TechRadar News

• CMA says UK cloud market is uncompetitive
• AWS and Microsoft account for 30-40% of the UK market each
• The two companies disagree with the CMA's findings

The UK's Competition and Markets Authority (CMA) has determined Britain's cloud market shows too many anticompetitive traits, with Microsoft and AWS each holding around 30-40% of the UK market in 2024 and hyperscaler concentration especially high in infrastructure-as-a-service.

At the same time, fewer than 1% of customers switch providers annually and multicloud usage is rare (particularly among SMEs with more limited budgets).

The CMA has blamed high egress fees, incompatible interfaces, latency and skills gaps for widespread vendor lock-in, which is ultimately weakening competition.

CMA worried about AWS and Microsoft cloud market dominance

Behind the two hyperscalers, Google accounts for just 5-10% of the market, with others like IBM and Oracle having even smaller shares. Although AI capabilities are yet to change market dynamics drastically, existing positions are likely to be amplified, thus the CMA has stepped in to ensure competition remains healthy.

In its Final Decision ruling, the CMA took the biggest hits at Microsoft over its unfair licensing practices, which make it costlier to run Microsoft software on rival cloud providers.

A Microsoft spokesperson told TechRadar Pro: "The CMA Panel’s most recent publication misses the mark again, ignoring that the cloud market has never been so dynamic and competitive, with record investment, and rapid, AI-driven changes. Its recommendations fail to cover Google, one of the fastest-growing cloud market participants."

"Microsoft looks forward to working with the Digital Markets Unit toward an outcome that more accurately reflects the current competition in cloud that benefits UK customers," they continued.

"The action proposed by the Inquiry Group is unwarranted and undermines the substantial investment and innovation that have already benefited hundreds of thousands of UK businesses," an AWS spokesperson added.

On the flip side, Google supported the CMA's findings: "The conclusive finding that restrictive licensing harms cloud customers and competition is a watershed moment for the UK."

Elsewhere in the industry, the CMA has been criticized for not acting fast enough and addressing persistent issues like cloud credits, lock-in and procurement bias.

"We urge the CMA to use the powers at its disposal now to address these harms, rather than embark upon a new investigation that may not give customers relief for years to come," Coalition for Fair Software Licensing Executive Director Ryan Triplette shared.

Looking ahead, the CMA's next step is to designate Microsoft and AWS with strategic market status (SMS) under the Digital Markets, Competition and Consumers (DMCC) Act, allowing it to impose legally binding, targeted conduct requirements on the two giants.

"A significant driver of high cloud computing bills is the consolidation of the market into a handful of players. Until recently, these companies have been the only game in town, so they’ve been able to set the rules of the market, for example, including egress fees for switching, long lock-in periods, and more. In fact, Gartner has observed that most customers spend 10% to 15% of their cloud bill on egress charges," noted Akamai's John Bradshaw.

"UK businesses are under huge cost pressures. We need to make it easier for them to switch cloud computing providers and find pricing options that better fit their balance sheets."

You might also like

• These are the best cloud hosting providers
• We've listed the best cloud storage and best cloud backup platforms
• Microsoft hits back against UK competition lawsuits, slams AWS and Google once again

Cyber-attacks continue to dominate headlines, disrupting operations and putting sensitive data at risk. In the wake of the AI boom, threats are growing more complex. The endless game between attacker and defender is intensifying, and defenders know the stakes are high. Operational, financial, and reputational damage can be severe when an attack succeeds.

At the same time, security teams face a widening skills gap, growing threat complexity and tighter budgets. It’s a perfect storm for burnout. In fact, 79% of cybersecurity professionals reported that escalating threats are impacting their mental health, highlighting the need for an empathetic approach to these challenges.

Prevention as the shield, resilience as the backbone

Historically, organizations have measured cybersecurity success by how well they prevent attacks. But with 90% of IT and security leaders reporting cyber incidents in the past year alone, it’s clear that prevention alone is no longer enough.

It’s time to shift the focus towards recovery, transparency, and resilience. Resilience shouldn’t be seen as a fallback – it needs to become the frontline. This shift in mindset not only better prepares organizations for inevitable breaches but also reduces pressure on teams by redefining what success looks like.

When teams are judged on their ability to recover and minimize disruption (not just prevent attacks), they’re empowered to focus on what matters; early detection, rapid response, and recovery planning. This reduces burnout and builds stronger long-term security posture.

We must also accept a hard truth; breaches will happen. Rather than fueling a culture of blame, we need to equip teams to respond effectively and confidently.

Securing the security team with transparency

As ever, collaboration in a crisis is critical. Security teams working closely with backup, resilience and recovery functions are better able to absorb shocks. When the business is confident in its ability to restore operations, security professionals face less pressure and uncertainty.

This is also true for communication, especially post-breach. Organizations need to be transparent about how they’re containing the incident and what’s being done to prevent recurrence. Trust drives everything and must be built into architecture, communication, and response, from user behavior to board confidence.

Shared risk, shared responsibility

As seen with the recent retail cyber-attacks in the UK, the implications of a cyber breach can be business critical. Yet many CISOs still struggle to get alignment at board level. Over three-quarters (77%) of UK CISOs feel that their IT budget is not completely reflected by their board’s objectives for cybersecurity.

To make matters worse, this is heightened when it comes to regulatory pressures. New legislation like DORA and the upcoming Cyber Security and Resilience Bill is turning up the heat, with over half (58%) of CISOs feeling the pressure as a direct result.

There is also an element of the blame game going on, with everyone keen to avoid responsibility for an inevitable cyber breach. It’s much easier to point fingers at the IT team than to look at the wider implications or causes of a cyber-attack. Even something as simple as a phishing email can cause widespread problems and is something that individual employees must be aware of. Security is everyone’s business - the attack surface isn’t just focused on IT, it’s every team, tool, and workflow.

This critical gap jeopardizes not only an organizations' security posture but also their ability to meet evolving regulatory demands. CISOs, boards, and other stakeholders must work together to ensure that cyber resilience priorities are clearly defined, adequately funded, and effectively implemented to meet the evolving regulatory landscape.

The weight of responsibility for cyber security shouldn’t just lie on the security team’s shoulders. Cyber resilience is business resilience and security leaders, boards and stakeholders all have a part to play.

Building teams that thrive

To build and retain a capable cybersecurity team amid the widening skills gap, CISOs must lead a shift in both mindset and strategy. By embedding resilience into the core of cyber strategy, CISOs can reduce the relentless pressure to be perfect and create a healthier, more sustainable working environment.

But resilience isn't built in isolation. To truly address burnout and retention, CISOs need C-suite support and cultural change. Cybersecurity must be treated as a shared business-critical priority, not just an IT function. This means aligning investment with board expectations, embedding security into daily operations and ensuring every employee understands their role.

With regulatory pressure rising and the threat landscape evolving, resilience isn’t just a technical necessity, it’s a strategic imperative. CISOs who champion collaboration, drive cultural change, and lead with empathy will be best positioned to build security teams that are not only effective but built to last.

We list the best employee recognition 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

• Nvidia has announced that support for GTX 10 series GPUs ends in October 2025
• After that, these graphics cards, including the GTX 1060, will only get security fixes
• It also announced that Windows 10 support will run through to October 2026, mirroring Microsoft's extended support program for the OS

Nvidia has released a new graphics driver and announced that it'll soon be drawing the curtain on support for GeForce GTX 10 series GPUs, as well as GTX 900 models - and the end for Windows 10 gamers will follow a year later.

As Ars Technica highlighted, the release notes for driver version 580.88 came with the revelation that graphics cards based on Maxwell and Pascal architecture - meaning GTX 900 and 10 series products - will witness their final driver release in October 2025.

After that, they will only get quarterly security updates to patch them against vulnerabilities, and that's all. Security patches will finish in October 2028 for these products, too.

If October 2025 rings a bell, that's because it's also the month when Microsoft casts aside support for Windows 10, and that's also wrapped up in this Nvidia announcement.

Team Green said that it's extending Game Ready Driver support for Windows 10 to October 2026, to mirror the extended support Microsoft is offering consumers who want to stick with the OS, and not upgrade to Windows 11 yet. Or indeed people who may not be able to upgrade to the newer operating system, due to their PC not meeting the hardware requirements.

This move comes as no surprise, as Nvidia already told us back at the start of July that the v580 drivers would be the last to support Maxwell and Pascal graphics cards - we just didn't know exactly when the cut-off was coming, and now we do.

(Image credit: Nvidia)If you're affected, what does this mean exactly?

As stated, there are two categories of PC gamers who this affects: those with GTX 10 model GPUs, like the GTX 1060, and those running Windows 10. Further, some folks will be in both camps, no doubt - maybe quite a few.

GTX 10 series graphics cards are still reasonably popular in some cases (whereas GTX 900 products have pretty much dwindled away to nothing). In fact, the GTX 1060 is actually the 12th most popular GPU according to the latest Steam hardware survey - and once reigned supreme - so it's still seeing a lot of use.

After October 2025, this GPU, along with other 10 series offerings like the 1070 and 1080, will only receive security updates. That means they'll still be safe to use - patched against any exploits in drivers that may be found by the bad actors out there - but they won't get support for new games or features.

So, as time rolls on, you'll find that your trusty GTX 1060 becomes wonkier and less reliable with new games, as its final driver version ages and generally gets more erratic. Note that if you stick with old games, which were catered for before game support was frozen, you should be fine, at least in theory.

As for those on Windows 10, you'll be okay for another year yet. You'll still have full driver support through to October 2026, as noted, so you'll be fine until then. Assuming you keep Windows 10 itself secure, of course - using Microsoft's offer of extended support, which is now free, with a slight catch.

After October 2026, though, you'll need to upgrade to Windows 11, or you won't get new drivers – so no game support, or security patches either – no matter how new your Nvidia GPU is.

At this point, you're really looking at a Windows 11 upgrade - or a switch to something else entirely - unless Microsoft extends Windows 10 support further for consumers beyond 2026 (which seems unlikely, but could happen). In which case, Nvidia might again mirror the move with its own drivers - given that's what has happened here - but nothing's guaranteed by any means.

You might also like

• Need a new GPU? Nvidia's RTX 5000 Super models may be coming sooner than you expect
• Nobody wants 8GB GPUs from Nvidia and AMD – and this retailer just made that clear
• AMD isn't giving up on high-end GPUs – a new leak hints at a new Radeon GPU challenging Nvidia's next-gen flagship graphics card

• One in six US workers say they lie about using AI to meet job expectations
• Engineers who use AI are the new threat, not the tools themselves
• Many workers copy AI-literate peers just to appear competent in modern workplaces

As AI tools spread across office environments, many US workers now find themselves in an odd situation: pretending to use artificial intelligence at work.

A recent survey by tech recruitment firm Howdy.com found that one in six employees claim to lie about using AI.

This phenomenon appears to be a reaction not only to managerial expectations but also to deeper insecurities around job stability in an AI saturated landscape.

Survival of the most artificial

Underneath the behavior is what some are calling “AI-nxiety,” an unease born from conflicting narratives.

On the one hand, companies urge employees to embrace AI to boost productivity; on the other hand, those same workers are warned that AI, or someone more skilled at using it, could soon replace them.

This sense of pressure is particularly acute when considering workers who fear being displaced by technically skilled peers, such as engineers who actively use LLM based systems and other AI tools.

As one commenter put it on The Register: “You may lose your job to an engineer who uses AI.”

For some, the message is clear: adapt or get left behind.

In late 2023, a survey by EY found that two thirds of white collar US workers feared being passed over for promotion by AI savvy colleagues.

In this environment, mimicking the behavior of the AI literate becomes a way to hedge against obsolescence.

Further complicating the picture is the lack of adequate training.

Howdy.com reports that a quarter of workers expected to use AI receive no instruction on how to do so.

Without proper guidance, many are stuck between expectations from management and the reality of poorly integrated AI systems.

Some give up on mastering the tools and simply act like they are already doing it.

Meanwhile, contradictory workplace norms deepen the confusion.

Another survey from Slack’s Workforce Index found that nearly half of global desk workers felt uncomfortable telling managers they use AI, worrying it may make them appear lazy or unoriginal.

Thus, some pretend not to use AI even when they do.

At the heart of the issue is a growing mismatch between what companies signal, “AI is the future,” and what employees experience: unclear expectations, low support, and shifting norms around competence.

Whether AI actually replaces jobs or not, the psychological toll is already here, and pretending to be an AI user has become a strange new survival strategy.

You might also like

• These are the fastest SSDs you can buy right now
• Take a look at some of the best external hard drives available
• The semiconductor industry is losing billions of dollars ever year because of this obscure little quirk

It feels like everyone and their aunt is making AI / AR smart glasses nowadays, especially as someone who tests the best smart glasses around. But something caught my eye when reading a description of Brilliant Lab’s new Halo glasses – as with their long-term memory capabilities, they promise to remind you of details of conversations and objects you’ve seen “years or even decades later.”

In real-time, Brilliant Labs’ specs can apparently offer contextual information based on what it hears and sees, too. This style of assistive help in the moment and later on sounds like a more ongoing version of features like the Ray-Ban Meta glasses’ visual reminders, features that Meta and others have said they plan to make (or have already made) an optionally always-on tool.

Now, Brilliant Labs has said its agent Noa will serve as a sort of AI VPN. Like a VPN reroutes your data to keep your online activity more private, Noa promises to offer similar levels of privacy as it communicates with the AI model powering its cognitive abilities.

Other Halo highlights are its “world’s thinnest AI glasses” design, its built-in display that sits in your periphery like some other AI specs we’ve seen announced this year, and it will have a relatively affordable $299 asking price (around £225 / AU$465) when it launches in November.

But even as someone who loves my Ray-Ban smart glasses and can see the benefits of these Halo glasses, I’m worried these smart specs are a sign we're continuing to race towards the death of privacy.

AI in your glasses can be handy (Image credit: Meta)Risk vs reward

Smart glasses wearables with cameras are already, admittedly, something of a privacy conundrum. I think the Meta Ray-Ban specs do it well – only letting you snap pictures or short videos (or livestream to a public Meta account on Facebook or Instagram), and have an obvious light shine while you do so.

But the next generation of utility wants to boast an always-on mentality – cameras that activate frequently, or microphones that capture every conversation you have.

This would be like the Bee wristband I saw at CES (which Amazon recently bought), which promises to help you remember what you talk about with detailed summaries.

You can instantly see the advantages of these features. An always-on camera could catch that you’re about to leave home without your keys, or remind you that your fridge is getting empty, and Bee highlighted to me that you could use it to help you remember ideas for gifts based on what people say, or recollect an important in-person work chat you might have.

However, possible pitfalls are close behind.

How private are we really? (Image credit: Shutterstock)

Privacy is the big one.

Not just your own, though you’re arguably consenting to AI intrusiveness by using these tools, but the privacy of people around you.

They’ll be recorded by always-on wearables whether they want to, or even know they are, or not.

Privacy makes up a big part of media law training and exams that qualified journalists (like me) must complete, and always-recording wearables could very easily enable people to break a lot of legal and ethical rules. I expect that without these people necessarily realising they’re doing something wrong.

Move fast and break everything

(Image credit: Shutterstock)

Big tech has always had an ask for forgiveness mentality. Arguably, because time after time, punishments (assuming they are even punished) are usually vastly outweighed by the benefits they reaped by breaking the rules.

This has seemed especially true with privacy, as our data seems to get mishandled by a company every other month – in small, but also sometimes catastrophic ways.

I’m looking at you Tea.

We’ve also already seen examples of AI companies playing fast and loose with copyright, and I expect the rulelessness will only get worse in the AI space as governments across the globe seem less than keen to properly regulate AI so they don’t hamper their country’s efforts to win the digital arms race.

AI wearables capturing every moment of our lives (from multiple angles to boot) with video and audio are a catastrophe waiting to happen.

Yes, there are always promises of privacy, and optional toggles you can switch on to supposedly enhance your data protection. Still, for every good actor that keeps its privacy promises, we can find plenty of companies that don’t – or quietly change them in new ToS you’re asked to sign.

Cooler than expected, just as scary (Image credit: Oakley / Meta)

We can hope that robust regulation and proper punishment for malpractice might come in and help avoid this disaster I foresee, but I’m not holding my breath.

Instead, I’m coming to terms with the demise of privacy – a concept already on its last legs – and accepting that while Big Brother might look different from how George Orwell pictured it, it will (as predicted) be watching us.

You might also like

• Meta's next wearables announcement might include a smartwatch
• These AI smart glasses are giving me a feature I’ve been desperate for the Meta Ray-Bans to get since launch
• I adore my Meta Ray-Bans, but these new Oakley smart glasses are making me jealous

• Alone – Charity Multipurpose Non-profit WordPress Theme has a 9.8/10 flaw
• The bug allows crooks to create rogue admin accounts
• More than 120,000 takeover attempts already blocked

The "Alone – Charity Multipurpose Non-profit WordPress Theme", a commercial theme used in many WordPress websites, contained a critical vulnerability that allowed threat actors to completely take over the website, experts have warned.

The WordPress theme, designed for charities, NGOs, and fundraising campaigns, features more than 40 ready-to-use demos, donation integration, and compatibility with Elementor and WPBakery.

According to Themetix, around 200 active WordPress sites are running this theme today.

Ongoing attacks

Wordfence researchers claim exploitation started on July 12, two days before the vulnerability was publicly disclosed. So far, the company blocked more than 120,000 exploitation attempts from almost a dozen different IP addresses.

In the attacks, the threat actors try to upload a ZIP archive with a PHP-based backdoor that grants them remote code execution capabilities, as well as the ability to upload arbitrary files. Crooks also used the flaw to deliver backdoors that can create additional admin accounts.

All versions up to 7.8.3 contained a vulnerability that allowed threat actors to upload arbitrary files, including malware that can create admin accounts. That way, crooks can completely take over websites and use them to host other malware, redirect visitors to other malicious pages, serve phishing landing pages, and more.

The vulnerability is now tracked as CVE-2025-4394, and has a severity score of 9.8/10 (critical). It was addressed in version 7.8.5, which was released on June 16, 2025. If you are using this theme, it would be wise to update it as soon as possible, since the bug is being actively exploited in the wild.

WordPress is generally considered a safe website builder platform, but third-party themes and plugins - not so much. That is why security pros advise WordPress users to only keep the plugins and themes they actively use, and to make sure they are always up to date.

Via The Hacker News

You might also like

• A popular WordPress theme has been hijacked by malware - here's what we know
• Take a look at our guide to the best authenticator app
• We've rounded up the best password managers

• AMD's head of client CPUs says it's looking into dedicated NPU accelerators
• These would be the equivalent of a discrete GPU, but for AI tasks
• Such boards would lessen demand on higher-end GPUs, as they'd no longer be bought for AI work, as they are in some cases

AMD is looking to a future where it might not just produce standalone graphics cards for desktop PCs, but similar boards which would be the equivalent of an AI accelerator - a discrete NPU, in other words.

CRN reports (via Wccftech) that AMD's Rahul Tikoo, head of its client CPU business, said that Team Red is “talking to customers” about “use cases” and “potential opportunities” for such a dedicated NPU accelerator card.

CRN points out that there are already moves along these lines afoot, such as an incoming Dell Pro Max Plus laptop, which is set to boast a pair of Qualcomm AI 100 PC inference cards. That's two discrete NPU boards with 16 AI cores and 32GB of memory apiece, for 32 AI cores and 64GB of RAM in total.

To put that in perspective, current integrated (on-chip) NPUs, such as those in Intel's Lunar Lake CPUs, or AMD's Ryzen AI chips, offer around 50 TOPS - ideal for Copilot+ PCs - whereas you're looking at up to 400 TOPS with the mentioned Qualcomm AI 100. These boards are for beefy workstation laptops and AI power users.

Tikoo observed: "It’s a very new set of use cases, so we're watching that space carefully, but we do have solutions if you want to get into that space - we will be able to."

The AMD exec wouldn't be drawn to provide a hint at a timeframe in which AMD might be planning to realize such discrete NPU ambitions, but said that "it's not hard to imagine we can get there pretty quickly" given the 'breadth' of Team Red's technologies.

(Image credit: Future / John Loeffler)Analysis: potentially taking the pressure off high-end GPU demand

So, does this mean it won't be too long before you might be looking at buying your desktop PC and mulling a discrete NPU alongside a GPU? Well, not really, this still isn't consumer territory as such - as noted, it's more about AI power users - but it will have an important impact on everyday PCs, at least for enthusiasts.

These standalone NPU cards will only be needed by individuals working on more heavyweight AI tasks with their PC. They will offer benefits for running large AI models or complex workloads locally rather than on the cloud, with far more responsive performance (dodging the delay factor that's inevitably brought into the mix when piping work online, into the cloud).

There are obvious privacy benefits from keeping work on-device, rather than heading cloud-wards, and these discrete NPUs will be designed to be more efficient than GPUs taking on these kinds of workloads - so there will be power savings to be had.

And it's here we come to the crux of the matter for consumers, at least enthusiast PC gamers looking at buying more expensive graphics cards. As we've seen in the past, sometimes individuals working with AI purchase top-end GPUs - like the RTX 5090 or 5080 - for their rigs. When dedicated NPUs come out from AMD (and others), they will offer a better choice than a higher-end GPU - which will take pressure off the market for graphics cards.

So, especially when a new range of GPUs comes out, and there's an inevitable rush to buy, there'll be less overall demand on higher-end models - which is good news for supply and pricing, for gamers who want a graphics card to, well, play PC games, and not hunker down to AI workloads.

Roll on the development of these standalone NPUs, then - it’s got to be a good thing for gamers in the end. Another thought for the much further away future is that eventually, these NPUs may be needed for AI routines within games, when complex AI-driven NPCs are brought into being. We've already taken some steps down this road, cloud-wise, although whether that's a good thing or not is a matter of opinion.

You might also like

• Need a new GPU? Nvidia's RTX 5000 Super models may be coming sooner than you expect
• Nobody wants 8GB GPUs from Nvidia and AMD – and this retailer just made that clear
• Where to buy AMD Radeon RX 9070 XT and RX 9070: the best retailers in the US and UK to check for stock

OLED vs mini-LED is a battle that’s been raging on for a few years now. While LG, Sony and Panasonic sit firmly in the OLED camp, Samsung, Hisense and TCL place their trust more in mini-LED. Granted, most of these brands have both TV types in their lineups, but it’s clear which one each favors.

Samsung and LG are makers of some of the best TVs, and you’ll regularly find Samsung at the top of our best mini-LED TVs list and LG at the top of the best OLED TVs. Each has had a strong showing with their flagship models in 2025, with both the Samsung QN90F (mini-LED) and LG G5 (OLED) earning five stars in their reviews. But what about the mid-range models?

I recently tested the Samsung QN80F and gave it four stars in my review, mainly due to its average sound quality, reflective screen and a sometimes inconsistent picture. I also had a chance to evaluate it side-by-side with LG’s most entry-level OLED: the LG B5. While these TVs sit in different places in their respective lineups, they’re priced very similarly, with the 55-inch QN80F selling for $1,299 / £1,199 and the 55-inch B5 for $1,299 / £1,399.

So, how does the QN80F fare against the ‘cheapest’ 2025 OLED on the market, and which TV is worth the money?

Color me impressed

Both the Samsung QN80F (left) and LG B5 (right) have great color reproduction, and while the QN80F may be brighter, the B5's picture has greater depth and clarity. (Image credit: Universal Pictures / Future )

Samsung and LG TVs both have exceptional color reproduction, with colors on Samsung mini-LED models generally looking bright and punchy, and LG OLEDs looking deeper and richer. Putting the QN80F mini-LED next to the LG B5, this was once again the case.

With both TVs set to their Movie picture modes, I watched the Wizard & I scene from a 4K Blu-ray of Wicked, and the B5 demonstrated bolder and more engaging colors. Pink flowers, blue details on walls and uniforms, and Elphaba’s green skin all looked more vivid on the B5, primarily due to the B5’s stronger contrast. Colors still appeared bright and vibrant on the QN80F with the same scene, but they didn’t have the same depth.

Measuring UHDA-P3 color gamut on both TVs yielded results of 93.05% on the Samsung QN80F and 99.5% on the LG B5, so it’s no real surprise that the B5 had the edge when it came to color depth and detail.

Brightness and reflections

The Samsung QN80F (left) has both higher peak and fullscreen brightness than the LG B5, and that can be a real benefit with certain movie scenes (Image credit: Universal Pictures / Future)

Brightness is where the Samsung QN80F takes a big win. While mini-LED is traditionally a brighter display technology than OLED, flagship OLEDs such as the LG G5 and Samsung S95F have made great strides in giving OLED a brightness boost, with both hitting over 2,000 nits. But, then again, these pricey TVs use new, advanced OLED panels (QD-OLED on the S95F, and ‘four-stack’ OLED on the LG G5).

The LG B5 uses a standard W-OLED panel, and when it comes to brightness, it shows. When I measured the B5’s peak HDR brightness, it hit 668 nits in Movie mode, whereas the Samsung QN80F clocked in at 1,106 nits in the same mode. A near-500 nit brightness difference is easy to appreciate, and as Elphaba ran through a wheat field in the same Wizard & I scene from Wicked, the sun appeared brighter on the QN80F thanks to its high HDR brightness in peak highlights, and the wheat itself had a glossier, more eye-catching look.

One thing that couldn’t be ignored in my comparison was how reflective the screens on both TVs were. I anticipated this from the B5, mainly due to its low measured fullscreen HDR brightness (a meagre 131 nits), but I expected the QN80F, which clocked in at 754 nits on the same test, to perform better. In bright room lighting conditions, both of these TVs suffer from distracting mirror-like reflections when watching darker scenes. In dimmed conditions, they perform better, but there can still be visible reflections.

Black levels: OLED’s best friend

Black levels are slightly raised on the Samsung QN80F (left), but look much richer and deeper on the LG B5 (right) (Image credit: Warner Bros. / Future )

Despite a very reflective screen on both TVs and their struggles with bright room viewing, they display decent contrast in most environments. The QN80F’s higher brightness gives a stronger perceived contrast thanks to the higher brightness difference between light and dark tones. But look closer, and it’s clear that contrast and black levels are areas where OLED shines.

Watching the subway fight scene and opening crime scene in The Batman, the B5 demonstrated superb contrast, and while there was a hint of black crush at times, shadow detail was also excellent. I found the QN80F to have solid contrast in isolated viewing during my testing, but next to the B5, there was an obvious winner. The B5’s sharper textures and finer details also gave the picture a more 3D-like effect.

Blacks on the B5 also have that inky quality that OLED fans love. In both The Batman and in Nosferatu, particularly the scenes where Hutler first arrives at Orlock’s castle, the QN80F’s black levels were slightly raised, taking on a grayer tone, whereas the black of the forest and the night sky were truly as foreboding and ominous as intended on the B5.

Final thoughts

Samsung QN80F (left) and LG B5 (right) (Image credit: Future)

Both the B5 and QN80F have a strong feature set, especially for gaming, and the new versions of LG’s webOS and Samsung’s Tizen smart TV platforms are arguably the best to date. But both TVs will require one of the best soundbars to provide a cinematic experience at home, so my comparison all comes down to picture quality.

The QN80F takes a big win for brightness, but I couldn’t help but prefer the B5’s overall picture during my side-by-side test. Its superior colors, deeper blacks and more realistic-looking textures gave it the edge. If I had the money to spend, I’d be picking the LG B5.

You might also like

• I tested LG, Samsung and Sony's elite 2025 OLED TVs side-by-side – here's the one I'd buy with my own money
• I tested Samsung and Sony’s 2025 QD-OLED TVs side-by-side, and it’s a battle for the ages
• Glare-Free vs anti-reflection: This is how Samsung and LG's flagship OLED TV screens fared when I tested them

• Asus updates MyAsus app to fix two new security issues affecting PCs
• Company confirms past issues in routers and software are now fully resolved
• Users urged to update software, use strong passwords, and disable unused services

Asus has released new security patches for its MyAsus software, following fresh reports from security researchers highlighting flaws across multiple services.

The update applies to all compatible desktops, laptops, NUCs, and All-in-One PCs, addressing two vulnerabilities tracked as CVE-2025-4569 and CVE-2025-4570.

The new versions are 4.0.36.0 for x64 systems and 4.2.35.0 for Arm-based computers - and users can update through the MyAsus app or via the Microsoft Store.

Patch now

The CVE-2025-4569 flaw received a CVSS v4.0 score of 7.7, marking it as high severity. CVE-2025-4570 was rated at 6.9, a medium-level risk.

Both are resolved in the update. Asus also confirmed that is has addressed related issues in Armoury Crate, DriverHub, and specific router models, although details were limited. The company says all reported vulnerabilities have now been resolved.

This patch follows a string of vulnerability disclosures this year.

Back in April 2025, the company patched a critical vulnerability in routers using AiCloud, a service that lets users access and share files over the internet.

Tracked as CVE-2025-2492, it had a severity score of 9.2 and could allow remote attackers to execute functions without permission.

Asus urged users to disable AiCloud on unsupported models and update the firmware immediately.

Then in June 2025, Asus also patched CVE-2024-3080, a critical authentication bypass flaw affecting seven popular router models, including ZenWiFi XT8 and RT-AX88U.

That update also covered a buffer overflow issue and a system command execution bug, both tied to admin privileges.

In a related incident, a high-severity flaw tracked as CVE-2023-39780 was exploited by attackers to build a botnet using compromised routers.

That vulnerability allowed persistent access via non-volatile memory.

Asus responded by urging users to reset affected devices and use stronger credentials.

It's important to keep your router secure - regardless of the make - by checking firmware regularly, updating software, and limiting remote access features whenever possible.

You might also like

• These are the best Asus routers you can buy right now
• Thousands of Asus routers hacked to create a major botnet
• FBI warns outdated routers are being hacked and hijacked for criminal purposes

• Weapons director Zach Cregger admits he's feeling the pressure after the success of his previous movie, Barbarian
• But he told TechRadar that he "feels secure" in the movie he's made
• Weapons is in theaters from August 8

Zach Cregger terrified us all in 2022 with his horror hit Barbarian and he is set to do it again with the new movie Weapons.

It's not easy following up from a breakout movie. With Barbarian scoring an impressive 92% on Rotten Tomatoes from the critics, horror fans are hoping for the same standard when it comes to Weapons.

In Cregger's latest movie, we follow the mysterious disappearances of multiple children who ran away from their homes at exactly 2:17am. Weapons viral marketing campaign had everyone talking, and now the pressure is on for it to be as good as Barbarian.

Speaking to TechRadar, Cregger told me about how he was feeling ahead of Weapons' release. He said: "I'm nervous just because I love the movie so much, and I want people to go see it, but I feel very confident because I feel secure in the movie that I made. "

"I love Weapons. I think it's a bigger riff on Barbarian. This is more and in a good way. I feel like I can sleep soundly at night no matter what happens, knowing like I did my job, but, you know, you hope people show up. "

Weapons is also Cregger's first time scoring a movie alongside two of his friends, so fans will get to hear some of his music on this project.

He told me: "I did it with two of my best friends who I was in a band with in high school, and we just kind of have the same musical mind and so the three of us, Hays and Ryan Holladay and myself, we kind of scored it.

"We scored the movie before I shot it, and then when I was editing it, you know, we just laid the score in."

One of the things I loved about Barbarian was the one word, impactful title, something that he did again with Weapons. I asked about the title and he said : "I think it's such a mysterious thing and with the mystery, it's just incredible. It just feels right."

You might also like

• Weapons has some of the best viral marketing I’ve ever seen, but the new horror movie’s director admits he ‘was not a believer at the beginning’
• Alien: Earth's official trailer is here – and the sci-fi horror series' xenomorph isn't the only frightening creature that'll need to be avoided
• Shudder’s new indie sci-fi shocker is one of the most unsettling horror movies I’ve seen in years

• Square Enix has announced The Adventures of Elliot: The Millennium Tales and Octopath Traveler 0
• The Adventures of Elliot arrives in 2026 for Nintendo Switch 2, PS5, Xbox Series X, Xbox Series S, and PC
• Octopath Traveler 0 is exclusive to Switch and Switch 2 and is releasing on December 4

To close out the July 31 Nintendo Direct Partner Showcase, Square Enix announced two new HD-2D role-playing games (RPGs), The Adventures of Elliot: The Millennium Tales and Octopath Traveler 0.

The Adventures of Elliot: The Millennium Tales is an all-new action-adventure RPG set to arrive in 2026, for the Nintendo Switch 2, PS5, Xbox Series X, Xbox Series S, and PC.

Although the game isn't releasing until next year, Switch 2 players can download a free 'Debut Demo' today that provides an early look at the game.

Developed by Team Asano, the developer behind Octopath Traveler and Bravely Default, The Millennium Tales tells the story of Elliot in a fast-paced adventure that combines thrilling real-time action and rewarding exploration.

Set in the world of Philabieldia, humanity’s last bastion is the Kingdom of Huther, protected by its tall walls and Princess Heuria’s powerful magic.

"When a mysterious set of ruins is discovered, a young adventurer named Elliot and his fairy companion Faie are sent off on a journey to investigate. What seems like a simple, if dangerous, mission soon becomes a grand saga that will be woven across time and space," Square Enix describes.

As Elliot, the player will explore an expansive world equipped with multiple weapon types, like swords, bows, chains, and sickles, which can be customized to increase stats. Faie will also have a 'Sprint' power that lets Elliot zip around faster.

The final reveal of the showcase was Octopath Traveler 0, which is the newest entry in the Octopath Traveler series, coming to Nintendo Switch and Nintendo Switch 2 on December 4, 2025.

"Start from zero and discover the newest entry in the OCTOPATH TRAVELER series. Experience a story of restoration and retribution over the divine rings—an epic saga that unfolds across the realm of Orsterra," the game description reads.

This new entry brings all-new features and returning core elements from the series, including the HD-2D graphics that blend retro pixel art and 3D CG art, Path Actions, the Break and Boost battle system, and more.

For the first time in the series, players will be able to play as their own customized protagonist and "rebuild their hometown after great destruction is brought to it".

Physical pre-orders for Octopath Traveler 0 are now available, with digital pre-orders coming later.

You might also like...

• The 14 best Nintendo Switch 2 games to play in 2025
• The Nintendo Switch 2 is the company’s least ambitious console to date, but its improvements are astronomical
• I’ve spent 150 hours with The Legend of Zelda: Breath of the Wild, and the Switch 2 Edition is an incredible upgrade

• FTX shut down in 2021, but continued to generate data in the backend
• Contact information on more than 35,000 FTX Japan users found leaking
• The company could face regulatory pressure as a result

FTX Japan, the Japanese arm of the dead cryptocurrency exchange FTX, leaked sensitive data on more than 35,000 users, putting them at risk of phishing and identity theft, experts have warned.

Researchers at Cybernews said they found an exposed database with 26 million files, including usernames and real names, email addresses, postal addresses, FTX account IDs, and detailed transaction logs including information on borrowing and lending, cryptocurrencies, collateral, margin rates, and risk flags.

The files are relatively fresh, too, as some of the logs were apparently generated in July 2024.

Troubling implications

Explaining how it’s possible that an exchange that shut down in late 2021 was still generating, and thus leaking files, as early as 2024, Cybernews said FTX Japan completed its bankruptcy and withdrawals in February 2023, but its backend systems probably remained active throughout 2024.

Following the fallout of FTX, its Japanese subsidiary was acquired by another Japanese crypto exchange called bitFlyer, and was rebranded to Custodiem in 2024.

“It is unclear whether the discovered leak belongs to the actively used Custodiem infrastructure, or is an abandoned, unmodified artifact remaining after the FTX collapse,” the Cybernews researchers explained.

The implications are troubling, since cybercriminals can use the information to target people who already lost a lot in the bankruptcy. For example, Celsius customers (another crypto company that went bankrupt at approximately the same time), are being bombarded with phishing emails in which crooks impersonate the company and claim the victims are eligible for withdrawals.

At the same time, the company itself is risking further regulatory pressure, and possibly fines, as a result.

Cybernews also said that the data leak raises concerns about privacy and regulatory compliance, since under Japanese laws, crypto firms need to uphold to high standards.

How to stay safe

The breach means cybercriminals could have a field day with the leaked data, which should be more than enough of sensitive information to launch highly personalized, successful phishing campaigns, leading to identity theft, wire fraud, and even ransomware attacks.

If you're concerned you may have been caught up in the incident, don't worry - there are a number of methods to find out. HaveIBeenPwned? is probably the best resource only to check if your details have been affected, offering a run-down of every big cyber incident of the past few years.

And if you save passwords to a Google account, you can use Google's Password Checkup tool to see if any have been compromised, or sign up for one of the best password manager options we've rounded up to make sure your logins are protected.

You might also like

• Ingram Micro confirms ransomware attack, internal systems affected and shut down
• Take a look at our guide to the best authenticator app
• We've rounded up the best password managers

• Weapons director Zach Cregger has admitted he "was not a believer" when it came to the movie's viral marketing
• However, the realistic marketing campaign has gone down well with fans
• The new movie is in theaters from August 8

Weapons is the hugely anticipated new movie from Zach Cregger, and while fans may be thoroughly spooked by the marketing they've seen so far, the director has revealed he was nervous at first.

The new horror movie, which arrives in cinemas on August 8, centers around a group of children who simultaneously disappear at 2:17am.

Speaking to TechRadar, Cregger said: "They came to me and they were like, 'we're thinking of doing this'. They had a term for it that I wasn't even aware of, but they're like, 'we're gonna make this website, we're gonna make it seem like it really happened.'

"And I'll be honest, I was like, 'I don't think I like this idea because we're trying to trick people', and then are people gonna get annoyed and they realize it's just a movie and they're gonna be like f*** this movie."

He added: "That was my fear, but they were like just trust us. So I was like, 'OK!', and then sure enough, everybody really responded to it, so they get all the credit, you know, and I, and I love what they did. I love it, but I was not a believer at the beginning, I'll tell you."

The website Cregger is referring to is MaybrookMissing.net, which is set up to look like a real missing person's case. There's teasers for the movie, fake CCTV footage and even a reference to his previous movie Barbarian for eagle-eyed fans.

Weapons has really gone hard on this marketing technique, which is known as an Alternate Reality Game. Movies such as Cloverfield and District 9 have adopted this approach in the past, setting up websites and hotlines to entice audiences into a story.

In my opinion, they work incredibly well with horrors or mysteries and it really got me hyped to see the movie, so it's a job well done.

This led me to ask about fan theories based on the marketing campaign, as I was sure people have drawn their own conclusions about what to expect from it. Cregger told me: "Oh yeah, I've seen a lot of theories and I don't even really want to comment on them because if I discredit some of them, then it's, I'm giving people ammo for for for spoiling it.

"But the one that I will say that I've seen a lot that I'm always just kind of like, 'wow', is that people say like an alien is involved in this, and I think that's probably my favorite fan theory."

With plenty of intrigue here, you'll have to go and see the movie if you want to unravel the mystery.

You might also like

• 7 new horror movies on Prime Video, Shudder, HBO Max and more in July 2025
• Alien: Earth's official trailer is here – and the sci-fi horror series' xenomorph isn't the only frightening creature that'll need to be avoided
• Five Nights at Freddy’s 2 has a trailer that’s even more discouraging than the first movie, but die-hard fans will be pleased with one detail

• Developers who published projects on PyPI with their email in package metadata are being targeted
• They are asked to "verify" their email address with a fake PyPI platform
• The "verification" process relays login credentials to attackers

Python developers are being targeted with dangerous phishing attacks, The Python Software Foundation (PSF) has warned .

PSF said threat actors were actively targeting developers who have published projects on PyPI with their email in package metadata. These developers are receiving emails asking them to “verify” their email address on the platform, providing a link to do so.

Clicking on the link redirects the victims to a page that looks seemingly identical to the original one - the URL for the original one is PyPI.org, and for the spoofed one - PyPJ.org, a difference small enough to pass under some people’s radars. This type of fraud is called “typosquatting” and is often used in attacks.

Disrupting the scam

The site looks almost the same as the real thing, and prompts the users to log into their accounts. However, sharing the credentials just relays them to the attackers, who can then log into the actual site, and tamper with the packages found there.

PSF is a nonprofit organization that manages and supports the Python programming language, and operates The Python Package Index (PyPI.org), the most popular package index for the programming language in the world.

Tainting legitimate PyPI packages with malware is also a common occurrence. Many Python developers trust the platform, and use the code found there in various projects. By downloading malicious packages, they can grant attackers access to their projects, and possibly even sensitive company files.

To tackle the impersonation campaign, PyPI admins added a banner to the homepage, and have reached out to CDN providers and name registrars to terminate the phishing sites.

Python developers who received such emails are advised not to click on any links, and just delete the emails immediately. Those who are unsure if the email they received is legitimate or not are advised to open up PyPi directly in their browser, instead of clicking any links in the email.

Via BleepingComputer

You might also like

• Malicious Python packages are stealing vital data, and have been downloaded thousands of times already
• Take a look at our guide to the best authenticator app
• We've rounded up the best password managers

• Kail Audio HP-1 are wireless cans with instant switching of sound profiles
• Sonic profiles for bass heavy and popular consumer headphones
• Launching at $199 / £199 / €199

If you make audio for other people to listen to, you'll know how much of a pain it is to have to listen on multiple things – but of course you have to do it, because what sounds good on neutral studio monitors may be too bassy on a pair of Beats or unclear on AirPods.

That means accumulating more earbuds and over-ears than you have ears, which is expensive and in my case, also very annoying because I keep forgetting to charge any of the wireless ones.

Kali Audio may have a better option.

Fancy a set of reference over-ear wireless headphones that can switch to emulate the sound of bassy boomers or airier options? That's what the new HP-1 promise to deliver, albeit without naming any specific headphone models. They're a single pair of headphones with a triple-split personality.

(Image credit: Kali Audio)Kali Audio HP-1: key features and price

The Kali Audio HP-1 headphones are over-ear closed-back headphones with 40mm drivers, a promised 18Hz to 22kHz frequency range, and custom digital signal processing profiles. There are three different voicings: studio, bass-heavy and consumer, all accessible instantly via the press of a button.

There's a 3.5mm cable (essential for latency-free music production) and the headphones also have Bluetooth for more relaxed listening, although that only has the basic standard SBC and AAC codecs, rather than aptX or LDAC.

Battery life is promised to be 40 hours, and there's active noise cancellation for when you need to take your show on the road.

The first profile, studio, is the standard one for these headphones, and it's designed to deliver transparent and accurate sound when you're mixing or mastering just like any other pair of studio headphones.

Press a button and you'll switch to the bass-heavy mode, which emulates the sound of "headphones that are popular with hip-hop and EDM creators and fans." The difference between fairly flat headphones and the more prominent low-end of such headphones can be quite considerable, so this setting should help keep the bass tight rather than overly boomy.

The third voicing is consumer, which according to Kali Audio "replicates the sound of popular headphones sold alongside phones and computers", which is an impressively phrased way of telling me you mean AirPods without telling me you mean AirPods.

I haven't tried these headphones, so I can't say how they compare to the headphones they aim to emulate. But if Kali has cracked the sound profiles, these could be a good option for those of us who can't afford to buy a library of headphones purely for mixing. With an official of $199 / £199 / €199 (about AU$400) the Kali Audio HP-1 are well within reach for bedroom producers as well as ones with bigger budgets.

You might also like

• Link 1
• Link 2
• Link 3

• Monster Hunter Stories 3: Twisted Reflection has been announced at the Nintendo Direct Partner Showcase
• The third game in the series is coming to Nintendo Switch 2, PS5, Xbox Series X, Xbox Series S, and PC via Steam
• The game has a 2026 release window

Capcom has announced the next game in its Monster Hunter Stories turn-based series, Monster Hunter Stories 3: Twisted Reflection.

Revealed during the July 31 Nintendo Direct Partner Showcase, Monster Hunter Stories 3 is coming to Nintendo Switch 2, PS5, Xbox Series X, Xbox Series S, and PC via Steam in 2026.

Capcom is calling the game "the next evolution of the turn-based, story-rich RPG series," and we got a first story trailer with an art style reminiscent of The Legend of Zelda: Breath of the Wild, teasing an open world, a new cast of characters, and Rathalos.

"Twin Rathalos, born in a twist of fate, ignite the long dormant flames of war between two fractured nations: Azuria and Vermeil," the trailer description reads.

"You are the heir to the kingdom of Azuria, and the sole Rathalos Rider in the kingdom. On the brink of war and amidst strange phenomena threatening the world, a life changing event propels you beyond the Meridian, and on an unforgettable journey in search of the truth."

Twisted Reflection follows the second game in the series, Wings of Ruin, which was released in 2021.

It will also be the first Monster Hunter Stories title to arrive on Nintendo Switch 2 since the console's release, and although the details haven't been confirmed, 60FPS is a possibility on the Switch 2 version.

There's no release date just yet, but the official website is now live.

You might also like...

• The 14 best Nintendo Switch 2 games to play in 2025
• The Nintendo Switch 2 is the company’s least ambitious console to date, but its improvements are astronomical
• I’ve spent 150 hours with The Legend of Zelda: Breath of the Wild, and the Switch 2 Edition is an incredible upgrade

The Zootopia 2 trailer has given us our first look at Judy Hopps (Ginnifer Goodwin) and Nick Wilde’s (Jason Bateman) next adventure almost a decade after the original movie was released. In a nutshell, the Disney movie will follow the pair hot on the trail of a mysterious snake who’s suddenly shown up in town, leading them to go undercover in parts of Zootopia they’ve never explored before.

Listen, we said ourselves that Zootopia deserved a sequel, but surely not like this. As you can tell from the premise above, Zootopia 2 is already setting itself up to be a total yawnfest. I’m not sure how much that matters to the target audience of under 10s, but hey, is there a more restless demographic than them? Regardless, I think our first few minutes of footage in the Zootopia 2 trailer highlight a real issue with Disney’s lack of creative engagement.

Following in the footsteps of Turning Red, Luca and Inside Out 2 (yes, I appreciate these are collaborations with Pixar), it feels as though the studio has regressed back to its cookie-cutter cash grab approach it took the minute it started churning out multiple streams of sequels and live-action adaptations. It’s not only the lack of imagination with the storyline that’s enraged me either – there was an obvious way to be much more playful with the title.

The Zootopia 2 trailer is lacklustre at best, but why isn’t the new movie called 2ootopia?

Disney, you can have this one on me. “2ootopia” (pronounced two-topia) would have been such a lightheartened way to inject some fun back into what is already proving to be a bit of a lazy launch. The internet often jokes about sequels adding their movie number into their title, and aside from us being genuinely hilarious on social media, it’s an easy win as far as kids films are concerned. Young’uns need something big and bold to grab onto, and that starts with the first impression a movie makes.

Surely I shouldn’t be telling the most famous animation company in the world how to do something as basic as their actual job, but if the best they have for Zootopia 2 is setting a snake loose in town only to be caught again, their heads clearly need a wobble. It’s certainly a nice touch to include a therapy session as a gentle way to poke fun at how we perceive ourselves – it’s definitely the strongest scene we’ve seen so far – but how much is that just superficial lip service? If you want to play it safe, it’s better to give the illusion of doing something different, and that’s exactly what Disney has done here.

I love Judy Hopps more than life itself, so I’m really hoping I’m wrong with my first-take assessment of the Zootopia 2 trailer. The fact I’ve not got a single interesting thing to say about it isn’t the best start, but at least there’s always time to rebrand to 2ootopia.

Zootopia 2 is set to be released on November 26, 2025.

You might also like

• When might Marvel release The Fantastic Four: First Steps on Disney+?
• Only Murders in the Building season 5 finally confirms a release date on Disney+ and Hulu – and it’s the show’s most impressive achievement yet
• Daredevil: Born Again season 2 has already wrapped filming – and I know when the Marvel TV show will return on Disney+

The enterprise software playbook seemed clear: everything moves to the cloud eventually. Applications, databases, storage: they all followed the same inevitable arc from on-premises to software-as-a-service.

But with the arrival and boom of artificial intelligence, we’re seeing a different story play out, one where the cloud is just one chapter rather than the entire book.

AI systems

AI workloads are fundamentally different beasts than the enterprise applications that defined the cloud migration wave. Traditional software scales predictably, processes data in batches, and can tolerate some latency.

AI systems are non-deterministic, require massive parallel processing, and often need to respond in real-time. These differences reshape the entire economic equation of where and how you run your infrastructure.

Take the challenge of long-running training jobs. Machine learning models don't train on schedule; they train until they converge. This could be hours, days, or weeks. Cloud providers excel at providing infrastructure at short notice, but GPU capacity at hyperscalers can be hard to get without a 1 year reservation.

The result is either paying for guaranteed capacity you might not fully use, or risking that your training job gets interrupted when using spot instances to reduce costs.

Then there's the inference challenge. Unlike web applications that might see traffic spikes during Black Friday, AI services often need to scale continuously as customer usage grows.

The token-based pricing models that govern large language models make this scaling unpredictable in ways that traditional per-request pricing never was. A single customer query might consume 10 tokens or 10,000, depending on the complexity of the response and the size of the context window.

Hybrid approaches

The most intriguing development involves companies discovering hybrid approaches that acknowledge these unique requirements rather than abandoning the cloud. They're using on-premises infrastructure for baseline, predictable workloads while leveraging cloud resources for genuine bursts of demand.

They're co-locating servers closer to users for latency-sensitive applications like conversational AI. They're finding that owning their core infrastructure gives them the stability to experiment more freely with cloud services for specific use cases.

This evolution is being accelerated by regulatory requirements that simply don't fit the cloud-first model. Financial services, healthcare, and government customers often cannot allow data to leave their premises.

For these sectors, on-premises or on-device inference represents a compliance requirement rather than a preference. Rather than being a limitation, this constraint is driving innovation in edge computing and specialized hardware that makes local AI deployment increasingly viable.

Infrastructure strategies

The cloud providers aren't standing still, of course. They're developing AI-specific services, improving GPU access, and creating new pricing models. But the fundamental mismatch between AI's resource requirements and traditional cloud economics suggests that the future won't be a simple rerun of the SaaS revolution.

Instead, we're heading toward a more nuanced landscape where different types of AI workloads find their natural homes. Experimentation and rapid prototyping will likely remain cloud-native. Production inference for established products might move closer to owned infrastructure. Training runs might split between cloud spot instances for cost efficiency and dedicated hardware for mission-critical model development.

The approach represents a step toward infrastructure strategies that match the actual needs of AI systems rather than forcing them into patterns designed for different types of computing

The most successful AI companies of the next decade will likely be those that think beyond the cloud-first assumptions and build infrastructure strategies as sophisticated as their algorithms.

We've featured the best cloud storage.

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

• The Lenovo Legion Go 2 prototype outperforms the MSI Claw A8 in new benchmark video
• The MSI Claw A8 featured in the video has 16GB of RAM, while the Legion Go 2 has 24GB
• The Legion Go 2 prototype is available on second-hand markets in China, but is yet to launch officially

As we approach the latter stages of 2025, several new contenders are about to enter the handheld gaming PC market – and one of them in particular, the Lenovo Legion Go 2, looks like it will make a big impression.

As highlighted by Notebookcheck, the Lenovo Legion Go 2 outperforms the MSI Claw A8, according to a new benchmark by Lines Tech on YouTube. In Shadow of the Tomb Raider, the MSI Claw A8 hit an average of 44 fps, while the Lenovo Legion Go 2 had an average of 49 fps, the latter producing 10% better performance.

The Lenovo Legion Go 2 is still in its prototype phase, but is already available on second-hand markets in China. Both the Legion Go 2 and the MSI Claw A8 utilize AMD's Ryzen Z2 Extreme processor, so it may be a bit confusing to see the former managing better performance.

However, it more likely comes down to the amount of RAM available; the MSI Claw A8 has 16GB of RAM, while the Legion Go 2 has 24GB, which is 8GB more. It's also worth noting that the Legion Go 2 appears to have a placeholder CPU name, but we know it has the iGPU that's in the Ryzen Z2 Extreme, which is the Radeon 890M.

For the MSI Claw A8, there has strangely been been little to no mention of other configurations featuring less RAM than the marketed 24GB (according to the specifications on its website), which should specify 'up to 24GB', but that's beside the point.

Ultimately, based on these benchmarks, the Lenovo Legion Go 2's top configuration (featuring 32GB of RAM) should outperform the MSI Claw A8, which will be limited to a maximum of 24GB of RAM. We'll need to see both handhelds and their 24GB configurations go toe-to-toe, and if Lenovo's handheld still comes out as the victor, it could simply be due to better cooling.

Analysis: None of this will matter if availability and pricing are an issue...

Having seen the performance benchmarks of the MSI Claw A8 and personally used its similar counterpart, the MSI Claw 8 AI+, I've already written off a purchase of both handhelds. While the performance leap from AMD's Ryzen Z1 Extreme is evident, it's not enough to warrant such a huge expense, but that's slightly different with the Lenovo Legion Go 2.

If the Lenovo Legion Go 2 doesn't come with an unreasonable price tag above $1,000, I could see it being a big hit considering all the benefits. It has an 8.8-inch OLED screen powered by the Ryzen Z2 Extreme, and if it wasn't clear already, the OLED display is what gives it an extra edge.

However, I'm worried that potential buyers will be thrown off by pricing – or worse, a lack of availability. Even if I had decided that I wanted to buy an MSI Claw 8 AI+, the stock would disappear before I could even come to a decision.

Fortunately, that's gradually changing with better availability at Currys in the UK, but I could easily see the same stock woes happening with the Claw A8 and the Lenovo Legion Go 2 when it eventually goes on sale.

Tariffs on Taiwan are certainly part of availability issues, as it's where both Lenovo and MSI manufacture these handhelds. Let's just hope it doesn't get out of hand...

You might also like...

• Finally, the MSI Claw A8 is almost here to rival the Switch 2 and other handhelds – just be ready to pay a hefty sum
• I ditched my PC for the MSI Vector A18 HX A9W gaming laptop - here's my verdict after three weeks
• Handheld PC makers are slowly losing touch with Valve's successful Steam Deck template of affordability, and that's very concerning

Cybersecurity and resilience are essential for organizations striving to succeed in our highly connected world. With threats evolving at a breakneck pace, protecting your business requires more than vigilance. It demands a two-pronged strategy. First, proactive measures that detect and stop attacks before they happen. Second, an agile, efficient response plan to recover quickly if a breach occurs.

Attack timelines in the age of AI

One of the most misunderstood aspects of cybersecurity attacks is timing. Cyberattacks don’t begin when they are detected; they begin when access is gained. This could be weeks or even months before detection. Studies suggest that the period between the disclosure and exploitation, known as “dwell time,” has a global median of around ten days.

During this window, attackers operate in stealth mode. They analyze your systems, identify gaps and vulnerabilities, and position themselves strategically for maximum impact. Attackers exploit this period to steal sensitive data, disrupt operations, or deploy malware like ransomware.

AI is now reshaping the timing and nature of these attacks. A report from the UK’s National Cyber Security Centre (NCSC) highlights a growing disparity between organizations equipped to counter AI-enabled threats and those that falter. Worse, AI has the potential to increase the attack surface while simultaneously reducing dwell time. This dual dynamic will challenge even the most prepared defenders.

The cybersecurity challenges of AI

Every technological leap creates the potential for new security gaps, and AI is no exception. The NCSC stresses that improperly integrated AI systems can expose vulnerabilities. Generative AI (GenAI), for instance, introduces risks related to data exposure, manipulation of outputs, sensitive information leaks and even injection attacks that compromise AI tools.

Maintaining the integrity of training and inference data is becoming a growing challenge, especially as the data itself becomes more distributed. Nearly 90% of companies surveyed in Dell’s Global Data Protection Index recognize AI will generate large volumes of data requiring robust protection.

Yet 65% of organizations back up 50% or less of their total AI data, according to Enterprise Strategy Group research. This gap underscores an urgent need for comprehensive data protection strategies across infrastructure, operations and governance.

Proactive defense for staying ahead

The modern world of cyber threats requires smarter, faster responses. Proactive defense rooted in zero-trust principles is critical. Zero-trust is an architectural approach to security versus a product you buy.

It never trusts and always verifies legitimate business use before granting anyone or anything access to resources. This means that users and devices are not trusted by default, even if they’re connected to a permissioned network and even if they were previously verified.

The ability to advance cybersecurity maturity starts by focusing on three core practice areas: Reducing the attack surface, detecting and responding to threats, and recovering from potential breaches.

Reducing the attack surface literally means making it harder for the bad actors to get into your network, move around and cause havoc. This requires a multilayered approach, starting with penetration testing and vulnerability assessments to identify and address potential security gaps that require immediate attention.

Other critical measures include network segmentation, strict access controls, isolating sensitive data and consistently updating software and systems to mitigate exposure to risks.

Yet, no matter how much the attack surface is reduced, breaches can still occur. Therefore, companies must also work on the second pillar - actively identifying and addressing potential security incidents and malicious activities in the earliest stages of a breach.

Managed Detection and Response (MDR) solutions harness the power of AI to monitor systems in real time, detect unusual activity and neutralize threats before they escalate. Whether it’s anomalies like strange login patterns, unusual traffic or tampered data, these systems catch issues early.

When combined with automation, these systems can immediately isolate compromised accounts or devices, preventing malicious actors from moving deeper into the network.

But detection goes beyond surface anomalies. Advanced analytics analyze data to granular levels, identifying even subtle warning signs of potential attacks. These tools can detect when data has been tampered with, signaling a possible breach before damage spreads. This foresight is invaluable, enabling organizations to act well before a threat spirals into a full-blown crisis.

When proactive measures aren’t enough

No defense strategy is foolproof, even with strict security protocols and advanced detection systems in place. Human error, insider threats or highly sophisticated cyberattacks can and will occasionally bypass safeguards. When this happens, recovery becomes the priority. Recovery can be guided by three AI-powered principles: isolation, immutability and intelligence.

An isolated recovery environment protects critical backups by separating them from normal operations. This containment prevents attackers from tampering with the stored data. Additionally, ensuring immutability means that backups cannot be altered, deleted or overwritten, providing a secure foundation for restoration. Automated System Recovery (ASR), for example, is a hardware-based tool that can bring compromised servers back to their last functioning state quickly.

AI-powered intelligence completes the recovery process. Analytical tools can review data for signs of corruption, look for anomalies like unauthorized encryption or mass deletions, and assess damage levels across the entire digital infrastructure. These forensic insights help businesses better understand the severity and scope of an attack while preparing for future resilience.

Navigating the threat landscape

We’re at a pivotal juncture for cybersecurity. Attackers are using AI to outmaneuver defenses more rapidly than ever before, but defenders are equally equipped to use the same technology to strengthen their protection and response strategies. By focusing on reducing vulnerabilities, detecting threats early and empowering teams with recovery mechanisms, organizations can achieve a resilience that’s critical for both surviving and thriving in today’s threat environment.

We've featured the best business VPN.

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