TechRadar News

In November last year, I wrote that the Google Pixel 9 Pro was the best-looking phone of 2024 and the only Android device that could tempt me away from my beloved iPhone. Ultimately, I still sided with the iPhone 16 Pro (the call of convenience was too hard to resist), but Google seems determined to win me over in 2025 with the Pixel 10 Pro.

Google’s next flagship is set to debut at this year’s Made by Google showcase on August 20, but it’s already been semi-unveiled in an official teaser video. The footage in question shows the back and one side of what’s purported to be the Pixel 10 Pro in a brand new ‘Moonstone’ color, and it’s heightened my excitement for the phone’s full-blown reveal.

Now, I’m not usually one to lose my head over colors, but this silverish-blue shade seemingly complements Google’s new (or rather, unchanged) flagship Pixel design perfectly. In fact, I’d go as far as to say that the Google Pixel 10 Pro in Moonstone is already the best-looking phone of 2025 – and yes, I’ve seen the rumored iPhone 17 Pro design.

To be clear: the Pixel 10 Pro looks almost identical to the Pixel 9 Pro. Leaked renders suggest that the phone’s metal frame will more closely resemble its rear panel glass (rather than being polished, as on the Pixel 9 Pro). But beyond that, it’s safe to expect a like-for-like handset; I don’t think this is a bad thing.

An official image of a phone believed to be the Pixel 10 Pro in Moonstone (Image credit: Google)

Google hit a home run with the design of the Pixel 9 Pro, which, in my opinion, is the Android phone that’s come closest to replicating the iPhone’s unique synergy of simplicity and style. It remixed the Pixel formula with flat aluminum sides, a frosted glass rear panel, and a perfectly sized pill-shaped module, all of which made the Pixel line feel more grown-up.

Google is right to stick by this cleaner, more industrial aesthetic with the Pixel 10 Pro, and the addition of a new ‘Moonstone’ color (as opposed to something more outlandish) suggests Google wants its next flagship phone to be perceived as just that: a grown-up flagship.

Of course, design is subjective. I’m sure many people continue to prefer the uniquely Pixel aesthetic of the Pixel 8 Pro, Pixel 7 Pro, and so on. I can understand why.

But I genuinely believe that Google is one of the few – perhaps the only – smartphone manufacturers to have successfully designed a better-looking phone with each new generation. And if the below X post is anything to go by, Google thinks so too:

10 generations in the making. Look for #Pixel10 in 10 days at #MadeByGoogle: https://t.co/VxzWRaYodH pic.twitter.com/ER8qtFcDv5August 10, 2025

Will I finally switch sides and buy the Pixel 10 Pro is year? Probably not. But that’s only because a decade of investing in Apple hardware has made ditching the iPhone more trouble than it's worth. The Pixel 9 Pro was the best-looking phone of last year, and, for me, the Pixel 10 Pro has already repeated the feat in 2025.

For more on Google's next flagship, check out our roundup of the five upgrades to expect from the Google Pixel 10 Pro.

You might also like

• 5 ways the Google Pixel 10 Pro can get ahead of the iPhone 17 Pro and Galaxy S26 Ultra
• The Google Pixel 10 Pro XL just got shown off in two subdued shades
• The Google Pixel 10 might let you edit photos with your voice (yes, really)

• President Trump set to meet with Intel CEO Lip-Bu Tan
• The President recently called for the CEO's immediate resignation
• Trump accused Tan of major "conflicts of interest" because of his investments in Chinese companies

Intel CEO Lip-Bu Tan is set to meet with President Trump to discuss the former's professional and personal history, and explore ways in which the US government and Intel can work together, sources have told the Wall Street Journal.

Tan has recently faced a series of attacks from President Trump, who has demanded his resignation over alleged investments and dealing with Chinese businesses.

Trump has accused Tan of being “highly conflicted” and demanded his “immediate resignation,” stating that there is “no other solution to this problem.”

Rocky history

Intel has been steadily trimming its global workforce in recent months, leading to a 20% reduction in headcount that has seen thousands of employees lose their jobs in US factories

Tan has also recently butted heads with Intel board members over the direction of Intel’s investments in the US.

For example, the new CEO has reduced the pace of construction of a new factory in Ohio in order to meet perceived customer demands, causing issues with Republican Ohio senator Bernie Moreno.

Both the job cuts and slowed construction directly contradict President Trump’s desire to increase domestic investment from companies that have typically manufactured goods outside of the US.

What’s more, Tan’s previous position as CEO of Cadence Design Systems appears to have ruffled a few feathers in the Trump administration due to the company recently handing over $140 million in penalties after pleading guilty to unlawfully exporting chip design tools to restricted Chinese military organizations.

Tan himself also holds numerous investments in Chinese businesses, including buying $200 million worth of stakes in Chinese manufacturing and chip companies between 2012 and 2024 according to a Reuters report.

However, Tan has not only been on the board of numerous enterprises involved in chip manufacturing, but founded a venture capital firm named Walden International that focused heavily on tech startups in the Asian market.

Tan’s holdings in Chinese companies could therefore simply be a smart investment based on decades of experience, rather than the shady under-the-table dealings Trump has alleged in his Red Scare tirades on Truth Social.

Trump appears to be deaf to his own accusations of Tan’s “conflicts of interest,” with the President frequently using his position to market his own goods from the Trump Store, as well as launching the $TRUMP cryptocurrency of which the President controls 80% of the coin's supply.

(Image credit: Photo by Chip Somodevilla/Getty Images)

In a recent letter to employees, Tan said the US had been his home for more than 40 years, and the company was communicating with the White House to make sure the President has “the facts.”

Trump has recently levied significant tariffs on countries that he perceives to have an unfair trade balance, causing the price of some goods to increase as the cost of imports are passed on to the consumer - a fact that many within the Trump administration and his MAGA base are struggling to come to terms with.

Trump’s pursuit of an America First policy has drawn significant investment from many major companies, including Nvidia, Intel, and Apple, to build factories in the US to produce goods domestically. However there is a reason that these companies have historically manufactured their goods outside of the US. It’s cheaper.

Recent studies into the costs of a US manufactured iPhone for example have theorized that the price tag could rise by between 25-100%.

In an added contradiction, Trump’s own branded handsets from Trump Mobile likely won’t be manufactured in the US, with Eric Trump claiming that “Eventually, all the phones can be built in the United States of America.”

You might also like

• Intel loses another top exec as Nokia poaches AI and data center chief as new CEO
• These are the best AI tools
• Take a look at our guide to the best password manager

• Columbia University files form with the Maine Attorney General's Office
• It confirmed the number of victims, and the type of data stolen
• Victims are being offered free credit and identity theft monitoring services

A massive cyberincident affecting almost 870,000 Columbia University students, employees, and other individuals occured in May 2025, the university has confirmed.

In a new filing with the Office of the Maine Attorney General, the university said in late June 2025 it experienced an outage in its IT systems, prompting an investigation with the support of third-party cybersecurity and forensics experts.

The investigation confirmed the outage was the result of a cyberattack, in which yet unidentified perpetrators stole sensitive data on exactly 868,969 people, including employees, applicants, students (both current and former), and various family members.

460 GB of data taken

"Our investigation determined that, on or about May 16, 2025, an unauthorized third-party gained access to Columbia's network and subsequently took certain files from our system," Columbia University said.

"To date, we have no evidence that any Columbia University Irving Medical Center patient records were affected."

The university recently started notifying affected individuals via letters which detailed the type of information stolen in the breach:

"The affected data included your name, date of birth, and Social Security number, as well as any personal information that you provided in connection with your application to Columbia, or that we collected during your studies if you enrolled," the university apparently said.

"This included your contact details, demographic information, academic history, financial aid-related information, and any insurance-related information and health information that you shared with us."

Some attackers confirmed the breach, and claimed to have stolen 460 GB of data. So far, there is no evidence that the data was abused in the wild, but Columbia University will provide victims with two years of free credit monitoring, fraud consultation, and identity theft protection services, through Kroll, regardless.

How to stay safe

While victims directly cannot do much about the stolen data, they can make sure the attackers don’t use it against them. The best course of action is to remain vigilant with incoming communications, especially those claiming to come from Columbia University.

Unsolicited emails, instant messages, or phone calls, particularly those “threatening” to terminate accounts or otherwise prevent services, are most likely bogus.

If you suspect you’re being targeted, the best course of action is to stop all communication, and then reach out to Columbia University directly, through proven channels.

Via BleepingComputer

You might also like

• "No evidence" - here's why the massive 16 billion record data breach may not be as bad as first thought
• Take a look at our guide to the best authenticator app
• We've rounded up the best password managers

Alien: Earth is finally close to hatching on TV screens across the globe. The highly-anticipated project and first-ever show set in the Alien franchise will burst forth later this week, so you'll want to know when and where you can stream it.

Below, I've provided more details on its various release dates and launch times. I'll also fill you in on which streaming platforms it'll be available on and provide you with a full release schedule, so you can block out time in your busy schedule to watch new episodes when they drop.

What is the release date and launch time for Alien: Earth episodes 1 and 2?

Alien fans heading over to their streamer of choice to watch the franchise's maiden TV show like... (Image credit: FX Network/Hulu/Disney+)

That depends on where you live. Those of you in North and South America can tune in to watch its two-episode premiere on Tuesday, August 12. Meanwhile, viewers in the UK, Europe, Africa, and Asia can catch them on Wednesday, August 13.

As for those all-important launch times, here's when Alien: Earth will debut on Hulu (US only) and Disney+ (everywhere else). NB: if your nation isn't listed below, use the 5pm PT option to work out when you should load up Disney+, aka one of the world's best streaming services.

• US – Tuesday, August 12 at 5pm PT / 8pm PT
• Canada – Tuesday, August 12 at 5pm PT / 8pm PT
• UK – Wednesday, August 13 at 1am BST
• India – Wednesday, August 13 at 5:30am IST
• Singapore – Wednesday, August 13 at 8am SGT
• Australia – Wednesday, August 13 at 10am AEST
• New Zealand – Wednesday, August 13 at 12pm NZST

When will new episodes of Alien: Earth air on Hulu and Disney+?

Yep, the Alien property's famous Xenomorphs are part of proceedings (Image credit: FX Networks/Hulu)

You can catch new entries every Tuesday in North and South America, and Wednesdays everywhere else. For more information, check out the list below:

• Alien: Earth episode 1 – Tuesday, August 12 / Wednesday, August 13
• Alien: Earth episode 2 – Tuesday, August 12 / Wednesday, August 13
• Alien: Earth episode 3 – Tuesday, August 19 / Wednesday, August 20
• Alien: Earth episode 4 – Tuesday, August 26 / Wednesday, August 27
• Alien: Earth episode 5 – Tuesday, September 2 / Wednesday, September 3
• Alien: Earth episode 6 – Tuesday, September 9 / Wednesday, September 10
• Alien: Earth episode 7 – Tuesday, September 16 / Wednesday, September 17
• Alien: Earth episode 8 – Tuesday, September 23 / Wednesday, September 24

Before its first two chapters arrive, read my Alien: Earth review and our dedicated Alien: Earth guide. Then, check out the section below for more exclusive coverage on the forthcoming sci-fi horror TV Original from FX Networks.

You might also like

• Alien: Earth isn't ignoring what happened earlier on the Alien timeline – but only because 'there isn't much mythology' to begin with, producer argues
• Alien: Earth introduces new creatures as terrifying as the Xenomorph, but the show's creator didn't want to squeeze in 'a vending machine of alien life' for the sake of it
• Alien: Earth star Timothy Olyphant only needed to do one thing to get into character for the sci-fi horror show: 'It does a lot of the work for me'

• Four Polk speaker models, all with improved audio and weatherproofing
• Easier to install, even in hard-to-reach spaces
• From $225 per pair

Polk Audio has updated its Atrium series of outdoor speakers, which start at a very affordable $225 (so around £169 or AU$345, give or take).

Polk has been making speakers for the great outdoors for ages – the Atrium first launched way back in 1991 – and the latest models come with improved weather resistance, more versatile mounting options and "enhanced sonic performance", which from Polk should not be taken lightly.

(Image credit: Polk Audio)Polk Audio Atrium speakers 2025: models and pricing

There are four models in the range: the Atrium 4, the Atrium 5, the Atrium 6 and the Atrium 8SDI. Prices range from $225 per pair for the Atrium 5 to $550 for the Atrium 8SDI.

The mid-range drivers are now injection-molded mineral filled or aerated polypropylene, the same materials used in Polk's LSi Series, and the tweeters are anodized aluminum. The Atrium 6, 7 and 8SDI also get water-resistant PowerPort bass venting, which ups the low end by around 3dB.

Both tweeters and woofers have rubber surrounds, and the speakers have a new broad coverage baffle design that Polk says offers better sound dispersion than previous models. The grilles and brackets are aluminum to eliminate rusts issues and make the speakers last longer outdoors.

Another key update is the new Speed-Lock mounting system, which Polk says makes installation much easier – so much so that you can install the speakers one-handed "even in hard-to-reach locations".

As before the Atrium speakers come with a wide range of standards compliance including ASTM D5894-UV Salt Fog, Mil Standard 810 Immersion, and Mil-Std 883 Method 1009.8 for salt and corrosion.

The new Atrium speakers are available now in black or white, and Polk says they're also paintable so you can customize them to fit any outdoor décor.

You might also like

• See our pick of the best stereo speakers
• And here's our roundup of the best Bluetooth speakers
• Want wi-fi streaming? Our guide to the best wireless speakers is for you

• President Trump allows chip sales to China to resume - with one major caveat
• AMD and Nvidia will have to hand over 15% of the sales revenue
• No additional measures have been made to address national security threats

The US government has reversed a ban on the sale of Nvidia's H20 chip and AMD’s MI308 chip to China under the condition that 15% of the revenue is paid to the US government.

The Trump administration expanded efforts to curb China’s ability to obtain powerful chips used to develop AI models by expanding Biden-era sanctions to an outright ban on the sale of the processors.

Many security experts, including those working under the Trump administration, have warned the government the sale of powerful processors to China will help it to expand both its consumer and military AI efforts to surpass the US.

What happened to national security?

The Nvidia H20 and AMD MI308 processors were developed to comply with Biden administration restrictions on AI processing chips that could be sold in China. Restrictions were imposed on the sale of powerful AI chips to China due to national security concerns surrounding the development of AI models for the People’s Liberation Army.

Trump banned the sale of these chips to China following a trade spat in April 2025, and then reversed the ban in July. Now, seeing the apparent worth of the sales and hoping to cash in, the Trump administration is issuing a 15% revenue cut for itself.

Speaking to the BBC, Nvidia said, “We follow rules the US government sets for our participation in worldwide markets. While we haven't shipped H20 to China for months, we hope export control rules will let America compete in China and worldwide.”

Whether this 15% revenue cut will be passed on to the Chinese market by Nvidia and AMD remains to be seen. But the elephant in the room still remains. What happened to the national security problem?

(Image credit: Getty Images)

Deborah Elms, head of trade policy at the Hinrich Foundation, said “You either have a national security problem or you don't. If you have a 15% payment, it doesn't somehow eliminate the national security issue."

Despite sanctions and trade bans, China has still managed to import powerful AI processing chips through loopholes and third-parties, and is likely using the chips to train AI systems to be used for military purposes.

20 security experts wrote a letter to US Commerce Secretary Howard Lutnick advising against the sales of the chips to the Chinese market, adding that even though the majority of the buyers would be civilian companies, the chips would still be used by China’s military.

“Chips optimized for AI inference will not simply power consumer products or factory logistics; they will enable autonomous weapons systems, intelligence surveillance platforms and rapid advances in battlefield decision-making,” the letter said.

Charlie Dai, vice president and principal analyst at global research firm Forrester called the deal “unprecedented,” further stating that, “The arrangement underscores the high cost of market access amid escalating tech trade tensions, creating substantial financial pressure and strategic uncertainty for tech vendors.”

You might also like

• Take a look at the best AI tools and best AI writers
• These are our picks for the best antivirus software
• OpenAI has new, smaller open models to take on DeepSeek - and they'll be available on AWS for the first time

• Buds 2a are predicted to launch on 20th August 2025
• Leaks suggest ANC and Spatial Audio, but no head tracking
• Reports also suggest a 50% price hike in Europe

It's been a few days since the last Google Pixel Buds 2a leak, so we're due another big one – and well-known leaker Evan Blass is happy to oblige. Posting on X, Blass has published a leaked spec sheet that fills in some of the blanks from previous leaks.

The spec sheet also does something useful: it shows how the more affordable buds compare to the current Google Pixel Buds Pro 2.

(Image credit: Evan Blass / X)Google Pixel Buds 2a: key features and pricing

Like the Buds Pro 2, the Google Pixel Buds 2a should have active noise cancellation, but they don't appear to have the Pros' Silent Seal, which adapts the ANC further based on the shape of your ears.

There are three other key differences here: the Buds 2a get spatial audio, but they aren't getting head tracking; the Buds 2a don't have wireless charging; and they don't have conversation detection. The only other spec on the sheet is battery life, which is a promised 7 hours – and 20 with the charging case. By comparison the Buds Pro 2 get 8 hours and 30 hours respectively.

This isn't the only Google Pixel Buds 2a leak. According to the German tech site WinFuture, which has also posted new renders of the earbuds, there's going to be a price hike in Europe: where the Pixel Buds A were €99, the Buds 2a will reportedly be €149. That's a huge price increase and if it's reflected in other markets would mean $149 / £149 / AU$195).

On a related note: if you're already a Google Pixel Buds Pro user and don't want to buy new buds any time soon, there's a little bit of good news for you: Google is rolling out a nifty visual upgrade to Android's headphone and earbud status notifications to make it easier to check battery levels. Thanks to Android Police for that one.

Of course, the Pixel Buds 2a leaks above are unofficial and should be treated as such – speculation and possible feature sets, rather than stone cold reported fact. But when we know more about the unreleased Pixel Buds 2a, so shall you.

You might also like

• Google Pixel Buds Pro 2 review: tiny earphones that are drowned out by the competition
• Samsung Galaxy Buds 3 Pro vs Google Pixel Buds Pro
• The best wireless earbuds, all real-world tested by us

• M&S has finally restored its in-store click and collect service
• The cyber incident was disclosed in April 2025
• Online orders for delivery were restored in June

Marks and Spencer (M&S) has finally restarted click and collect orders for clothing, home and beauty products after a nearly four-month suspension following an apparent major cyberattack.

Although the company had resumed online orders for delivery on June 10 after it disclosed details of a cyber incident on April 22 (and stopped deliveries and collections from April 25), M&S took a further 15 weeks to switch back on its click and collect services.

The cyber incident was previously expected to cost the company around £300 million in lost operating profit for this fiscal year, but M&S hopes to halve the impact via insurance and cost controls.

M&S click and collect back online

However, although the restoration of click and collect signals a 'back to normal' for customers, analysts aren't expecting a sudden resurgence as M&S continues to battle with reputational damages.

Although the British retail giant took a major hit, the industry didn't, and rivals like Sainsburys and Next were able to scoop up some of the lost business.

M&S CEO Stuart Machin had previously stated the effects of the incident could continue into June and July, signalling a restoration in August, and the company has been able to adhere to that.

Detailing its learnings in Parliament, M&S has urged stronger cyber-incident disclosure norms. The counsel also noted that companies should be able to operate manually during outages.

The UK's National Crime Agency arrested four people in a probe tied to attacks on M&S, Co-op and Harrods, however the true cause of the incident still remains uncertain.

The attacks against Marks and Spencer (M&S) and the Co-op supermarket were combined into a single incident by the same attacker by the Cyber Monitoring Centre (CMC), an independent, non-profit body established to categorize major cyber events by the insurance industry.

It had been reported the group known as Scattered Spider was behind the ordeal, but TCS, which has been servicing M&S for more than a decade, is also investigating whether it was the stepping stone to the attack.

You might also like

• Take your store online with the best ecommerce platforms and best payment gateways
• Mystery of M&S hack deepens as TCS claims none of its systems were compromised
• We've listed the best ransomware protection tools around

There's a calculation that every AI executive should know by heart, but most have never done: an on-premises GPU server costs roughly the same as six to nine months of renting equivalent cloud capacity.

Given that hardware typically runs for three to five years, the mathematics are stark, yet somehow this isn't common knowledge in boardrooms making million-pound infrastructure decisions.

The issue stems from a fundamental mismatch between how we think about AI costs and how they actually accumulate. The operational expenditure over capital expenditure model feels intuitive when you pay as you go, scale as needed, and avoid big upfront commitments.

But AI workloads break these assumptions in ways that make traditional cloud economics misleading.

What the cloud isn't telling you

For example, renting a single NVIDIA H100 GPU instance from a hyperscaler cloud provider can cost around $8/hour, or over $5500 per month. Over 12 months, that's upwards of $65,000.

By contrast, purchasing equivalent hardware outright might cost around $30,000 to $35,000, with three to five years of usable life. Add power, cooling, and maintenance and you still come out ahead after just 6 to 9 months of usage. Plus, you own the hardware so you don’t have to return it after 12 months.

But the pricing hierarchy is more complex than it appears. While neocloud providers like Fluidstack offer H100s at that $2/hour rate, hyperscalers charge closer to $8/hour, making the on-premises case even stronger.

The real-world comparison gets harder to ignore when you consider actual deployments: 8xH100 systems from Dell or Supermicro cost around $250,000, versus $825,000 for three years of equivalent hyperscaler capacity (even with reserved pricing). NVIDIA's own DGX systems carry a punishing 50-100% markup over these already substantial prices.

The missing numbers in most AI budgeting conversations represent real savings, not theoretical ones. The problem compounds when you examine specific use cases.

Consider training runs. Most cloud providers only guarantee access to large GPU clusters if you reserve capacity for a year or more. If your training only needs two weeks, you're still paying for the other 50.

Meanwhile, inference demands create their own mathematical puzzle. Token-based pricing for large language models means costs fluctuate with the unpredictability of the models themselves, making budget forecasting feel more like weather prediction than financial planning.

Elasticity, but with fine print

The cloud’s promise of elastic scale feels tailor-made for AI – until you realize that scale is constrained by quota limits, GPU availability, and cost unpredictability. What’s elastic in theory often requires pre-booking in practice and cash upfront to make costs acceptable.

And once your usage grows, discounts come with multi-year commitments that mirror the CapEx models cloud was meant to replace.

It's not that the cloud isn't scalable. It's that the version of scale AI teams need (cost-efficient, high-throughput, burstable compute) isn’t always what’s on offer.

The irony runs deeper than pricing. Cloud providers market flexibility as their core value proposition, yet AI workloads, which are the most computationally demanding applications of our time, often require the least flexible arrangements.

Long-term reservations, capacity planning, and predictable baseline loads start to look suspiciously like the traditional IT procurement cycles cloud computing was supposed to eliminate. The revolution becomes circular.

Hidden costs, visible friction

The hidden complexity emerges in the details. Teams preparing for usage spikes often reserve more capacity than they use, paying for idle compute "just in case."

Data migration between providers can consume non-trivial amounts of engineering time, representing an opportunity cost that rarely appears on infrastructure budgets but significantly impacts small, time-constrained teams.

These opportunity costs compound over time. When teams switch between cloud providers – driven by pricing changes, performance issues or compliance needs, they often face weeks of rewrites, re-optimizations, and revalidations.

It’s not just the IT infrastructure that changes, but all the code that manages it, internal expertise in that provider disappears and deployment pipeline needs to be rewritten. For lean teams, this can mean delayed product updates or missed go-to-market windows, which rarely get factored into the headline GPU bill.

Perhaps most surprisingly, the operational burden of managing on-premises infrastructure has been systematically overstated. Unless you're operating at extreme scale, the complexity is entirely manageable through in-house expertise or through managed service providers.

The difference is that this complexity is visible and planned for, rather than hidden in monthly bills that fluctuate unpredictably.

From budgeting to strategy

Smart companies are increasingly adopting hybrid approaches that play to each infrastructure model's strengths. They use owned hardware for predictable baseline loads like the steady-state inference that forms the backbone of their service.

Cloud resources handle the spikes: time-of-day variations, customer campaign surges, or experimental workloads where spot pricing can soften the blow.

Companies taking this approach have moved beyond anti-cloud thinking toward financially literate engineering.

The cloud remains invaluable for rapid experimentation, geographic scaling, and genuinely unpredictable workloads. But treating it as the default choice for all AI infrastructure ignores the mathematical reality of how these systems actually get used.

Companies getting this calculation right are doing more than saving money. They're building more sustainable, predictable foundations for long-term innovation.

These conversations aren’t just technical, they’re strategic. CFOs may favor cloud for its clean OpEx line, while engineers feel the pain of FinOps teams desperately chasing them to delete resources as month-end cost spikes and poor support hit.

That disconnect can lead to infrastructure decisions driven more by accounting conventions than real performance or user experience. Organizations getting this right are the ones where finance and engineering sit at the same table, reviewing not just cost, but throughput, reliability, and long-term flexibility. In AI, aligning financial and technical truths is the real unlock.

Understanding these hidden mathematics won’t just help you budget better, it’ll make sure you’re building infrastructure that works the way AI actually does, freeing up headspace to focus on what matters most: building better, faster, and more resilient AI products.

We list the best IT management tool.

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

• Hostinger Horizons integrates ecommerce directly, removing reliance on plugins or third-party connections.
• The update uses a tested ecommerce engine from Hostinger’s existing website builder system.
• Users can add up to 600 products with no extra transaction fees charged.

Hostinger has introduced built-in ecommerce platform functionality to its Horizons vibe coding platform, making it the first such tool in its category to offer a complete online store setup without relying on third-party integrations.

The website builder company claims this update removes the need for plugins, lengthy configuration, or technical expertise, which allows users to establish an online retail presence in minutes.

Users can list up to 600 products, configure over 100 payment gateway options, set up shipping methods, and apply taxes or discounts as needed.

Reducing time and effort for digital storefronts

“We’re building Horizons for people who don’t want to worry about technical setup or to have to figure out how ecommerce works," said Dainius Kavoliūnas, Head of Hostinger Horizons.

"Creating an online store was technically possible before, but it took too much time and effort - fortunately, a tested solution was right next door – our Hostinger Website Builder team already had a powerful ecommerce engine. We just needed to integrate it."

Vibe coding is a relatively new approach to web development that replaces manual coding with conversational AI prompts - all users need to do is describe their desired website or application in natural language, and the platform generates a ready-to-publish version.

Additionally, the Horizons update integrates an ecommerce platform directly into the interface, enabling seamless store management without needing to leave the system.

Hostinger states there are no additional transaction fees, and inventory management can be done manually without consuming paid AI prompts.

While this eliminates recurring costs for simple updates like price changes or stock adjustments, scaling beyond the provided capacity or customizing complex workflows may still require additional resources.

The inclusion of Hostinger’s existing e-commerce engine, previously part of its standalone website builder, suggests the company is repurposing proven infrastructure rather than introducing an untested solution.

This could offer some reliability, but whether it meets the expectations of experienced merchants remains to be seen.

Although AI can be used for storefront customization, such as rearranging products or altering visual elements, the long-term success of any online shop will still rely heavily on marketing, product quality, and customer service

These factors are not automatically solved by a fast setup process.

Hostinger launched Horizons in March 2025 to enable non-technical users to build and publish websites or applications through simple text prompts.

Earlier updates included generative engine optimization, manual editing tools, free automatic error correction, and database integration.

“After analyzing 200,000 prompts, we learned that business websites are the most popular use case among Hostinger Horizons clients, representing around a third of all projects built with the AI tool.”

“Understanding that our clients want to sell online, we delivered an easy, intuitive ecommerce feature,” Kavoliūnas added.

You might also like

• We've rounded up the best password managers
• Take a look at our guide to the best authenticator app
• Massive leak of over 115 million US payment cards caused by Chinese "smishing" hackers - find out if you're affected

Fifty years ago, it was heists like the one that hit the Baker Street Bank that had the power to shock the nation. Now, in the digital world, heists look starkly different and cybersecurity threats are constant, with banks like NatWest facing a “continuous arms race” with around 100 million cyber-attacks every month. What used to be gangs of robbers digging tunnels and smuggling deposit boxes full of cash are now groups of hackers sending phishing emails and holding some of the most notable companies to ransom for hundreds of millions of dollars.

This transition from physical to digital theft is evident. No longer confined to vaults and getaway cars, today's high-stake heists are executed remotely, by online threat actors. These modern-day criminals operate across borders, targeting vulnerabilities in systems and human behavior to extract data and money. The sheer volume and relentless nature of these digital assaults, as exemplified by financial institutions battling millions of cyber-attacks monthly, highlight a new era of crime.

The growing problem of cyber-attacks

Cyber-attacks are a growing problem, amongst a growing number of sectors, and confronting this escalating issue is vital. It’s not just banks that are facing the constant threat of cyber-attacks; cyber threats are growing at an exponential rate, while becoming increasingly sophisticated and targeted.

Data breaches have hit a myriad of industries: from luxury brands like Dior and supermarkets like M&S, to cryptocurrency exchange Coinbase and UK government organization Legal Aid.

The dangers to personal data are being felt across all sectors, at all digital touchpoints. Amid this battleground of immediate cyber threats comes a growing demand for robust security solutions that address company concerns.

From advanced antivirus technologies to endpoint backup software, AI-powered security is evolving rapidly to stay ahead of such attacks - and it’s essential that companies invest in these defenses in order to stay more than one step ahead.

Evolution of technology

As technology evolves at a rapid pace, companies must keep up with advancements made by cyber-attackers. As businesses of all sizes continue to embrace digital transformation, the need to strengthen their cybersecurity grows increasingly critical.

The UK Government’s recently published Cyber Governance Code of Practice highlights that management of cyber risks is vital for modern businesses to function, and effective management requires collective input from across an organization. This Code of Practice and governance framework package guides boards and directors in managing digital risks and safeguarding their businesses and organizations from cyberattacks.

The framework encourages companies to take four employee-focused actions: foster a cybersecurity culture; ensure clear policies support a positive cybersecurity culture; improve their own cyber literacy through training; and use suitable metrics to verify the organization has an effective cybersecurity training, education, and awareness program.

The report is a clear reminder that the human firewall, that is, the employees who encounter an attack and respond, is just as important as technological defenses.

More than a simple fix, a culture shift is needed

It’s not enough to roll out generic training. The reality is that in today’s world, one wrong click can bring a business to a complete halt. According to the latest insights, the approximate amount of ransoms paid globally in 2024 reached $813.55 million.

When requested to pay a ransom, companies know that refusing to do so runs the risk of their customers’ personal information being leaked publicly, which would additionally require them to pay the associated financial penalties and legal payouts, not to mention reputational damage.

Addressing the threat of cyber-attacks must be embedded in a company’s culture, given the fact that if threat actors are successful, the impact of their actions would be felt not only company-wide but also by the ecosystem within which the organization operates.

Leadership and security

Organizations can bolster their security by cultivating strong leadership, providing tailored training, and building a proactive security culture to create a ‘human firewall’ of colleagues armed with know-how.

Employees of all skillsets and seniorities should undergo comprehensive and ongoing cyber awareness training, whatever their role and seniority, to drive the defenses forward and cultivate a mindful culture.

When employees are provided with the knowledge and tools to maintain awareness of the dangers their company is facing, they can be the most effective method to keep the business secure.

Building a mindful culture

Building a mindful culture can be complemented by a Zero Trust approach, which creates a robust defense against evolving cyber threats. This strategic approach mandates rigorous verification for all access requests, irrespective of their origin or the user's location within the network, thereby yielding exceptionally strong results that effectively eliminate a significant portion of potential threats.

For example, when an employee receives an email requesting sensitive information or a link to a suspicious website, they should be trained to recognize it as a potential phishing attempt right away, verify the sender's identity, and report the email to the IT department for further investigation.

This proactive stance, ingrained through a Zero Trust philosophy and continuous education, significantly reduces the likelihood of successful breaches. It’s better safe than sorry, and in the realm of cybersecurity, this means being diligent about taking the extra steps to fortify an organization's digital defenses.

Don't stop at basic protections

Don't stop at basic protections, make ongoing training a priority.
Defenses can’t stop at antivirus technology and endpoint protection, and training isn’t a one-time solution. While these are the necessities, they are simply not enough for the twenty-first century heist as businesses continue to battle millions of cyber-attacks each month.

As threats advance or teams become complacent, ongoing phishing simulations, tests and education are key in maintaining a robust human firewall. Companies must invest in technology and ongoing training to equip employees across all roles and levels with the skills and awareness to stay alert. A company’s greatest weapon can be its workforce, if leveraged.

Cybersecurity needs tech, but it's nothing without people who are well trained to understand the latest attack methods and protect against the digital transition's inherent risks.

We list the best ransomware protection.

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

An old proverb famously states, "If you want to go fast, go alone. If you want to go far, go together."

This is especially true when it comes to artificial intelligence, where breakneck advances happen seemingly every day. And while individual companies are rapidly fielding their own AI-powered chatbots and analysis tools, real long-term improvement and innovation in this new scientific frontier often requires broad collaboration in developing open and trusted AI systems that produce accurate, reliable, and safe outputs.

Early conventional wisdom held that only so-called 'closed' AI systems controlled by one company could be safe and trusted. Some argued that open models would inevitably undermine safety or lead to misuse. But experience is quickly showing that open source models and the collaboration they bring are a powerful tool for promoting security and trust.

The power of collaboration

Collaboration is a powerful force for AI advancement because it fosters diverse perspectives and capabilities. When it comes to AI, collaboration can, in many cases, be optimized by leveraging open source to reduce bias, increase transparency, gain greater control over our data, and ultimately, accelerate time to innovation.

According to McKinsey, organizations that view AI as essential to their competitive advantage are far more likely to use open source AI models and tools than organizations that do not. Open source AI models, tools, and frameworks enable developers and researchers to build upon existing work, rather than starting from scratch, to achieve higher-quality outputs more quickly.

The open source software approach thrives on community contributions, bringing together individuals, companies, and organizations from around the globe to collaborate on shared goals. This is where organizations like the AI Alliance—which was spearheaded by IBM and others, and is comprised of technology creators, developers, and adopters collaborating to advance safe and responsible AI—play a crucial role.

By pooling resources and knowledge, the AI Alliance provides a platform for sharing and developing AI innovations. This meritocracy yields immediate value, both for the broader technology ecosystem and the world at large.

Why the AI Alliance matters today

There are many practical and ethical reasons for such broad-based AI partnerships. AI research and development require substantial resources, including data, computing power, and expertise. The availability of open source models keeps costs down, broadening choices and helping to prevent the concentration of the AI industry in the hands of a few major players.

The AI Alliance also offers a forum to hold honest conversations among like-minded organizations about AI-related legislation and its impacts on greater innovation and adoption.

In a short time, the AI Alliance has blossomed into a vibrant ecosystem, bringing together a critical mass of data, tools, and talent. Today, more than 140 organizational members from 23 countries collaborate through the alliance to address some of the most pressing challenges in AI.

Open source is particularly critical to members of the alliance, including Databricks, which has long championed the democratization of AI. We’ve open sourced many critical big data processing and analytics projects, like the Delta Lake, MLflow, and Unity Catalog tools that underpin many large data and AI deployments today.

When it comes to today’s AI ecosystem, we need to ensure that everyone, including academics, researchers, non-profits, and beyond, can access and understand the best AI tools and models. The more we all understand these models and how to utilize them, the more we can share ideas on how to safely shape the future of AI and subsequently use it to solve today’s toughest challenges.

But we can’t do it alone.

Collaborate, code, and create the future of AI

We established a policy working group within the Alliance to focus not only on advocacy but also on developing responses to government requests that could impact open-source AI development. For example, last year, we contributed to the landmark National Telecommunications and Information Administration study examining potential benefits and risks of open weight frontier AI models.

The final NTIA report strongly underscored the valuable role of open models in today’s AI ecosystem, while also highlighting the need for vigilant monitoring and ongoing evaluation of policies to manage emerging risks in the future.

Our intention is to ensure that AI regulation is thoughtfully crafted so that open source AI thrives. Organizations like the AI Alliance have laid a solid foundation for international cooperation, but it's just the beginning.

If you work at a business that prioritizes artificial intelligence, you too can be part of this important work. Start by developing educational programs, workshops, and training sessions – and joining AI-related projects and communities – to share knowledge and build tools that benefit others.

You can create and share your own open source projects, such as datasets, pre-trained models, or utilities, which build on a foundation of AI fairness, transparency, and accessibility to ensure the benefits of AI are widely distributed. Check out GitHub or Hugging Face to look for AI/ML projects that align with your skills and interests.

The advent of AI is a pivotal moment in our collective human history. Experience shows that collaboration will be key to our success in advancing AI innovation with safety and trust. We must move into this promising future with open arms and open software models and tools, adequately prepared for the challenges ahead. Let's go far—together.

We list the best IT Automation 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

• Experts warn a single calendar entry can silently hijack your smart home without your knowledge
• Researchers proved AI can be hacked to control smart homes using only words
• Saying “thanks” triggered Gemini to switch on the lights and boil water automatically

The promise of AI-integrated homes has long included convenience, automation, and efficiency, however, a new study from researchers at Tel Aviv University has exposed a more unsettling reality.

In what may be the first known real-world example of a successful AI prompt-injection attack, the team manipulated a Gemini-powered smart home using nothing more than a compromised Google Calendar entry.

The attack exploited Gemini’s integration with the entire Google ecosystem, particularly its ability to access calendar events, interpret natural language prompts, and control connected smart devices.

From scheduling to sabotage: exploiting everyday AI access

Gemini, though limited in autonomy, has enough “agentic capabilities” to execute commands on smart home systems.

That connectivity became a liability when the researchers inserted malicious instructions into a calendar appointment, masked as a regular event.

When the user later asked Gemini to summarize their schedule, it inadvertently triggered the hidden instructions.

The embedded command included instructions for Gemini to act as a Google Home agent, lying dormant until a common phrase like “thanks” or “sure” was typed by the user.

At that point, Gemini activated smart devices such as lights, shutters, and even a boiler, none of which the user had authorized at that moment.

These delayed triggers were particularly effective in bypassing existing defenses and confusing the source of the actions.

This method, dubbed “promptware,” raises serious concerns about how AI interfaces interpret user input and external data.

The researchers argue that such prompt-injection attacks represent a growing class of threats that blend social engineering with automation.

They demonstrated that this technique could go far beyond controlling devices.

It could also be used to delete appointments, send spam, or open malicious websites, steps that could lead directly to identity theft or malware infection.

The research team coordinated with Google to disclose the vulnerability, and in response, the company accelerated the rollout of new protections against prompt-injection attacks, including added scrutiny for calendar events and extra confirmations for sensitive actions.

Still, questions remain about how scalable these fixes are, especially as Gemini and other AI systems gain more control over personal data and devices.

Unfortunately, traditional security suites and firewall protection are not designed for this kind of attack vector.

To stay safe, users should limit what AI tools and assistants like Gemini can access, especially calendars and smart home controls.

Also, avoid storing sensitive or complex instructions in calendar events, and don’t allow AI to act on them without oversight.

Be alert to unusual behavior from smart devices and disconnect access if anything seems off.

Via Wired

You might also like

• Microsoft urges users to be on alert following high-severity flaw in hybrid Exchange deployments
• Here's a list of the best firewalls around today
• These are the best endpoint security tools right now