Technology

• Nintendo Switch 2 Edition upgrade pack pricing has been officially revealed
• They seemingly cost either $9.99 / £7.99 or $19.99 / £16.99 depending on the content included
• "Selected" upgrade packs will also be available via Nintendo Switch Online at no extra cost

Pricing for some Nintendo Switch 2 Edition upgrade packs have finally been revealed, putting an end to some of the confusion surrounding these paid enhancements.

Nintendo has now confirmed, via its US and UK online stores, that the Nintendo Switch 2 Edition upgrade packs for both The Legend of Zelda: Breath of the Wild and The Legend of Zelda: Tears of the Kingdom will cost $9.99 / £7.99 each.

Both of these upgrade packs will also be included in the Nintendo Switch Online + Expansion Pack subscription tier at no extra cost. Nintendo has also revealed that "selected upgrade packs" will be featured on the service, which implies it's not going to be all of them.

Elsewhere, Nintendo UK has confirmed upgrade packs for the Nintendo Switch 2 Editions of Super Mario Party Jamboree + Jamboree TV and Kirby and the Forgotten Land + Star-Crossed World will cost £16.99 each. Nintendo hasn't officially revealed US pricing for these upgrades yet, but - as spotted by Nintendo Life - Best Buy has them listed at $19.99 each.

Chances are these two games have more expensive upgrades due to them both featuring sizeable content expansions. Whereas the Zelda games appear to be performance upgrades only, hence the reduced cost here.

It's likely, then, that Metroid Prime 4: Beyond's Switch 2 Edition upgrade pack could also fall in that $9.99 / £7.99 price bracket. Exclusive Switch 2 content hasn't been revealed for this upcoming game yet, but Nintendo has confirmed greater image quality and performance for it at the Nintendo Direct presentation for Switch 2.

I'm personally hoping that upgrade pack pricing doesn't fluctuate any more than this. The cost of the simpler upgrades isn't too bad, but I certainly won't be making a habit of dropping 17 quid on some of these others - even if the additional content is substantial.

You might also like...

• Shopping for a Nintendo Switch 2 on eBay? Beware of deals that aren't worth the paper they're printed on – and that's the point
• Your Nintendo Switch 2 pre-order might not arrive on release day, Nintendo warns
• Hori is releasing another Nintendo Switch 2 camera intended for handheld mode

• Developer Remedy has shared details regarding FBC: Firebreak
• The game will launch on June 17, 2025
• There will be two editions available: standard and Deluxe

Developer Remedy Entertainment has revealed new information about the upcoming co-op first-person shooter FBC: Firebreak in a recent developer livestream.

The game, which is set in the same universe as Control, will launch on June 17, 2025 for PC, Xbox Series X and Series S, and PlayStation 5. It will be available on day one as part of Xbox Game Pass and the PlayStation Plus Game Catalog (for Extra and Premium members).

In addition to giving us the release date, the live stream also showed off a new job - Hot Fix. In Hot Fix, players go up against a supernatural furnace that has gone completely out of control. It features particularly hot environments, challenging players to rely on fans or each other's water cannons in order to cool down.

The stream went on to offer details on the game's pricing and editions. The standard version of FBC: Firebreak will cost $39.99 / £32.99, which seems like a pretty fair price for a multiplayer focused game.

Paired with the availability via Xbox Game Pass and the PS Plus Game Catalog, I don't think it will be a huge struggle to get my friends on board for a session.

There will also be the more premium FBC: Firebreak Deluxe Edition, which comes in at $49.99 / £39.99.

It includes the following in-game bonuses:

• “The Firestarter” Premium Voice Pack
• “The Pencil Pusher” Premium Voice Pack
• Firestarter Armor Set, Apex Revision (Helmet, Body Armor, Gloves)
• Scorched Remnant Double-Barrel Shotgun Skin
• Golden Firebreak Spray
• Classified Requisition: “Firestarter”: A collection of 36 unlockable cosmetic items including weapon skins, sprays, and armor sets

A Deluxe Edition upgrade will be offered to owners of the base game for $10 / £7.

You might also like...

• Nintendo Switch 2 pre-orders live: US retailers to check, UK stock live at The Game Collection, and all the latest updates
• Forza Horizon 5 shines on PS5, but the Realms update left me wanting more
• F1 25’s Braking Point 3 won't be for everyone, but I’m still enjoying its dramatic take on the sport

• M&S facing ongoing disruption after a cyberattack
• The attack has affected contactless payment and Click and Collect systems
• It's still not clear if any customer data is affected

British retail giant Marks and Spencer has had to take some systems and processes offline after suffering a cyberattack which disabled contactless and Click and Collect services in stores.

The disruption has now continued for several days, with many stores still unable to process contactless payments, and Click and Collect now paused in all stores.

The retailer confirmed in a statement that in order to protect colleagues, partners, suppliers, and the business, M&S has “made the proactive decision to move some [of our] processes offline”, which would be consistent with the response to a ransomware attack - although its not yet clear if this is the case.

Get Keeper Personal for just $1.67/month, Keeper Family for just $3.54/month, and Keeper Business for just $7/month.

Keeper generates and stores strong passwords so you never have to remember them again. Don’t let one weak password leave you exposed.

Preferred partner (What does this mean?)View Deal

Retail at risk

Physical stores, the website, and the M&S App are still up and running, but this disruption could be seriously costly for a store as big as this - as operational losses and damage to reputation for stores can be costly.

The retail industry is a common target for cybercriminals, as even a few hours of downtime can cost millions of dollars, making firms more likely to pay a ransom, and therefore more vulnerable.

Earlier in 2025, Walmart membership programme ‘Sam’s Club’ was hit with a ransomware attack that reportedly affected thousands of staff - illustrating the sector’s vulnerability.

“The retail industry is operating on a very small profit margin, and therefore the amount of attention or budget they can give to addressing their cybersecurity posture is usually scarce,” explains Pierre Noel, Field CISO EMEA at Expel.

“To address this, retailers must implement a continuous cyber risk quantification programme. One of its outcomes is to generate and price credible incident scenarios, as well as to identify mitigating controls and their associated costs. This information is very meaningful for senior executives and the board, communicates effectively, and places the responsibility on them to determine which risks are acceptable and which are not.”

You might also like

• Take a look at our picks for the best malware removal software around
• Check out our choice for best antivirus software
• Starbucks has gone back to pen and paper after vendor ransomware attack

• Fresh clues about the 25H2 update have been uncovered
• References in a file mention 25H2 and tie it to preview builds in the 26200 range that were recently kicked off by Microsoft
• As 26200 is a small increment from previous 26100 builds, it is likely the 25H2 update will be a minor affair in the form of an ‘enablement package’

More clues have been picked up suggesting that Microsoft is indeed working on Windows 11 25H2 – speculation which was first fired up last month – and that it’s likely to be a relatively minor update.

Windows Latest noticed that one of the more regular leakers of Microsoft-related info on X, XenoPanther, spotted what’s apparently a reference to the 25H2 update in a recent preview build of Windows 11.

GE25H2 is mentioned in appraiserres.dll26200=FT_ALL_CompatIndicatorHelper_WritingGE25H2April 23, 2025

The reference to ‘GE25H2’ is present in a DLL file, and that stands for ‘Greater than or Equal to 25H2’ and there’s another mention of 25H2 which specifically connects it to the series of builds numbered from 26200.

Windows Latest has verified this, and notes that the ‘appraiser’ DLL in question pertains to checking whether a PC qualifies for the upgrade. In other words, this is part of the code that verifies whether any given system is compatible and okay to have 25H2 installed.

(Image credit: Future / Jasmine Mannan) Analysis: Probably an enablement package for 25H2?

All of this is in theory, of course, as Microsoft hasn’t said that it’s working on Windows 11 25H2 officially, or even mentioned the name at all.

What Microsoft has told us, back in March if you recall, is that it is making “behind-the-scenes platform changes” in the new preview builds in the 26200 range. And as noted above, 26200 is mentioned and tied to 25H2 specifically in this DLL file.

Those changes being made in the background are theorized to be tweaks to the platform that underpins the desktop OS, which was refreshed to a new model called Germanium with Windows 11 24H2. As another leaker, Zac Bowden, informed us last month, it’s very likely that all this is wrapped up with laying the early groundwork for 25H2, which could be a much more minor update compared to 24H2, which was a huge undertaking (with that shift to Germanium).

The change from the previous 26100 builds to the 26200 range is a small increment, suggesting that 25H2 will be an equally scaled-down update. Indeed, as Windows Latest points out, it’ll probably be what’s called an ‘enablement package’ in the same way that 23H2 was built on 22H2. This simply means any new features (doubtless a small number of them) are already in place in Windows 11, and will simply be enabled by the update.

All of this is guesswork at this point, although with this new leak, it seems just a tad more likely that this is how things will unfold.

The potentially good news on 25H2 being a lesser update is that with fewer changes, there should be fewer bugs, too. The 24H2 update has proven seriously problematic with gremlins in the works partly because of all the tinkering going on deep in the guts of Windows 11 that was required to usher in the Germanium platform.

You might also like...

• Windows 11 users get ready for more ‘recommendations’ from Microsoft – but I’m relieved to say these suggestions might actually be useful
• Windows 11 is getting a very handy change to the taskbar, as Microsoft takes a leaf from Apple’s Mac playboo
• Windows 11 fully streamlined in just two clicks? Talon utility promises to rip all the bloatware out of Microsoft’s OS in a hassle-free way

• A new leak claims iPadOS 19 will get a Mac menu bar
• Stage Manager 2.0 will also come to iPad and iPhone, the report claims
• There are reasons to be doubtful of the rumors, however

For years, a small but vocal group of tech fans has been calling for Apple to essentially merge iPadOS and macOS into some sort of hybrid operating system that works on both platforms. Apple has persistently pushed back against this idea, but a new rumor suggests that the two systems could soon get even closer when iPadOS 19 launches later this year.

According to leaker Majin Bu, the main changes will take place when you connect a Magic Keyboard to an iPad running iPadOS 19. When that happens, Bu claims that a menu bar will appear at the top of the iPad’s display, giving you a more Mac-like experience.

As well as that, Bu argues that iPadOS 19 will come with Stage Manager 2.0, an update to the multitasking feature that “activates automatically when the keyboard is attached.” Bu states that the new version of Stage Manager will “make managing apps and windows smoother and more productive than ever,” but doesn’t go into any more detail.

As well as iPadOS 19, Bu touched on what might arrive in iOS 19. There, the leaker claimed that iPhones with a USB-C port – that is, those from the iPhone 15 and iPhone 16 ranges – will be able to run Stage Manager when connected to an external monitor.

This would extend your iPhone display rather than simply mirroring it, as is currently the case. Bu noted that this feature might come with some limitations, “possibly in resolution or the number of apps displayed at once.”

Big questions remain

(Image credit: Daniel Romero / Unsplash)

Majin Bu has a somewhat spotty record when it comes to Apple leaks and rumors, so their latest claims should be treated with a degree of skepticism. But aside from their track record, there are other reasons to be doubtful.

For one thing, Apple has long said that merging elements of iPadOS and macOS will mean both systems miss out, with too many compromises being needed to make the hybrid platform work. With that in mind, directly porting a macOS element like the menu bar across to iPadOS feels a little too close to the kind of thing that Apple has long argued against.

As well as that, adding Stage Manager support to the iPhone feels incredibly niche, as the number of people who will want to use an iPhone for productivity in this way is likely to be very small.

However, might Stage Manager support on iOS be something that is built for the upcoming foldable iPhone? That device will likely have a display around the size of an iPad mini, where better productivity tools might make more sense.

We’ll have to hang tight on that one, as the foldable iPhone isn’t expected to arrive until 2026 or 2027. But while we wait, iPadOS 19 and iOS 19 are set to land at Apple’s Worldwide Developers Conference (WWDC) in June. That could be when we finally see if Majin Bu’s claims are on the money.

You might also like

• iPadOS 19: 4 rumored new features, and 2 I’d like to see
• iOS 19: new features, a new design, and everything you need to know
• How to use Stage Manager on a Mac in macOS Ventura

• Intel didn't lay off 20,000 workers after all, but layers of managers are at risk
• CEO Lip-Bu Tan has also asked teams to remove unnecessary meetings
• Workers will need to be in the office four days a week, up from three

Despite recent rumors that it was planning to lay off thousands of workers, cutting its headcount by around one-fifth, Intel did not announce such measures as part of its most recent quarterly results.

However, not all jobs are safe, with the company's new CEO Lip-Bu Tan criticizing existing and inefficient hierarchical structures, and alluding to future job cuts.

In an open letter to all company employees, Tan also confirmed Intel would be requiring workers to return to the office, adding that managers should remove unnecessary meetings to streamline processes and boost productivity.

Intel workers are in for a change

"We are seen as too slow, too complex and too set in our ways," Tan said, after hearing from customers. The new leader, who has been in post since March 18, 2025, revealed ongoing plans focused on "flattening the organization," including removing "organizational complexity."

Tan explained that many teams are eight or more layers deep, and the constant upstream reporting can slow teams down significantly. Intel's execs have been tasked with "tak[ing] a fresh look at their respective orgs," which is corporate speak for removing some of the unnecessary layers.

"There is no way around the fact that these critical changes will reduce the size of our workforce," Tan added. The Santa Clara chipmaker had already laid off 15,000 employees in August 2024.

In his memo to staff, Tan also drew attentnion to unproductive administrative work and meetings. "Too much valuable time is being wasted," he added.

Tan's final change – the big one – is that workers will have to spend more time in the office. He described workers' adherance to the existing three-day policy as "uneven," implying enforcement could be stricter for the upcoming four-day policy for in-office working, which will come into force on September 1, 2025.

The news came as the company posted flat quarterly revenue compared with the same period last year, at $12.7 billion.

"I am taking swift actions to drive better execution and operational efficiency while empowering our engineers to create great products," the CEO said in a statement.

You might also like

• Worried about the news? These are the best job sites around
• We've also rounded up the best recruitment platforms for hirers
• Boost your skills with the best online learning platforms

• Samsung has accidentally shared Galaxy S25 Edge pricing on its own site
• The leak reveals Canadian pricing, putting it between the S25 Plus and the S25 Ultra
• Elsewhere, leaked marketing text reveals several possible specs, along with the weight and thickness

Most leaks have suggested that the Samsung Galaxy S25 Edge will have a price tag somewhere between the Samsung Galaxy S25 Plus and the Samsung Galaxy S25 Ultra, and now that’s looking almost certain, as Samsung itself has accidentally revealed Canadian pricing.

As spotted by Roland Quandt, the company made the pre-launch blunder in a promotional offer for the Galaxy Tab S10 FE, revealing that the 256GB Galaxy S25 Edge will cost $1,678.99 Canadian dollars, while a 512GB model will set you back $1,858.99.

Price conversions are rarely accurate, so we won’t bother with them here, but the key thing to note is that this starting price puts the Galaxy S25 Edge between the Canadian starting price of the Galaxy S25 Plus ($1,438.99) and the Galaxy S25 Ultra ($1,918.99). So, the same will almost certainly hold true elsewhere.

This lightweight phone is leaky

The Samsung Galaxy S25 Edge (Image credit: Future/Lance Ulanoff)

And this isn’t the only new Samsung Galaxy S25 Edge information, as reputable leaker Evan Blass has shared a screenshot of what reads like official Galaxy S25 Edge marketing text.

While we’d take this with a pinch of salt, the source has a superb track record, and the text mentions things like a 200MP camera, a Snapdragon 8 Elite chipset, all-day battery life, a thickness of just 5.8mm, and a weight of just 163g, most of which we’ve heard before.

For comparison, the standard Samsung Galaxy S25 is 7.2mm thick and weighs 162g, but the Edge is expected to have a much larger screen, so if it only weighs 1g more than the Galaxy S25, that’s still quite impressive.

But at 5.8mm thick, it might not be quite as slim as the iPhone 17 Air, which is reportedly just 5.65mm thick. Whether such a tiny difference will have an impact remains to be seen, though.

We should find out just how slim the Samsung Galaxy S25 Edge feels soon, as according to FN News (via @Jukanlosreve), the phone will be announced on May 13 and ship globally on May 30.

You might also like

• Samsung Galaxy S25 Edge: latest news, rumors, and everything we know so far
• Best Samsung phones: top Galaxy handsets ranked
• Samsung Galaxy Z Flip 7: latest news, rumors, and everything we’ve heard so far

• Elipson Connect 2130 Xi is a powerful streaming amp and DAC
• 2 x 130W, capable of up to 500W peak in mono mode
• AirPlay 2, Chromecast, UPnP and DLNA with Sabre ES9023 DAC

Some audio components are beautiful, with gorgeous details that make it hard to tear your eyes away. The Elipson Connect 2130 Xi amplifier, it's safe to say, is not one of them – its efforts have been put elsewhere.

The French audio brand – Elipson is France's oldest hi-fi manufacturer, and was founded in 1938 – has taken a very simple approach to the industrial design here by not really doing any of it, preferring to concentrate instead on what the amp can do rather than what it looks like. And it can do a lot.

Inside its compact casing the Connect 2130 Xi has 2 x 130W of Class D amplification in normal mode; Wi-Fi and Ethernet for streaming; line-in and RCA inputs; and a subwoofer output too. And it has AirPlay 2, Chromecast, UPnP and DLNA for local streaming playback.

This amp is aimed more at the professional installer market than the home market – which is why it lacks the visual flair of Elipson's distinctive high-end home speakers – but if you're not fussed about its looks, it delivers plenty of power for high-quality streaming.

(Image credit: Elipson) Elipson Connect 2130 Xi: key features and pricing

The DAC inside is an ESS Sabre ES9023, delivering up to 24-bit/192kHz hi-res audio. And the Connect 2130 Xi works with all the usual streaming services – such as Qobuz and Tidal if you want to get the most from its hi-res support, plus Spotify if you, uh, don't.

The amplification has a PFC power factor correction system that Elipson says guarantees harmonic distortion of less than 0.003% and a signal to noise ratio of over 119dB.

You can run the amp in mono mode to put out 250W RMS and 500W peak (at 8 ohms), and it's compatible with 100V loudspeakers for large-space installations. The amp is also rackable for professional installations.

Final pricing hasn't been officially disclosed at the time of writing but some UK specialist retailers are currently listing it for £659, which is very reasonable indeed, and makes it a super-interesting option for powering the beefier options among the best stereo speakers. The US price is expected to be under $700 (about AU$1,244).

You might also like

• TEAC's new hi-fi amplifier is made to power hardcore audiophile systems, but in a super-compact size
• Cambridge Audio's elite EXA100 amplifier covers the past, present and future
• Best stereo speakers 2025: stereo systems for all budgets

The AI landscape is shifting—fast. With DeepSeek delivering ChatGPT-level performance at a fraction of the cost, we’re witnessing the next evolution of AI: one that makes advanced models more affordable, accessible, and industry-specific.

We’ve seen these technology shifts before:

• Mainframes evolved into PCs
• On-prem solutions gave way to the Cloud
• Custom chips were overtaken by GPUs
• Expensive, general-purpose AI is transitioning to more affordable, specialized solutions

Each transition unlocks waves of innovation at the application level. Now, AI is following the same trajectory, and insurance will benefit immensely.

The future isn’t about building massive, expensive foundation models—it’s about developing AI tools that solve real, industry-specific problems.

From General-Purpose AI to Industry-Specific Intelligence

The real impact of AI isn’t in the foundation models themselves—it’s in how businesses apply them. Large language models (LLMs) like GPT-4 are powerful, but they weren’t designed for nuanced, industry-specific tasks. That’s why the next wave of AI innovation will focus on specialized, smaller models that are trained on deep, proprietary data sets rather than general internet knowledge.

In insurance, this means AI that understands policy language, risk factors, and claims trends at a granular level. Instead of generic AI tools that require extensive customization, insurers can now adopt purpose-built AI solutions that deliver more accurate risk assessments, streamline underwriting, and optimize claims processing.

The emergence of DeepSeek highlights this shift. Its ability to deliver OpenAI-level performance at a fraction of the cost is a glimpse into the near future—where AI models are cheaper and more efficient, accelerating the development of vertical AI solutions.

Why the Next AI Unicorns Won’t Be Building Foundation Models

For years, AI has been dominated by companies building massive, foundational models—requiring billions of parameters and millions of dollars to train. But history tells us that when core infrastructure costs drop, application-level innovation takes off.

For example, AWS made cloud computing infrastructure widely available, leading to an explosion of SaaS startups. We’re now seeing the same shift with AI: the real breakthroughs won’t come from those training massive models, but from those applying AI in novel, industry-specific ways.

For insurers, this shift is a game-changer. Instead of investing in generic AI tools that require extensive customization, they can now access affordable, purpose-built AI designed specifically for underwriting, claims processing, and risk management. The next AI unicorns won’t be competing with OpenAI or DeepSeek to build foundation models—they’ll be the companies applying these models to solve real-world problems in insurance and beyond.

The Business Case for Specialized AI in Insurance

With AI becoming more cost-effective, insurers have a unique opportunity to embrace vertical AI solutions that provide immediate, tangible benefits.

For example, AI models tailored for insurance can analyze massive amounts of historical claims data to refine risk assessment and pricing strategies. Traditional AI requires insurers to build custom solutions from scratch—often leading to long, expensive development cycles. With specialized AI, insurers can quickly deploy models that understand their business from day one.

In addition, these industry specific models lead to:

Better Accuracy – AI trained on insurance-specific data provides deeper insights, improving underwriting and claims assessment.

Faster Implementation & ROI – Specialized AI solutions require less customization and can be deployed faster, delivering value more quickly.

Enhanced Decision-Making – AI trained on insurance-specific data can process complex data more effectively, leading to smarter risk management.

Cost Efficiency – As AI infrastructure becomes more accessible, insurers can invest in tailored solutions without the high costs of general AI platforms.

This shift also levels the playing field. Previously, only large insurers with deep pockets could afford AI-driven insights. Now, smaller insurers can harness the power of AI without massive infrastructure investments, making them more competitive in the market.

Looking Ahead: AI’s Role in Shaping the Future of Insurance

DeepSeek marks the shift of the next wave of AI—a wave that prioritizes affordability, accessibility, and industry-specific applications. History shows us that technological advancements not only make better tools, but they also change how industries operate. Just as cloud computing enabled the SaaS revolution, affordable AI models will empower insurers to make smarter decisions, reduce risk, and operate more efficiently.

The bottom line? The next AI revolution won’t be led by massive, generalized models. It will be driven by specialized AI applications that solve real problems in insurance and beyond. The companies that embrace this shift early will be the ones leading the industry into the future.

Final Thoughts

For insurers, this is an opportunity to rethink their approach to AI. Instead of viewing AI as an expensive, broad tool that requires extensive customization, they should be asking: What problems do we need AI to solve?

The answer isn’t a massive, general-purpose model—it’s an AI solution purpose-built for their industry.

DeepSeek has shown that the economics of AI are changing. The question isn’t whether AI will reshape insurance—it’s how quickly insurers will adapt to this new reality.

We've rated the best personal finance 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

• Google's report reveals worrying adoption gaps among UK workers
• The Government should "guarantee AI training" for these workers
• Enabling the workforce to adopt AI effectively would account for half of the economic growth

A new report from Google has claimed the UK could potentially double its adoption of artificial intelligence, resulting in boosted productivity and an estimated £400 billion in economic growth.

The company's latest AI Works report plays into the role of human workers in the successful deployment of AI – effective workforce adoption could account for £200 billion, or half, of the gains.

However, as it currently stands, two in three UK workers have never used generative AI at work – a trend that's most prominent among "older women from lower socio-economic backgrounds."

Effective AI adoption

Google's report centers around the accessibility concerns surrounding artificial intelligence. Women over 55 are said to be four times less likely to use AI than men under 35, the report reveals.

It doesn't all need to be lost hope, though, with the tech giant spelling out a handful of key considerations businesses can explore to boost adoption across all workers.

The primary advisory is that workers need "permission to prompt" – giving workers explicit permission to use AI and setting out clear AI policies will help them to understand that AI's advantages are legitimate, fair and comparable to other internet-based tools.

Google also noted that AI habits are easy to form, revealing that just a few hours of training can double daily AI usage, which remains high several months after that initial lesson. Consequentially, this habit formation leads to further exploration and encourages ongoing learning.

Naturally, the report also explores how AI can save workers time – an average of 122 hours per year across all sectors, according to the figures, which exceeds the previously modelled 100-hour estimate.

Finally, Google says that supporting workers with upskilling efforts can help to narrow adoption gaps.

"We want these pilots to encourage more leaders to act now and seize this opportunity to equip everyone with the skills needed to unlock economic growth and change ways of working for the better," noted Google's EMEA President, Debbie Weinstein.

Google has also urged the UK Government to "guarantee AI training for all public sector workers."

Peter Kyle, Secretary of State for Science, Innovation and Technology, committed to supporting the UK workforce: "We will support workers to develop the skills they need for jobs in and with AI, so that all parts of society can benefit from this technology."

You might also like

• We've listed the best AI tools and best AI writers
• Support your workers by giving them access to the best online learning platforms
• British businesses are getting used to AI at work - but there are still plenty of hurdles to overcome

Hoo boy, it's almost time to deal with the fallout from last week's episode of The Last of Us season 2.

Indeed, with viewers proclaiming that episode 2's shocking moment "ruined my entire week, month, and year", this season's third chapter is sure to be a solemn and anger-filled affair.

So, when will The Last of Us TV show's next episode be available to watch? And where will it be streamable in the US, UK, and Australia? You'll get the answers you seek in this article, as well as a complete rundown of season 2's full release schedule for the weeks ahead. Wipe away the last of those tears, dear reader, and proceed when you're ready.

What is the release date and time for The Last of Us season 2 episode 3 in the US?

Tommy is going to be hurting for many, many reasons in this season's third entry (Image credit: Liane Hentscher/HBO)

The next installment of The Last of Us' sophomore season will air at 6pm PT / 9pm ET on Sunday, April 27.

Just like this installment's first two episodes, you can watch it on Max if you're subscribed to one of the world's best streaming services. Alternatively, those who aren't can catch it on Warner Bros. Discovery's (WBD) linear TV channel – that being HBO.

Where can I watch The Last of Us season 2 episode 3 in the UK?

Ellie is going to need Dina more than ever after what happened to Joel in season 2's last episode (Image credit: Liane Hentscher/HBO)

UK fans can tune into Sky Atlantic and Now TV to watch episode 3 of The Last of Us' second season. It'll be a late night or very early morning for those of you who want to watch it as soon as it drops, too, with the show set to return on Monday, April 28 at 2am BST.

When does the next episode of The Last of Us season 2 air in Australia?

Jesse is as unimpressed as the rest of us over the week-long wait for episode 3 (Image credit: Liane Hentscher/HBO)

One of the best Max shows will return for its latest chapter Down Under on Monday, April 28 at 11am AEST.

With WBD's main streaming platform launching in Australia last month, Max is now the primary home for The Last of Us TV series. However, anyone who's signed up to Foxtel can also stream HBO's live-action adaptation of Naught Dog's multi-award-winning video game series on this service, too.

When do new episodes of The Last of Us season 2 come out?

So long, Joel (Image credit: Liane Hentscher/HBO)

There'll be four more episodes of The Last of Us season 2 to enjoy (if that's the right word to use) before the critically-acclaimed and hugely popular TV program leaves our screens once more. Here's when you can watch that quartet:

• Episode 1 – out now
• Episode 2 – out now
• Episode 3 – April 27 (US); April 28 (UK and Australia)
• Episode 4 – May 4 (US); May 5 (UK and Australia)
• Episode 5 – May 11 (US); May 12 (UK and Australia)
• Episode 6 – May 18 (US); May 19 (UK and Australia)
• Episode 7 – May 25 (US); May 26 (UK and Australia)

You might also like

• The Last of Us season 2 cast and character guide: who's playing who in the popular HBO TV show's next chapter?
• The Last of Us season 2 won't be the hugely successful Max TV show's final entry as HBO officially announces its third season
• First trailer for Predator: Badlands drops two huge clues that it could be a new Alien Vs Predator movie in all but name

From DeepSeek to Anthropic’s Computer Use and ChatGPT’s ‘Operator,’ AI tools have taken the world by storm, and this may be just the beginning. Yet, as AI agents debut with remarkable capabilities, a fundamental question remains: how do we verify their outputs?

The AI race has unlocked groundbreaking innovations, but as development surges ahead, key questions around verifiability remain unresolved. Without built-in trust mechanisms, AI’s long-term scalability — and the investments fueling it — face growing risks.

The Asymmetry of AI Development vs. AI Accountability

Today, AI development is incentivized for speed and capability, while accountability mechanisms lag behind. This dynamic creates a fundamental imbalance: verifiability lacks the attention, funding and resources needed to keep pace with AI progress, leaving outputs unproven and susceptible to manipulation. The result is a flood of AI solutions deployed at scale, often without the safety controls needed to mitigate risks like misinformation, privacy breaches and cybersecurity vulnerabilities.

This gap will become more evident as AI continues to integrate into critical industries. Companies developing AI models are making remarkable strides — but without parallel advancements in verification, trust in AI risks being eroded. Organizations that embed accountability from the outset won’t just mitigate future risks; they’ll gain a competitive advantage in a landscape where trust will define long-term adoption.

AI’s rapid adoption is an incredible force for innovation, but with that momentum comes the challenge of ensuring robust verification without slowing progress. Rather than leaving critical concerns for later, we provide a seamless path to integrate verifiability from the start — so developers and industry leaders can move full speed ahead with confidence. The current AI gold rush has unlocked massive opportunities, and by closing the gap between capability and accountability, we ensure that this momentum not only continues but strengthens for the long term.

Verifiability as a Catalyst for AI’s Future

Recently, many were surprised when one of the largest tech companies in the world pulled the plug on its AI features. But as AI capabilities expand, should we really be caught off guard when verification challenges surface? As AI continues to scale, the ability to prove its trustworthiness will determine whether public confidence grows or diminishes.

Recent surveys indicate that skepticism is rising, with a significant portion of users expressing concern over AI’s reliability. The next evolution of AI requires accountability to grow in tandem with development, ensuring trust scales with innovation.

The future of AI needs to be reframed: The question is no longer just ‘Can AI do this or that?’ but rather ‘Can we trust AI’s outputs?’ By embedding trust and verification into AI’s foundations, the industry can ensure AI adoption continues to expand with confidence.

But to return to the fundamental question at hand: how? More precisely, how do you know if the information generated from AI is accurate? How can the privacy and confidentiality of that information be verified? Anyone using ChatGPT, Copilot, Perplexity or Claude, among countless others, has faced these questions. Addressing them requires leveraging the latest advancements in cryptographic verification.

Enter zkML: A Framework for AI Trust

AI’s ability to generate complex outputs is growing exponentially, but verifying the accuracy, security and trustworthiness of these outputs remains an open challenge. This is where zero-knowledge machine learning (zkML) presents a breakthrough solution.

Zero-knowledge proofs (ZKPs), originally developed for cryptographic security, provide a way to prove the validity of an AI-generated output without revealing the underlying data or model details. By applying these techniques to machine learning, zkML ensures that AI-generated outputs are produced as expected while preserving privacy and integrity.

Inference generated using zkML confirms that AI models operate as intended, while verifiable AI training ensures that the training data remains untampered. Additionally, private input protection allows AI to be leveraged securely without exposing sensitive information, and compliant, confidential AI helps meet regulatory requirements while preserving data confidentiality. This means AI systems can prove their outputs — without disclosing the full details of their processes, including model weights.

Unlike traditional verification methods that rely on centralized oversight or controlled environments, zkML enables decentralized, trustless verification. This allows AI developers to demonstrate the authenticity of their models without requiring external trust assumptions, paving the way for scalable and transparent AI verification.

The Future of AI Trust Hinges on Verifiability

AI’s credibility hinges on its ability to prove its outputs are trustworthy. The industry has an opportunity to integrate verifiability now — before trust erodes.

A future where AI operates without trust mechanisms will struggle to scale sustainably. By integrating cryptographic verification techniques like ZKPs, we can create an AI ecosystem where transparency and accountability are built in, not an afterthought.

Verifiable AI is more than a theoretical solution; it’s the next frontier of AI innovation. The shift toward verifiable AI is not only necessary — it’s the next step in ensuring AI’s long-term success. The time to act is now.

We've listed the best encryption 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

The new financial year is the time for businesses to consider any unnecessary expenditure to ensure financial efficiency. However, as budgets reset, it also presents an opportunity to reassess and streamline supply chain operations.

Conscious buying is on the agenda for many businesses as consumers look to make more thoughtful purchasing decisions. In fact, research has shown that according to consumers, sustainable procurement results in as much as a 15 to 30% increase in brand value.

This is only putting more pressure on businesses to ensure that their suppliers are sustainable and are aligned with their values and own sustainability ambitions.

When it comes to meeting socially responsible commitments, for example, procurement can make or break a business’ sustainability trajectory, with up to 90% of a company's carbon footprint linked to its supply chain. With so much pressure on businesses to be more socially responsible, procurement needs to take the lead on not only ensuring that expectations are met, but that purchasing processes also support the long-term growth of the company against established targets.

The case for more sustainable procurement is strong, but how can businesses reflect this priority in their supply chain? Following recent research from Amazon Business, which reveals that 49% of procurement professionals see the complexity of their supply chain as a major obstacle at work , here are some ways to spring clean your supply chain and align with sustainable growth ambitions for the new financial year.

1. What businesses don't know will hurt them as AI raises the bar for reporting

The complexity of supply chains often leaves valuable purchasing data lying dormant across different tools and platforms. With so much data to manage, it is no wonder that businesses can struggle to keep tabs on inventory, spend, shipping routes and supply chain disruptions.

It is essential that procurement teams have visibility of where exactly their vulnerabilities lie and the impact they might have on the wider business. Put frankly, measurable progress towards sustainable procurement goals is not achievable without the right data, and the tools to make it seen.

AI can provide a major competitive advantage here with the ability to use large language models (LLMs) and cognitive analytics to recognize and predict patterns in purchasing data. Natural language processing (NLP) then allows procurement teams to interact with this data, ask it questions, and pull-out actionable insights.

For example, procurement leaders can use AI tools to estimate the carbon footprint of a product, helping to map its environmental impact and assess its performance against internal criteria for sustainable purchasing – unlocking a new level of data-driven decision making.

It is clear that there is appetite for AI innovation as 96% of procurement professionals have plans to invest in AI tools. By embracing advancements in real-time data analytics, businesses can have access to the data they need to stay one step ahead of changes in their supply chain.

2. A diverse supply chain is a strong supply chain

With pressure on businesses to be more sustainable in their purchasing, an overhauled supply chain should focus on diversifying supplier relationships and investing in partnerships with local and small businesses. In fact, achieving a broad supplier base is cited as a top priority for the majority of UK procurement leaders as a way to ensure resilience in supply and distribution.

By making sure the business is not reliant on a limited pool of suppliers, procurement teams can be more assured of the continued growth and performance of their department, in turn contributing to more reliable results for the business. A diverse supply chain also allows for greater flexibility in the long term, making it easier for businesses to adjust and refine their supplier base as goals and ambitions change.

With this in mind, variety can be an effective solution for managing a volatile global landscape. By encouraging businesses to prioritize more local suppliers and minimize transportation costs, a diverse supplier base also indirectly helps businesses to play a part in the circular economy and cut down on emissions.

3. Stay on track with universal standards and purchasing policies

Whilst a diverse supplier base can increase supply chain resiliency, managing a dispersed network might sound like an overwhelming logistical challenge, adding even more admin to procurement teams’ day-to-day system maintenance. This is where digital tools and platforms such as Amazon’s Business ‘Guided Buying’ feature are transforming the future of procurement, establishing a ‘one-stop shop’ for business purchasing that frees up time and capacity for time-stretched procurement teams.

By setting up simple policies across the organization, ‘Guided Buying’ helps employees identify which sellers are preferred by their organization, as well as sellers that might be restricted or blocked. Visual signposts make this a natural part of the decision-making process and takes the guessing game out of compliance.

Creating clear guidelines for business purchasing can also prove to be a significant time-saver for businesses as reporting requirements become tighter under new and developing regulations such as The Corporate Sustainability Reporting Directive (CSRD) (5). As social value weighting becomes a necessary part of managing bids and tenders, early adopters of purchasing policies will benefit from a more consistent and reliable track record of purchasing that aligns with social value ambitions.

The new financial year will no doubt present procurement teams with a fresh set of challenges and opportunities. By focusing on cleaning up their supply chain, improving supplier diversity and organizing internal processes, businesses can put themselves in the best position to proactively respond to these changes. Making an investment now in a transparent, resilient and data-forward procurement strategy, will pay off for businesses in the long term, unlocking savings in cost, sustainability, and efficiency.

We've listed 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

• Perplexity AI has brought a new voice assistant to iOS
• The assistant can open apps like OpenTable or YouTube and prefill tasks like reservations or video searches
• Perplexity offers a streamlined alternative to Siri that may beat the native voice assistant in most ways

AI conversational search engine Perplexity is coming for Siri in the form of a new iOS voice assistant. Previously limited to Android, Perplexity's voice assistant wants users to turn to it before the native option. Further, there are a few good reasons why iPhone owners might be inclined to do so.

Basically, it's more proactive and able to go a few extra steps beyond Siri's abilities. Ask it to find a dinner reservation, and it will dive into the OpenTable app to fill in your reservation requests, including guests, date, and time, without you having to say another word, just leaving the final tap on the Book button.

The same goes for hunting for moments in YouTube videos. You can describe the climactic win from a niche sports documentary, see it queued up on YouTube right away.

Of course, some of what Perplexity can do are things that Siri already handles, like writing emails and setting up calendar events. But, even with Apple Intelligence helping out, Perplexity is better at understanding more casual language. And that's before considering the more proactive approach.

Ask Siri about signing up for an event this weekend, and you'll hear the familiar “Here's what I found on the web.” Do the same with Perplexity's voice assistant, and (depending on the circumstances) the AI might say, “I already filled out the form. Just click send.”

Of course, it’s not all-powerful. You need to open the app and tap the microphone icon to start talking to the AI. However, the responses often let you refine your request without having to start over from scratch.

Additionally, the iOS version of Perplexity’s assistant has a few notable limitations. It can’t set alarms or control core iPhone functions, including muting notifications or taking photos. It also can’t access your camera to “see what you see,” which other AI assistants like ChatGPT’s voice mode can.

Introducing Perplexity iOS Voice AssistantVoice Assistant uses web browsing and multi-app actions to book reservations, send emails and calendar invites, play media, and more—all from the Perplexity iOS app.Update your app in the App Store and start asking today. pic.twitter.com/OKdlTaG9COApril 23, 2025

Perplexity popularity

Perplexity is definitely angling to take the place of Siri by not just telling you things, but doing them too. This “agentic AI” approach is gaining popularity across various AI services, such as ChatGPT and Gemini, which are both experimenting with similar ideas.

The aim is to cross the bridge from traditional voice AI to fully independent digital agents. Right now, it won’t book the reservation unless you make your final click. But that might change in a year or two.

Apple isn't ignoring this concept, but has been slow off the mark in some ways. Although Siri's intelligence has been upgraded in recent months, we are still awaiting the full generative AI overhaul that was originally promised to launch this year within a future version of iOS 18. Apple has since delayed the AI-infused Siri and said it will arrive at some point in the future, more specifically, "in the coming year."

Still, by opening its voice assistant to iOS users and layering in real-world tools like OpenTable and YouTube, Perplexity is carving out a space as a nimble alternative to native AI assistants.

And if you just want to say, “Find me tacos and make the reservation,” and have the bot say, “Done," Perplexity's voice assistant might be your new favorite iPhone aide.

You might also like

• Perplexity AI drops new Squid Game-inspired ad that pokes fun at Google starring Lee Jung-jae
• Perplexity's voice mode gets a futuristic makeover on your iPhone
• I tried Perplexity’s Deep Research and it doesn't quite live up to ChatGPT's research potential

Here are the answers for The New York Times Mini Crossword for April 25.

Hand-cranked windows and a DIY dashboard join a high-tech electric vehicle platform in the affordable, customizable Slate Truck.

The bare-bones Slate Truck is designed to be customizable and affordable, with a flat-pack DIY SUV kit that turns the 2-seat pickup into a 5-passenger SUV.

Nampa has several internet service providers to choose from. These are the best plans they offer.

Two new mice for productivity still look like they were made for gamers and that's okay.

• LG Smart Monitor Swing is inspired by StanbyME, but built for everything
• Charge your laptop with 65W USB-C — no extra charger needed
• Touchscreen and webOS make it usable without a computer

LG has announced the Smart Monitor Swing, a 31.5-inch 4K UHD touchscreen monitor designed to rethink how and where users interact with their screen.

The monitor features an IPS panel with a resolution of 3,840 x 2,160 and supports multitouch for direct interaction with apps and files.

The Smart Monitor Swing includes three USB-C ports and two HDMI inputs for connectivity. One of the USB-C ports supports 65W Power Delivery, allowing it to charge most business laptops on the market. On the rear, a concealed adapter helps reduce cable clutter, a feature not commonly found on even the best business monitors.

(Image credit: LG) More than just mobility

A key selling point is its all-in-one wheeled stand, adapted from the portable StanbyME smart TV design. The stand offers more than just mobility: it supports height adjustment, up-and-down tilt, left-right swivel, and a pivot for switching between landscape and portrait modes.

LG’s proprietary webOS smart platform is also built in, giving users access to apps like Microsoft Office, YouTube, and others directly from the monitor. Like LG’s MyView Smart Monitor, it can run cloud PC services without needing a separate computer.

While the Smart Monitor Swing borrows some standout features from the StanbyME, it omits the built-in battery and remote control in favor of a cleaner, productivity-focused design.

Combining smart functionality with a portable frame, the LG Smart Monitor Swing is positioned to compete with the best portable monitor options available.

While LG has not yet confirmed global pricing or availability, the monitor will launch in Korea on April 24 at 1,049,000 won (approximately $735).

You may also like

• We've rounded up the best monitors for photo editing on the market today
• And here are the best monitors for programming we've found
• Western Digital plans to launch 40TB HDD next year using HAMR technology