pause sign-ups

• Xbox has partnered with Fanta in a new giveaway
• Prizes including a custom Xbox Series X console bundle and Xbox Wireless Controllers are up for grabs
• To enter, you need to scan QR codes on eligible drinks in the Coca-Cola app

Xbox has teamed up with soft drink brand Fanta to give true Fanta lovers the chance to own an exclusive Fanta-themed Xbox Series X or a colorful Xbox Wireless Controller.

The competition will run across the UK until March 25 and in other regions in Europe. To enter, all you need to do is buy an eligible Fanta product and scan its QR code using the Coca-Cola mobile app. There are loads of prizes up for grabs, including 1 month Xbox Game Pass Ultimate membership.

I’m far more interested in the exclusive Xbox Series X console bundle, however, which decks out the monolithic system in a glorious neon yellow print complete with its own Fanta branding. The bundle also includes a matching Xbox Wireless Controller, with a yellow face plate and blue details.

You could also win one of a handful of Fanta-themed Xbox Wireless Controllers, with designs inspired by popular flavors like tangy lemon and fruit twist.

The competition’s fine print reveals that, in the UK, there will be a total of 1600 Xbox Game Pass Ultimate memberships up for grabs. It’s important to note that they will only be valid for new members and will require a debit or credit card to sign up, which is a bit of a shame.

There will be 15 Xbox Wireless Controllers, available via weekly prize draws, plus two Xbox Series X Console bundles as part of a grand draw. Thankfully, unlike the incredible Final Fantasy 14 Xbox Series X that was up for grabs in a previous promotion, it does seem like this custom console actually works.

All things considered, it seems like a pretty solid giveaway and I will definitely be trying my chances at winning next time I find myself craving a sweet orangey treat.

You might also like...

• Sniper Elite: Resistance review: more of the same, and that’s great
• Xbox games are coming to Nintendo Switch 2, per Phil Spencer
• Doom: The Dark Ages is id Software’s biggest ever Doom game, and includes ‘the largest AI we’ve ever created’ on the studio’s ‘most powerful engine yet’

Generative AI (gen AI) isn’t just a tech trend; it’s reshaping industries at every level. From personalizing shopping experiences to enhancing fraud detection in finance, we’re beginning to see its transformative potential. But what’s next? Enter enterprise AI.

Enterprise AI is set to blend business strategy with advanced technology, driving continuous transformation across organizations. By combining AI tools like gen AI, machine learning, and robotic process automation (RPA) with orchestration, enterprise AI optimizes operations, boosts productivity, and upholds high standards of quality, security, and governance. This approach enables businesses to apply gen AI strategically across their operations, maximizing its potential impact.

As we look toward 2025, companies investing in enterprise AI will benefit not only from gen AI but from a comprehensive suite of AI tools purpose-built to reshape the workplace. With enterprise AI as a true vehicle for realizing the promise of gen AI, we anticipate three major trends that will define how industries leverage AI to meet future demands.

Welcome to workplace 5.0

Workplace 5.0 takes the era of automation further, emphasizing and strengthening the collaboration between humans and technology. Industry 5.0 is where humans, AI, robotics and digital-integrated activity combine.

Central to this is orchestration: with business process management (BPM) and other tools, organizations can integrate human and digital workflows end-to-end. The key to successful operations is no longer simply orchestrating human work but integrating work as a whole. This will bring together workflows, inputs, outputs and the resources that drive the work, humans, gen AI and enterprise agents to create a smarter, more collaborative, engaging and productive work environment.

Enterprise AI will allow us to move on from the individual efficiency gains we have seen through the likes of ChatGPT, to a grown-up implementation of gen AI where leaders have the opportunity to revolutionize the end-to-end processes on a large scale. This shift will lead to more human-centric work, making enterprises more creative, adaptive and customer-focused.

Human-centric collaboration with enterprise AI agents

The next evolution of gen AI’s integration with the enterprise is to make interactions feel more lifelike. This will be achieved through experiential interfaces, enabling humans to communicate with enterprise agents in natural language. We’ll see this manifest in the form of copilot-like interfaces, which are posed to dramatically enhance productivity, drive innovation and transform industries such as healthcare, manufacturing and financial services.

While AI agents are not new, 2025 will mark the rise in more advanced business agents with specialized knowledge on specific industries. These enterprise agents are not meant to replace human workers, but rather to function as coworkers that support and enhance performance. They will autonomously handle delegated tasks, continuously learning and improving through agentic AI.

Previously, these tools required coding and a structured approach to building applications. Now, gen AI has levelled the playing field, allowing people to turn their ideas into reality through natural language prompts.

Mitigating risks with enterprise AI governance

As generative AI tools advance quickly, there’s growing concern about their potential to create privacy and security risks. With the possible consequences of a rogue agent (rogue AI) such as business/supply chain disruption, reputational damages and customer loss, it’s worth considering these concerns.

To navigate this, organizations will need enterprise-grade guardrails, succinct integrations of gen AI into critical workflows and a robust AI governance model that prioritizes transparency, documentation, and bias prevention. This ensures AI solutions enhance, not jeopardize, the business.

Enterprise AI: turning potential into impact

Enterprise AI will be the game-changer for organizations ready to get the most out of generative AI, applied strategically and effectively across their entire operations. As we look ahead to 2025, companies that embrace enterprise AI will gain unparalleled access to gen AI’s benefits—not just in isolated applications, but through integrated processes combining RPA, orchestration, machine learning, and more. This convergence will enable organisations to turn the potential of gen AI into tangible results, achieving the scale and impact that early AI applications only hinted at. The era of delivering on AI’s promise across the enterprise has arrived.

We've compiled a list of the best Enterprise Resource Planning (ERP) software.

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

New AI chatbot DeepSeek is taking the technology world by storm, and has even dethroned ChatGPT from top spot on the iOS App store.

There's certainly a lot of speculation about the potential DeepSeek holds, as well as what this could mean for some of the biggest players in tech right now. Plenty of both new and existing firms have invested billions of dollars into Large Language Models in the last few months alone - but DeepSeek has the opportunity to cause chaos in the venture capital world.

Having caused the biggest drop in the US stock market in history, the chatbot is understandably a huge talking point - but what are the experts’ predictions?

Will the hype last?

There’s no denying that DeepSeek arriving on the scene has been disruptive, and although the expert opinion isn’t unanimous, the significantly lower development cost is definitely turning heads.

If you're confused about what DeepSeek is, or why its causing so much of a stir, check out our article here explaining all you need to know,

The pressure is certainly on for US tech firms to respond, but it may not be as destructive as it seems.

We heard from 10 experts in the technology industry;

Steve Povonly, Senior Director of Security Research & Competitive Intelligence at Exabeam:

"The release of Chinese-developed DeepSeek has thrown US tech markets into turmoil; this is both justifiable and also perhaps, a bit overblown. The emergence of a technology that ultimately optimizes chip usage and efficiency is likely to apply pressure on existing large chip vendors, which is a very good thing. As the adage goes: "Pressure yields diamonds" and in this case, I believe competition in this market will drive global optimization, lower costs, and sustain the tailwinds AI needs to drive profitable solutions in the short and longer term."

Mike Follett, CEO and Founder of Lumen Research:

“Last week, Sam Altman and Elon Musk were AI monopolists. Now, it is clear that competition will reduce costs and loosen the grip hold of the biggest players. This represents an opportunity for marketers enabling them to build agents designed for specific problems, at a speed, scale and at a price that drives ROI.”

Nigel Green, CEO of deVere Group:

“This is a wake-up call for markets. The assumption that tariffs could contain China’s technological ambitions is being dismantled in real time. DeepSeek’s breakthrough is proof that innovation will always find a way forward, regardless of economic barriers.”

“By restricting China’s access to high-end semiconductors, Washington sought to slow its progress in AI. Instead, it has fueled an acceleration in domestic innovation, forcing Chinese firms to find alternatives. DeepSeek’s achievement is a direct result of this shift.

“Rather than being crippled by US sanctions, Beijing has cultivated AI models that require significantly less computing power, diminishing its reliance on American technology and eroding US leverage over global supply chains.”

The value

Aleksandr Yampolskiy, CEO of SecurityScorecard:

“DeepSeek is trained on 14.8 trillion diverse tokens whereas OpenAI is trained only on 13 trillion. It also costs radically less to train DeepSeek at $6M while OpenAI costs allegedly $100M, making DeepSeek 16.6X more efficient.”

“We are living in fascinating times. While "constraints in capital" may seem like a challenge, history has shown us (and DeepSeek has demonstrated) that these constraints often spark innovation and creativity. Security for AI will only become more critical. In a world where the lines between deepfake and human-generated content blur, and where biased information can shape our opinions, the need for robust security and ethical practices will grow exponentially.”

Dr. Kjell Carlsson, Head of AI strategy at Domino Data Lab:

“Deepseek’s success serves as a powerful reminder that the value of AI lies not in the size of your infrastructure or the exclusivity of your models, but in how effectively they are leveraged to deliver impact. By developing cutting-edge generative AI models without relying on the latest, most expensive hardware, Deepseek has demonstrated that agility and strategy can outpace raw computational power. Their achievements also highlight the vulnerability of incumbents in the generative AI space—proving that open-source innovation continues to be a powerful equalizer, enabling challengers to match and even surpass established players years into the revolution.”

“For companies seeking to maximise value from AI, the lesson is clear: success hinges on flexibility and capability, not exclusive partnerships or infrastructure scale. Rather than locking into specific LLM providers or focusing solely on hardware access, organisations should prioritise building the end-to-end capabilities to source AI innovations, design solutions tailored to their unique needs, and operationalise them effectively. This approach ensures that businesses remain agile, competitive, and prepared to harness the next wave of AI advancements, wherever they emerge.”

Bradford Levy, Assistant Professor of Accounting, University of Chicago Booth School of Business:

“DeepSeek has sent shock waves through the tech industry – directly challenging tech giants like Meta, Microsoft and Open AI."

“Until now, it’s been assumed their expertise in designing and operating large-scale distributed systems are essential for training state of the art models. But the development of R1 suggests otherwise – if these models can be trained using 90% fewer chips, the implications for valuation models are massive."

“This opens the door for smaller, more agile players to compete, potentially driving more innovation. With limited resources, they proved that scrappy, innovative teams can shake up the industry, even on a shoestring budget."

“While impressive, we should remain skeptical of any claims made by those with a vested interest in their own success. Before jumping to conclusions about the broader AI landscape, we need more time to test these models and understand how they achieved these numbers.”

The revolution

Professor Geoff Webb, Department of Data Science & AI, Faculty of Information Technology, Monash University:

“The emergence of DeepSeek is a significant moment in the AI revolution. Until now it has seemed that billion dollar investments and access to the latest generation of specialised NVIDIA processors were prerequisites for developing state-of-the-art systems.”

“This effectively limited control to a small number of leading US-based tech corporations. Due to US embargoes on exporting the latest generation of NVIDIA processors, it also locked out China.”

“DeepSeek claims to have developed a new Large Language Model, similar to Chat GPT or Llama, that rivals the state-of-the-art for a fraction of the cost using the less advanced NVIDIA processors that are currently available to China. If this is true, it means that the US tech sector no longer has exclusive control of the AI technologies, opening them to wider competition and reducing the prices they can charge for access to and use of their systems.”

“Looking beyond the implications for the stock market, current AI technologies are US-centric and embody US values and culture. This new development has the potential to create more diversity through the development of new AI systems.”

“It also has the potential to make AI more accessible for researchers around the world both for developing new technologies and for applying them in diverse areas including healthcare.”

The risks

Aditya Sood, VP of Security Engineering and AI Strategy at Aryaka:

"Open-source AI models like DeepSeek, while offering accessibility and innovation, are increasingly vulnerable to supply chain attacks triggered during large-scale cyberattacks. These attacks, where adversaries exploit the reliance on third-party dependencies, pre-trained models, or public repositories, can have severe consequences. Adversaries may tamper with pre-trained models by embedding malicious code, backdoors, or poisoned data, which can compromise downstream applications. Additionally, attackers may target the software supply chain by manipulating dependencies, libraries, or scripts used during model training or deployment. This can lead to systemic AI functionality corruption."

Renuka Nadkarni CPO at Aryaka:

"The sudden popularity of DeepSeek comes at a price. There are two dimensions of this. First, threat actors are likely to adopt this new tool now that it's widely available. Second, DeepSeek was a victim of a large-scale malicious attack. This means that their system could be compromised and subject to several of the known AI model attacks. Known AI model vulnerabilities, data risks, and infrastructure threats come into play here."

“While the unavailability of the service is an easy and visible attack on its infrastructure, the bigger concern lies in the undetected attacks on its model and data. These hidden threats could compromise benign users and enable other malicious activities."

The sceptics

Dan Goman, CEO, Ateliere Creative Technologies:

"The market’s reaction to the latest news surrounding DeepSeek is nothing short of an overcorrection. While the enthusiasm around breakthroughs in AI often drives headlines and market speculation, this feels like yet another case where excitement has outpaced evidence. Investors should be cautious about blindly jumping on the hype train without asking the tough questions."

"In summary, while Deepseek’s story is intriguing, it’s imperative to separate fact from speculation. The market needs to temper its enthusiasm and demand more transparency before awarding DeepSeek the crown of AI innovation. Until then, skepticism remains a healthy and necessary stance."

You might also like

• Check out our list of the best AI tools around today
• Scammers have a new phishing trick for iPhone users
• We've also rounded up the best AI website builders on offer right now

• Dyson has launched the Dyson Car+Boat handheld vacuum
• It has many of the same features as Dyson's full-sized stick vacuums
• Available in US and UK from 4 Feb, priced $279.99 / £249.99

I've been waiting for Dyson to release a handheld vacuum, and my patience has finally paid off: today, the brand unveiled the Dyson Car+Boat, and it looks set to outshine the rest of the best handheld vacuums currently on the market.

The Car+Boat looks like someone took a Dyson stick vacuum and shrunk it in the wash. You can't really tell from the press photos, but this thing is dinky. That pint-sized build packs a punch though – this handheld model has a lot in common with the best Dyson cordless vacuums from the wider range, including powerful suction, useful detail tools, advanced filtration and a long battery life. Unfortunately, the similarity also extends to the price – its list price is $279.99 / £249.99, which is cheaper than the cheapest Dyson stick vacuum, but much more expensive than your average car vacuum.

Before I get too far into it, I'll be clear that despite being called the Car+Boat, this is not a wet-dry vacuum. It's just a regular handheld vacuum, aimed at people who might own a boat. Dyson is nothing if not aware of its target market. Other things its designed to tackle include dirt, debris, allergens, pet hair, upholstery and bedding.

(Image credit: Dyson)

Let's start with battery life, which is the bit that I'm probably most excited about. Most cordless handheld vacuums offer you a maximum of 30 minutes of cleaning, but sometimes it's more like 15 minutes. Here you have a full 50 minutes, which is up there with the best full-sized cordless vacuums on the market – and, in fact, longer than you'll get with the Dyson V8 .

That accounts for the fact that cleaning cars, furniture and, uh, boats, is fiddly. It takes time to do it properly, and there's nothing more annoying than running out of juice once you've wedged yourself into a footwell. With nearly an hour of cleaning, you have time to get into every nook and cranny.

Dyson also promises top-notch suction. The proof will be in the testing (a full review is in the works) but the specs look promising – the motor spins at 110,000rpm, same as on the V8, which is the oldest Dyson stick vacuum in the current lineup. There's a fully sealed filtration system that can capture 99.99% of particles as small as 0.3 microns (the same as all but the very newest and most expensive Dyson stick vacuum). That includes dust mites, pet dander and other allergens.

(Image credit: Dyson)

The Dyson vacuum attachments are similar to those you'll get with a full-sized stick vac, too. There's a Mini Motorized Tool for pulling up stubborn dirt and sticky hair, a 2-in-1 Combination Tool (pictured above) that combines a dusting brush and wide nozzle, and a Crevice Tool for edges and narrow gaps.

While this is the only dedicated handheld in the current Dyson lineup, its not the first handheld from the brand. Most recently, there was the Dyson Humdinger, which has just been discontinued in the US, and which left the UK lineup back in 2022. Of course, any of Dyson's cordless stick vacuums can be used in handheld mode by removing the floor wand and adding a detail tool, but the main vacuum unit tends to be too bulky and heavy to be an ideal choice as a car vacuum.

The Dyson Car+Boat will go on sale in the US and UK direct from Dyson from 4 February, at a list price of $279.99 / £249.99 [US release date and pricing has been updated from a previous version of this article]. It will also be available from various third-party retailers.

You might also like...

• Dyson Airwrap's dreamy new colorways might finally tip me over into buying one
• What is the Dyson Digital Slim vacuum, and should I buy one?
• Shark UltraCyclone Pet Pro Plus review: read about our #1 handheld vacuum

The IT industry is undergoing seismic change as businesses adapt to a rapidly evolving technological landscape. Emerging trends such as generative AI, sustainable computing, and secure enterprise infrastructure are pushing organizations to rethink their approaches to scalability, energy efficiency, and innovation. These shifts are not merely incremental—they are transformative, reshaping the foundation of IT strategy for years to come.

Bold leaders are tasked with solving unprecedented challenges, including meeting increasing demand for compute power while maintaining operational sustainability. The rise of AI-driven workloads is placing new pressures on IT infrastructure, requiring companies to adopt architectures capable of delivering high performance at scale.

Simultaneously, the need for energy-efficient solutions is accelerating innovation in both hardware design and data center management, ensuring these systems remain cost-effective and environmentally responsible. Moreover, decision-makers must ensure their strategies align with growing concerns over data sovereignty, addressing regulatory demands while protecting sensitive assets.

From maximizing data center efficiency to adopting renewable energy solutions, companies face both challenges and opportunities as they prepare for the demands of tomorrow. The trends discussed here demand a holistic approach that balances innovation with responsibility, ensuring that technology not only meets current needs but also paves the way for a sustainable and secure future.

1. From Experimentation to Execution: Generative AI Inference Takes Center Stage

Generative AI is transitioning from merely experimental AI tools to fully integrated solutions that provide substantial business value. While the past year focused on chatbot use cases, largely using public data, the future lies in applying generative AI to private, secure datasets to create even more valuable tools. Enterprises in sectors like finance, insurance, and ecommerce are poised to adopt these technologies to extract meaningful insights from proprietary data.

Deployment flexibility will be critical. As AI workloads expand into diverse environments — on-premises, edge, and air- gapped hosting facilities — latency-sensitive applications will demand infrastructure closer to users, deployed in existing data centers and PoPs. Moreover, inference is no longer a standalone workload. Supporting tasks like retrieval-augmented generation (RAG) and app integration will require robust, general-purpose compute alongside AI-specialized resources, emphasizing efficiency and scalability.

2. Powering the Future: Renewable Energy Growth Plus Efficiency Gains

As compute demands surge, so does the need for power. However, overloaded grids and geographic power constraints are forcing industries to seek new solutions. Renewable energy sources like solar, wind, and geothermal are gaining traction as smaller, regionally distributed data centers emerge. These projects will take more time than is available to meet the immediate demands of IT infrastructure growth.

Efficiency, however, cannot wait. To avoid bringing new non-renewable energy sources online or prolonging their life in the short-term, hardware optimization will play a pivotal role in reducing power requirements. Replacing older, power- hungry systems with modern, efficient processors can dramatically cut energy use, making existing infrastructure more sustainable. This efficiency shift is critical to balancing the need for more energy with responsible environmental stewardship.

3. The Rise of Density: Maximizing Every Rack and Data Center’s Potential

Given the rapid increase in demand for AI compute, density at scale has become the new benchmark for efficiency in computing. Solutions are being built not at the node level, but at the rack and data center level. This means that organizations are moving toward maximizing workloads per rack by fully utilizing available hardware. Unlike legacy systems, where resources were often underutilized due to inefficiencies, modern architectures are designed to eliminate waste and improve average utilization at rack and data center scale without the negative side-effects of unpredictability.

The challenge of optimizing density at the solutions level is not limited to AI-only workloads. Certain AI workloads, particularly inference, are driving infrastructure changes to accommodate mixed-use environments, where general purpose compute density matters as well. In software engineering organizations, more efficient virtualization and containerization technologies combined with more efficient containers and power aware coding practices will enable better partitioning of resources, allowing enterprises to achieve higher utilization rates without compromising performance.

4. Sovereignty and Security: Enterprise AI on the Rise

Data sovereignty and security will heavily influence AI deployment strategies in 2025. Enterprises are increasingly aware of the value of their proprietary datasets, treating them as competitive assets. This shift will mean that AI inference workloads run not only on public hyperscale clouds, but also in more secure environments like private clouds, on- premises data centers or privately hosted facilities.

The risk of data breaches and tampering with AI algorithms underscores the need for secure, isolated infrastructure. As enterprises compete on AI-driven innovation, the ability to safeguard intellectual property and sensitive information will become a cornerstone of success. Furthermore, this trend will expand the role of enterprise-owned compute resources, creating a more decentralized and secure AI ecosystem. This sovereignty and security requirement combined with the need to place computing resources closer to users will disperse computing resource and give rise to a more compute heavy edge architecture.

Summary

The trends outlined here reflect a fundamental shift in how businesses harness technology to drive efficiency, cybersecurity, and innovation. Generative AI is advancing from experimentation to execution, energy optimization is becoming non- negotiable, and maximizing data center density has emerged as the new benchmark for scalable infrastructure. At the same time, the emphasis on data sovereignty and security will ensure enterprises remain in control of their competitive assets.

Organizations that succeed in this rapidly evolving environment will prioritize agility, leveraging AI-driven insights to optimize operations while addressing pressing concerns such as resource constraints and regulatory compliance. These efforts will not only improve performance but also position companies as leaders in the drive toward a sustainable future.

Forward-thinking businesses will explore partnerships that enable them to expand capabilities while minimizing risks, ensuring sustained growth in the face of uncertainty. Investments in cutting-edge architectures, renewable energy integration, and secure AI deployments will form the backbone of IT strategies in 2025 and beyond. By aligning innovation with accountability, businesses can unlock lasting competitive advantages while fostering resilience in the face of constant change.

Organizations ready to embrace these shifts will not only overcome today’s challenges but also set the stage for sustained leadership in a rapidly evolving technological landscape.

We've compiled a list of the best data visualization tools.

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