Feed aggregator

Jamilah, Zaynah and Latifah McBryde grew up wrestling one another in Buffalo, N.Y. Coaches recognized their talent, but they couldn't wear the required wrestling singlet due to their faith.

Cross-border payments have long carried the promise of something bigger than convenience. They imagine a more connected financial world—one where distance doesn’t delay progress, and borders don’t slow the movement of capital. As global commerce expands and digital economies accelerate, it becomes more urgent that we fully realize this vision and take advantage of the enormous benefits it affords.

But while we work towards this promise, we must manage complexities in getting there: fragmented systems, a patchwork of compliance requirements, and often, slow manual processes that can threaten to erode trust and efficiency.

It’s exactly here that agentic AI can have an impact. Not just by automating these manual processes, but by introducing intelligence and adaptability at the heart of payment infrastructure.

Smarter Systems for a Faster, Fairer Flow of Funds

Agentic AI refers to systems capable of independent decision-making, learning, and acting within defined objectives. Unlike traditional automation, which follows predetermined rules, agentic AI can work autonomously, responding to new data or situations in real time. In the context of cross-border payments, that adaptability opens up a range of opportunities.

Consider the coordination involved in sending money between countries: a single transaction might pass through multiple partners, banking systems, currencies, and regulatory frameworks.

When everything works perfectly, the process feels seamless. But when something goes wrong—delays, failed compliance checks, technical errors—resolution typically requires human intervention. Logs must be pulled. Partners contacted. Systems checked.

Agentic AI has the potential to radically speed up that process.

Intelligent systems trained on payment corridors, transaction histories, and partner system behaviors can flag issues early, pinpoint the likely causes, and suggest fixes fast. Instead of starting every investigation from scratch, agents can work alongside AI models that have already done the legwork, making operations more responsive and far less dependent on time-consuming manual deep dives.

Integration Without the Bottleneck

Another long-standing friction point for cross-border payments enterprises is the challenge of integrating new partners. Every integration means studying APIs, aligning data structures, and ensuring compatibility. It’s a time-consuming process that slows down innovation and expansion.

Agentic AI can simplify this. By learning the patterns and structures behind common integration workflows, it can assist in parsing APIs, mapping connections, and reducing the manual overhead typically required. This results in faster partner onboarding, more agile ecosystems, and a much smoother path to scalability.

This also rebalances the cost-benefit analysis for smaller institutions or partners looking to join global networks. Lowering the technical barriers means broader access—and broader access strengthens the entire system.

Compliance That Moves at the Speed of AI

No discussion of payments, especially international ones - is complete without compliance. From sanctions screening to AML checks, compliance teams face growing pressure to review increasing volumes of alerts with limited time and resources. And in many cases, today’s systems still rely on rigid rulesets that generate large numbers of false positives, overwhelming reviewers.

Agentic AI offers a way to improve both speed and accuracy. These models can learn from human decisions over time, categorizing alerts more effectively, recognizing patterns that traditional systems miss, and escalating only when needed. Crucially, AI doesn’t need to act alone. A layered approach, where multiple AI agents validate each other’s conclusions - can build greater confidence and traceability into the decision-making process.

To be clear: this isn’t about removing humans from the loop. It’s about giving them better tools - augmenting expert judgment with analysis that is both faster and broader in scope.

Laying the Data Groundwork

Of course, none of this works without good data. Agentic AI, like any system, is only as strong as the information it’s trained on. And for many financial institutions, the data challenge is real: important insights are locked in silos, scattered across SaaS platforms, legacy systems, and poorly indexed databases.

To take full advantage of intelligent systems, businesses need to invest in the foundational work: structured, unified data repositories that make relevant information accessible and usable. That means not only cleaning the data, but also breaking down internal barriers to visibility, rethinking fragmented architectures in favor of shared intelligence layers.

This will also reshape how financial firms choose and deploy their tech stacks. Stitching together dozens of narrow SaaS platforms may no longer be sustainable in a world where AI needs a full picture to function effectively. The focus will likely move toward more integrated environments, purpose-built for intelligent orchestration.

Even in these environments, agentic AI isn’t a silver bullet, but it is a powerful lever. Used wisely, it can help build payment systems that are faster, more reliable, and more inclusive. It can reduce the time it takes to connect with a new market, resolve an issue, or adapt to a regulatory change. And as the technology matures, cost barriers will fall, accessibility will rise, and open-source alternatives will drive further innovation.

Ultimately, this isn’t just about smarter or faster transactions, it’s about making the financial infrastructure of the future more human in its outcomes: more responsive, more accessible, and more aligned with the needs of real people moving money across borders every day.

The infrastructure we build today will define the shape of tomorrow’s financial landscape. And with agentic AI in the mix, that landscape has the potential to be more open, more intelligent, and more connected than ever before.

We list the best mobile credit card processor .

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

• Security researchers found nine flaws across two Ruckus products
• The flaws have not yet been patched, so users should beware
• Users advised to limit access to the wireless management environments

Almost a dozen vulnerabilities have been found in two Ruckus Networks products which could be abused to take full control over the network environments they operate in.

Ruckus Networks (formerly Ruckus Wireless) is a networking gear manufacturer, whose products include Virtual Smart Zone (vSZ) and Ruckus Network Director (RND).

VSZ is a virtualized network controller that manages Ruckus access points and switches. It is usually used by medium to large enterprises for centralized control, scalability, and advanced Wi-Fi management features. RND, on the other hand, is a centralized network management platform used for deployment, monitoring, and maintenance of large-scale Ruckus wired and wireless networks.

Meaningful disruption

At press time, the vulnerabilities remain unpatched, putting countless businesses at risk.

According to Noam Moshe from Claroty’s research arm Team82, these two carried nine vulnerabilities:

• CVE-2025-44957 – hardcoded secrets in vSZ that allow bypassing authentication and admin-level access using crafted HTTP headers and valid API keys
• CVE-2025-44962 – path traversal in vSZ that allows arbitrary file reads for authenticated users
• CVE-2025-44954 – vSZ has hardcoded default public/private SSH keys that allows anyone to connect to vulnerable devices with root access
• CVE-2025-44960 – vSZ has an API route with a user-controlled parameter that isn't sanitized, allowing execution of arbitrary operating system commands
• CVE-2025-44961 – command injection in vSZ allows an authenticated user to supply an unsanitized IP address to an OS command
• CVE-2025-44963 – RND uses a hardcoded backend JWT secret key, allowing anyone with it to forge valid admin session tokens
• CVE-2025-44955 – RND includes a "jailed" environment with a built-in jailbreak using a weak, hardcoded password to gain root access
• CVE-2025-6243 – RND includes a root-privileged user (sshuser) with hardcoded public/private SSH keys that allow root access
• CVE-2025-44958 – RND encrypts stored passwords with a hardcoded weak secret key and can return them in plaintext if compromised

Moshe reported his findings to Carnegie Mellon University’s CERT Coordination Center (CERT/CC), who confirmed that the flaws can be abused to cause meaningful disruption to businesses.

“Impact of these vulnerabilities vary from information leakage to total compromise of the wireless environment managed by the affected products. As an example, an attacker with network access to Ruckus Wireless vSZ can exploit CVE-2025-44954 to gain full administrator access that will lead to total compromise of the vSZ wireless management environment,” the organization explained.

“Furthermore, multiple vulnerabilities can be chained to create chained attacks that can allow the attacker to combine attacks to bypass any security controls that prevent only specific attacks.”

Severity scores have not yet been assigned, and Ruckus has not yet come forward with a patch.

Therefore, to mitigate the risk, CERT/CC advises network admins to limit access to the wireless management environments using the affected products, allowing a limited set of trusted users and their authenticated clients to manage Ruckus infrastructure through a secure protocol.

Via BleepingComputer

You might also like

• Sony, JBL and Bose headphones all affected by major Bluetooth security flaw which could let hackers spy on you via microphone
• Take a look at our guide to the best authenticator app
• We've rounded up the best password managers

• Apple’s M5 MacBook Pro might have been delayed until 2026
• No reason has been given for the apparent postponement
• It means Apple might launch two MacBook Pro models next year

For the last few years, Apple has launched a new MacBook Pro in the fall like clockwork – but this year, that could all change. That’s because a new report has claimed that the M5 MacBook Pro might be pushed back to early 2026, with Apple supposedly “internally targeting” that date instead of the regular late-year release window.

The surprising news comes from Bloomberg journalist Mark Gurman, who has been a reliable source of Apple leaks and rumors in the past. Gurman’s latest report contradicts his earlier reporting, which suggested that the M5 MacBook Pro would be unveiled near the tail end of 2025. That implies that something has changed Apple’s mind on when to push the new laptop into the limelight.

Gurman hasn’t provided a reason for the delay, so the cause of Apple’s decision is anyone’s guess. That said, the reporter noted that the timing is “fluid,” which suggests there’s a chance that the next MacBook Pro could make an appearance before this 2025 is over.

Still, launching a MacBook Pro early in the year is not without precedent, as the M2 MacBook Pro, for example, was released in January 2023. Still, that model has been the exception in the Apple silicon era, with the M1, M3 and M4 MacBook Pro editions all coming out in the fall.

Disappointing delays

(Image credit: Future)

What could be causing this delay? Gurman didn’t offer any hints here, and it’s not immediately clear what Apple’s thought process might be.

The M5 MacBook Pro is likely to be a minor update, with new Apple’s M5 chip being the main differentiator between it and the M4 MacBook Pro. That means issues with the chip could be one of the main reasons for the delay. Alternatively, Apple might simply be holding the device over until 2026 for sales reasons.

Next year could be a big one for the MacBook Pro, with a redesigned model expected to arrive in late 2026. This version is slated to come with a new design and an OLED display, marking the biggest update to the MacBook Pro since the last overhaul in 2021.

But with the M5 MacBook Pro potentially creeping into 2026, we’ll have to see whether the OLED MacBook Pro also gets punted into the long grass as a result.

You might also like

• The MacBook Pro could get Apple’s M5 chip before the iPad Pro, but don’t expect groundbreaking changes
• MacBook Pro M5 next year might be disappointing as rumor suggests the real fireworks are coming with MacBook Pro M6 with OLED in 2026
• Love the MacBook Pro? A new model could be on the way, with some significant upgrades

• There have been rumors of a Galaxy AI subscription
• Samsung has just confirmed many features will stay free
• But it's not clear how Galaxy AI may evolve in the future

Samsung certainly hasn't been shy about stuffing AI features into its phones and other devices in recent years, and there has been some uncertainty over whether these features might eventually cost money – but we now have a bit more clarity.

Speaking to Android Police at this week's Galaxy Unpacked event, a Samsung representative said that Galaxy AI features that are "on your phone by default" are always going to be available for free.

That includes Live Translate, Note Assist, Zoom Nightography, Writing Assist, Drawing Assist, Now Brief, and the Audio Eraser. There's a full list here of everything you get, AI-wise, when you buy a new Samsung phone or tablet.

What's not immediately clear is which AI features might not be covered – at the moment, there aren't any Galaxy AI add-ons that don't come bundled with devices, though perhaps we might see some introduced in the future.

The Gemini model

Samsung has introduced numerous AI features (Image credit: Samsung)

There is of course a freemium model to the Gemini AI assistant that now appears on most Android devices, including Samsung phones and tablets. Paying for Google AI Pro or Google AI Ultra gets you more features and access to better AI models.

Google AI Pro starts at $19.99 / £18.99 / AU$32.99 and includes extras such as 2TB of Google cloud storage, but you often get a few months free. If you buy one of the Google Pixel 9 phones, you get a full year of use before you have to start paying.

The newly-announced Samsung Galaxy Z Fold 7 and Samsung Galaxy Z Flip 7 come with six months free, and clearly Google is hoping that by the time that free period is up, you won't be able to live without all the AI goodness you get with a subscription.

Samsung has dropped hints about introducing a Samsung Health subscription package, granting you extra features for a monthly fee, so it's possible that some additional AI tools might be included in that – though we'll have to wait and see.

You might also like

• Photo Assist is one of the best Samsung Galaxy AI features
• 9 things we learned from Samsung's Unpacked July 2025 event
• We've spent some hands-on time with the Galaxy Watch 8

• Samsung says it didn't give the Z Fold 7 a bigger battery because users wanted it to prioritize other features
• Instead, the company prioritized improving the cameras, making the phone thinner, and giving it a wider cover screen
• However, with silicon-carbon battery tech now available, it seems like the company could still have upgraded the battery too

The Samsung Galaxy Z Fold 7 is an upgrade on the previous model in numerous ways, but one notable area where Samsung hasn’t made any improvements is the battery capacity, with both phones having a 4,400mAh unit.

This is especially odd as the battery capacity was one of the main things the company did improve on the Samsung Galaxy Z Flip 7 – and now Samsung has explained its reasoning.

Speaking to SamMobile, the company has claimed that in the case of the Samsung Galaxy Z Flip 7, a bigger battery was the most-wanted feature by users, but with the Galaxy Z Fold 7, there were other priorities from foldable fans, including “better cameras, a thinner and lighter body, and a wider cover screen.”

But we’re not entirely convinced by this explanation for a few reasons. First off, 4,400mAh really is a very low capacity for a phone of this size, with the much smaller Samsung Galaxy S25 Ultra for example having a far larger 5,000mAh battery.

So the Galaxy Z Fold 7 could certainly have done with a larger battery, and with the phone now being longer and wider (thanks to its larger screens) Samsung surely could have found space for a bigger battery if it hadn’t chosen to further slim the phone down instead.

The best of both worlds

The Samsung Galaxy Z Fold 7 (Image credit: Lance Ulanoff / Future)

But Samsung didn’t even necessarily need to choose between a thinner phone and a bigger battery. That's because some handsets – including the OnePlus 13 and the foldable Honor Magic V5 – now use silicon-carbon batteries.

This type of battery tech allows you to increase the battery capacity without increasing its size, and it’s a tech that Samsung is reportedly considering using for the Galaxy S26 series – a line of phones that’s arguably less in need of this, since the Samsung Galaxy S25 series already have decent battery capacities for their screen sizes.

So why couldn’t Samsung offer this tech in the Galaxy Z Fold 7? Perhaps it’s a matter of cost, with a silicon-carbon battery possibly pushing the price up. Or perhaps Samsung is exhibiting an abundance of caution in increasing capacities, so as not to risk a repeat of the widespread battery issues faced by the Samsung Galaxy Note 7.

The good news at least is that if Samsung really is considering silicon-carbon batteries for the Galaxy S25 line then the same tech may well be used for the Samsung Galaxy Z Fold 8 – it’s just a shame we'll have to wait another year to possibly get it.

You might also like

• 9 things we learned from Samsung's Galaxy Unpacked July 2025 event
• Hands on: The Samsung Galaxy Z Fold 7 captures the essence of Ultra in a ludicrously thin device, and I still can't believe it
• Hands on: I opened the case file on the new Galaxy Z Flip 7 FE, and it was all black and white

New Disney+ movies have been non-existent for a month. Finally, though, the streamer's latest film drought is over, albeit thanks to the debut of a new Disney Channel Original Movie in Zombies 4: Dawn of the Vampires.

But, fear not, Disney+ user, because more new movies will land on one of the world's best streaming services next week, too. In short: I'll be updating this article again very, very soon. In the meantime, if none of the forthcoming films are of interest to you, our best Disney+ movies article is full of great recommendations for you to check out instead. So, read that after you've perused what's on offer in this guide.

New Disney+ movies: JulyZombies 4: Dawn of the Vampires

Release date: July 11
Runtimes: 88 minutes
Age rating: TV-G (US); U (UK)
Directors: Paul Hoen
Main cast: Milo Manheim, Meg Donnelly, Kylee Russell, Chandler Kinney, Malachi Barton, Freya Skye, Julian Lerner, Swayam Bhatia, Mekonnen Knife, Lisa Chappell, and Jonno Roberts
Rotten Tomatoes (RT) score: TBC (critics); TBC (audiences)

• Watch Zombies 4: Dawn of the Vampires on Disney+

New Disney+ movies: JuneSnow White

Release date: June 11
Runtimes: 114 minutes
Age rating: PG (US); 9+ (UK)
Directors: Marc Webb
Main cast: Rachel Zegler, Gal Gadot, and Andrew Burnap
Rotten Tomatoes (RT) score: 40% (critics); 71% (audiences)

• Watch 2025's Snow White on Disney+

New Disney+ movies: coming soon

The Amateur will join Disney+'s movie library very soon (Image credit: 20th Century Studios)

As I mentioned in this article's intro, there'll be a few more new Disney+ movies to enjoy in the weeks to come. Here's a brief guide on all of the films you'll be able to stream soon:

• Transformers – July 16 (UK only)
• Transformers: Age of Extinction – July 16 (UK only)
• Transformers: The Last Knight – July 16 (UK only)
• The Amateur – July 17

Not up for a movie? Try our pick of the best Disney+ shows, see if there's a free Disney+ trial available, read our Disney+ review, or get the lowdown on the new Disney+ price points.

Marvel Rivals Season 3 launches later today (July 11), adding Phoenix as a brand new Duelist. So far, we've gotten an in-depth look at the new Heroe's moveset, a deep dive into the changes to Team Ups and Daily Missions, and an extended peek at Klyntar: Celestial Husk, the game's new Domination map.

Now that we're in the build-up of Marvel Rivals' next season, I'll be providing you with up-to-date information on what's set to change, what you'll want to be focusing on before the end of Season 2.5, and some impressions from my time playing the new season as part of my early access preview opportunity, provided by NetEase.

Marvel Rivals Season 3 key info

Release date and time
Downtime
Phoenix info
Patch notes
New map
Ranked changes
Jump to our live coverage as it happens

NetEase has now unveiled the next season of Marvel Rivals, revealing Phoenix's kit and a deep dive into the latest patch notes and balance changes. There's a new Domination map to look at, too: the dark and mysterious lair of the villainous Knull. If my early impressions are correct, Season 3 seriously shakes up the meta, especially with its new Team Ups for Heroes like Spider-Man and Wolverine.

Having covered Season 2.5's launch and the subsequent balancing changes to the Hero roster, I'm excited to see what this new era of Marvel Rivals has in store. I'm especially interested to see how the new release schedule - NetEase is planning a new Season, and two new Heroes every two months - will change the pace of the meta.

I've now spent hundreds of hours playing Marvel Rivals Competitive Mode since launch, mostly focusing on Vanguard Heroes like Venom and Magneto, but I'm really liking how Phoenix is shaping up pre-release. Here's the launch of Marvel Rivals Season 3 as it happens. I'll be covering today's launch live, focusing on downtime, and the early hours of Season 3.

Marvel Rivals Season 3 - cut to the chase

Looking for a quick overview? Here are the most important parts of Season 3 to know about:

• Start time: July 11 at 7AM EDT / 4AM PDT / 12PM BST (after estimated downtime)
• New Hero: Phoenix
• New map: Klyntar - Celestial Husk (Domination)
• Season event: Milano Repair Logs
• Hero balancing
• New Rank Rewards: Phoenix - Emerald Flames

Marvel Rivals Season 3 release date and estimated release time

(Image credit: NetEase)

Marvel Rivals Season 3 will launch on July 11, 2025. Downtime is set for 5AM EDT / 2AM PDT / 10AM BST, and based on previous season launches, I expect it to last for two hours. Given this estimate, here's when I'd expect Season 3 to launch in your timezone:

• 4am PDT
• 7am EDT
• 12pm BST
• 1pm CEST
• 8pm JST
• 9pm AEST

Marvel Rivals Season 3 downtime is estimated to be 1-2 hours, though this is subject to change. I'll be keeping this part of our live coverage updated during downtime, and will bring you the news when servers go back online. You can also look out for intel over on the Marvel Rivals Twitter account.

Marvel Rivals Season 3 downtime

(Image credit: NetEase)

Marvel Rivals Season 3 will arrive after a period of downtime. This is scheduled to kick in at the following time (depending on your timezone):

• 2am PDT
• 5am EDT
• 10am BST
• 11am CEST
• 6pm JST
• 7pm AEST

Downtime has typically lasted two hours for major seasonal updates to Marvel Rivals. I expect it to be the same this time around, but stay tuned on the live section of this blog for updates on when the game is back online.

Marvel Rivals Phoenix

(Image credit: NetEase Games)

Phoenix is the new Hero being added to Marvel Rivals Season 3. I've spent some time trying her out, and she's primarily a ranged Duelist who works best at off-angles. She plays very similarly to Hela, though she has a passive that stacks damage on enemies with each hit. Every third hit triggers an explosion, dealing extra damage, hitting any nearby enemies, and tagging them with a Phoenix mark. As such, Phoenix is extremely powerful when teams are grouped together.

As a Duelist, Phoenix is mostly about dealing huge amounts of damage, though she's also great at removing enemy support passives, like Loki clones and Peni Parker mines. Her Ultimate is a dive bomb attack that deals massive area of effect damage, before returning Phoenix back to where she triggered the ability. She has a new Team Up called Primal Flame, which buffs Wolverine with flame damage with his Primal Leap. She also applies life steal to Wolverine's attacks once the Team Up is triggered.

Marvel Rivals Season 3 patch notes

(Image credit: NetEase)

NetEase has now released the full patch notes for Marvel Rivals Season 3. It outlines loads of bug fixes for Heroes like The Thing, as well as some adjustments to in-game audio. There's a new Accessory system too, allowing players to add small trinkets onto Heroes for an extra bit of visual flair. You can read the full patch notes here.

Marvel Rivals Season 3 new map

(Image credit: NetEase)

Klyntar: Celestial Husk is a brand new Domination map set to be added in Marvel Rivals Season 3. It's the setting for Knull's seasonal debut, and it's appropriately menacing. Having spent some time with the new map, I can say it's one of the best released so far. There are tight corridors, perfect for the new Hero Phoenix to unleash AoE attacks in. There are a lot of possibilities for flanking and off-angle play, too. You can check out a deep dive on the new map in the video below:

Marvel Rivals Ranked changes

(Image credit: NetEase)

In addition to the new map, there's going to be a rank reset for Competitive Mode players. If you've reached the higher tiers in Marvel Rivals, expect to be demoted by seven tiers upon the launch of Season 3. This means that if you hit Diamond I in Season 2.5, you'll go down to Gold II. There's a Phoenix skin to earn by reaching Gold by the end of this season, and several nameplates and banners corresponding to Grandmaster, Celestial, and Eternity ranks.

Marvel Rivals Season 3 live coverageWelcome to my live coverage of Marvel Rivals Season 3

(Image credit: NetEase)

Welcome to my live coverage of Marvel Rivals Season 3, which is set to launch after some downtime. There's just fifteen minutes until the game goes offline, so let's get started.

Downtime triggers in ten minutes

Marvel Rivals will go offline in just 10 minutes, as downtime kicks in. This should last for around two hours, but I’ll keep you updated on exactly when the game is back online.

Costume Coins can be exchanged for skins

(Image credit: NetEase)

You can earn a Costume Coin from the Venom’s Bubble Pop event, which ends once Season 3 downtime kicks in (you can make it a permanent event by purchasing the Premium version). This can be exchanged for a number of skins in the Store. I went with the Master of Magnetism skin for Magneto.

Spider-Man gets his flame on

(Image credit: NetEase)

Ever-Flaming Bond is a brand new Team Up between Human Torch and Spider-Man. It adds a Burn-Tracer to Spidey’s kit, tagging enemies and sending Spider-Man flipping backwards. This opens up new combo abilities, and in my testing, it’s really powerful.

Season 3.5 will launch August 8

(Image credit: NetEase)

Thanks to the Season 3 roadmap provided by NetEase yesterday (July 10), we know that Season 3.5 will launch on August 8. This should be when we get Blade as a new Hero, and he's rumored to be a Duelist.

Marvel Rivals will go offline any minute

(Image credit: NetEase)

Marvel Rivals downtime is scheduled to end in a few minutes. You should still be able to finish off your match, but expect to be kicked from servers too.

Downtime has started

(Image credit: NetEase)

Marvel Rivals is now offline. I’ll still be covering the launch of the new season live, so let’s continue with some of the key changes coming once things go back online.

Marvel Rivals Server connection failed error message

When trying to load up Marvel Rivals, you'll now be greeted with the 'Server connection failed' error message. Don't worry, this us ask normal for downtime.

Phoenix could be a real contender

(Image credit: NetEase)

Phoenix will be added as a new Hero in Season 3, and she’s already feeling very powerful indeed. Her primary damage output is especially high, and thanks to a dash ability, she can escape dive characters quite easily.

• Experienced developers actually spent 19% more time on tasks when using AI, report claims
• Only 44% of AI-generated code was accepted by experienced devs
• Developers still feel that work is easier when using AI

Artificial intelligence might not be as beneficial to experienced developers as it can be to new starters and those who are yet to develop the right skills, new research has claimed.

A new study conducted by Model Evaluation & Threat Research (METR) has suggested not only were the developers less optimistic following the study, but the real results suggest that artificial intelligence actually ended up costing them time.

The study found that, with AI, devs spent less time coding and searching, and more time prompting, waiting for and more important, reviewing, AI output. An estimated 9% of the time went on reviewing and cleaning up AI-generated code, with AI suggestions generally on the right tracks but lacking in detail.

AI doesn't actually same these developers any time

Observing 16 experienced developers across 246 real tasks on mature open-source projects that they were already familiar with, the researchers analyzed how the developers interacted with popular tools from Cursor Pro and Claude 3.5/3.7.

Before the study, the 16 experienced developers in question expected to reduce task time by 24% when combining their expertise with artificial intelligence. After the study, they reduced their expectations to just 20%, however post-study analysis reveals that AI actually increased task completion time by 19%.

Fewer than 44% of the suggestions were accepted, with a lack of contextual knowledge and large, complex repositories highlighted as contributing factors to developer slowdown. The study also noted that the experienced developers already had high familiarity with the codebases, leaving little room for AI to add any meaningful value.

However, despite the slowdown, many developers continued to use AI tools because the work felt less effortful, making work feel more pleasant even if it wasn't faster.

"AI capabilities in the wild may be lower than results on commonly used benchmarks may suggest," the research paper concludes.

You might also like

• We've listed the best productivity tools
• Fancy an upgrade? Ask your boss for the best laptops for programming
• AI coding assistants are getting ever more popular - especially in this country

A federal judge in New Hampshire on Thursday blocked President Trump's executive order that attempted to end birthright citizenship, stopping it from taking effect anywhere in the U.S.

Elon Musk and his AI have been busy. So has the TSA. And Amazon. Were you paying attention?

The 900-page tax and policy package fulfills several key campaign promises but also makes cuts that could boot millions from health care and food assistance programs.

(Image credit: Brendan Smialowski)

Rep. Robert Garcia is the new top Democrat on the House Oversight Committee. At a moment when his party is craving more confrontation with President Trump, he says he's ready to lean into the fray.

(Image credit: Rod Lamkey)

In the wake of the deadly flash floods in Texas, state leaders are exploring whether to install more flood warning sirens. Such sirens can save lives if they're part of a larger warning system.

(Image credit: Brandon Bell)

Many Americans received an email from the Social Security Administration applauding the megabill's passage. Experts say it was misleading.

(Image credit: Saul Loeb)

President Trump has become a major booster of digital currency. The Planet Money team went to look into how the president's crypto enthusiasm plays out inside the industry.

Artificial Intelligence (AI) is being adopted across industries at remarkable speed. From finance to healthcare, AI is driving new services and unlocking new business models – fundamentally changing the way we all live, learn and work. But with progress comes challenges, and in the case of AI, fast adoption means that the infrastructure supporting it is under increasing pressure.

Data centers, once optimized for traditional enterprise workloads, are being pushed to accommodate entirely new operating patterns. The rise in high-performance computing means more power, more heat, and more volatility. Established systems are struggling to keep up.

AI workloads don’t just demand scale. They require IT infrastructure that can react to dynamic, unpredictable demand. And as organizations expand their use of AI, the supporting environment must evolve too.

Rack density is climbing quickly

One of the clearest shifts data center operators are experiencing is in rack density. Standard deployments have typically operated at around 10 kilowatts to 15 kilowatts per rack. But AI hardware - especially clusters of Graphics Processing Units (GPUs) - consumes much more power and generates far more heat.

In many AI deployments, racks now draw 40 kilowatts or more. Some experimental training environments exceed 100 kilowatts. This isn’t just about energy consumption. It’s a challenge for every part of the power chain, from uninterruptible power supply (UPS) systems to power distribution units (PDUs), to the facility’s own switchgear.

Older data centers may not be able to support these loads without major upgrades. For those expanding into AI, the layout, redundancy, and zoning of rack space needs careful planning to avoid creating thermal or electrical bottlenecks.

Cooling is reaching its limits

Conventional air cooling was never designed for today’s thermal loads. Even with hot aisle containment and optimized airflow, facilities are finding it hard to remove heat fast enough in high-density zones.

This is why liquid cooling is gaining ground. Direct-to-chip cooling systems, already common in high-performance cloud computing environments, are being adapted for data centers supporting AI and where densities exceed 50kW/rack. Immersion cooling is also being explored where space is tight or where energy efficiency is a priority where densities exceed 150kW/rack.

Installing liquid cooling involves significant changes - from plumbing and pumping systems to maintenance protocols and leak prevention. It’s a major shift, but one that is becoming necessary as traditional cooling approaches run out of headroom.

Load volatility is forcing a rethink

AI workloads behave differently from legacy compute. Training cycles can move from idle to peak and back in a matter of seconds. And inference jobs often run continuously, putting steady pressure on electrical and cooling infrastructure.

That variability puts systems under stress; power systems need to be fast and responsive and cooling systems must avoid overshooting or lagging behind, sensors and controls need to act in real time, not based on average load assumptions.

This is driving investment in smarter infrastructure. Software-based power management, predictive analytics, and environmental telemetry are no longer add-ons. They are becoming essential for resilience and efficiency.

Commissioning is getting more involved

Designing infrastructure for AI is one thing. Proving that it works under pressure is something else.

Commissioning teams are having to simulate conditions that didn’t exist just a few years ago. That includes sudden spikes in compute load, failure scenarios under high thermal pressure, and mixed environments where air and liquid cooling run side by side.

This means that simulation tools are being used earlier in the design process, with digital twins helping to test airflow and thermal modelling before equipment is installed. On-site commissioning now includes more functional testing, and more collaboration between electrical, mechanical and IT teams.

Power constraints are slowing progress

In some parts of the UK and Europe, getting access to the grid has become a significant barrier. Long connection times and limited capacity are delaying new builds and expansion projects.

This real and growing challenge is leading some operators to turn to on-site energy generation, energy storage systems, and modular buildouts that can grow in stages. Others are prioritizing regions with better access to power - even if they aren’t the original target location.

Cooling strategies are also directly affected. Liquid cooling systems require consistent energy supply to maintain stable operation. Any power disruption can quickly become a cooling issue, especially when workloads can’t be paused. And, in high-density environments, even brief interruptions to power can have thermal consequences within seconds - leaving no room for infrastructure to catch up after the fact.

Heat reuse is being taken seriously

AI workloads generate a lot of heat- and more than ever, operators are exploring ways to efficiently reuse waste heat.

In the past, heat recovery was often seen as too complex or not cost-effective. But with higher temperatures and more concentrated thermal output from liquid cooling systems, the picture is changing.

Some new facilities are being designed with heat export capabilities. Others are considering connections to local district heating systems. Where planning authorities are involved, expectations around environmental performance are rising, and heat reuse can be a strong point in a project’s favor.

Infrastructure is becoming more adaptive

AI is creating new expectations for data centre infrastructure. It needs to be fast, scalable and adaptable. Standardization helps, but flexibility is becoming more important - particularly as AI workloads evolve and spread from central hubs to the edge.

The next generation of data centers will need to manage high loads with minimal waste. They will need to recover energy where possible, stay efficient under pressure, and respond in real time to shifting demand.

This isn’t just about capacity. It’s about designing flexible systems that stay effective as conditions change.

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

Two managers at a San Diego construction project were trying to boost the morale of workers in the middle of the COVID pandemic. In this StoryCorps, hear how some monarch butterflies helped.

• NCA arrests four people suspected of participating in recent attacks
• M&S, Co-op and Harrods all hit by cyberattacks
• Arrested group are accused of organized crime, money laundering, and more

The UK National Crime Agency (NCA) has arrested four people suspected of taking part in recent cyberattacks against M&S, Co-op, and Harrods.

An NCA press release said the police apprehended two 19-year-old males, one 17-year-old, and a 20-year-old female in the West Midlands and London, all of whom had their electronic equipment seized.

They are suspected of Computer Misuse Act offenses, as well as blackmail, money laundering, and participating in the activities of an organized crime group.

DragonForce

"Since these attacks took place, specialist NCA cybercrime investigators have been working at pace and the investigation remains one of the Agency's highest priorities,” commented Deputy Director Paul Foster, head of the NCA's National Cyber Crime Unit. "Today's arrests are a significant step in that investigation but our work continues, alongside partners in the UK and overseas, to ensure those responsible are identified and brought to justice.”

In late April and early May 2025, three major UK retailers, M&S, Co-op, and Harrods, all suffered major cyberattacks.

The hit on Marks and Spencer affected stores nationwide, and resulted in “small changes” to store operations in order to protect customers “and the business." The retailer confirmed Click and Collect services were impacted by technical issues as a result, and some stores were unable to process contactless payments.

Co-op suffered differently, as threat actors (who named themselves “DragonForce”) walked away with a hefty archive of sensitive company data, including customer information.

"This data includes Co-op Group members' personal data such as names and contact details, and did not include members' passwords, bank or credit card details, transactions or information relating to any members' or customers' products or services with the Co-op Group," the company said at the time.

"Cyber attacks can be hugely disruptive for businesses and I'd like to thank M&S, Co-op and Harrods for their support to our investigations. Hopefully this signals to future victims the importance of seeking support and engaging with law enforcement as part of the reporting process. The NCA and policing are here to help."

You might also like

• M&S, Co-Op and Harrods got hit by a cyberattack: here's what retailers need to do to stop this
• Take a look at our guide to the best authenticator app
• We've rounded up the best password managers

• Perplexity’s new Comet browser promises an AI assistant that travels the web for you
• Comet joins other AI browsers aiming to beat Chrome and other popular platforms.
• Though AI companies are betting on their browser's enticing users, their mainstream appeal is uncertain.

Perplexity has officially made the move to web browsers, embedding its AI tools into Comet, its new Chromium browser.

It’s available now, initially only to some subscribers of Perplexity’s $200-a-month Max plan. At first glance, Comet is like most browsers, but Comet has a unique sidebar. You can highlight a word, sentence, or image, and Comet will discuss it with you. You can get a summary of an article, write an RSVP, or organize the itinerary of your next vacation. There's also a privacy benefit. Perplexity says all of the AI processing stays local and that it won’t train its AI using your site visits.

It makes sense. Browsers are central to modern work and life. If AI models can latch onto our browsing flow, then we will be using them all the time. But will it stick?

The browser space has seen a lot of failed Chrome clones, from Yahoo Browser to Internet Explorer 6. Chrome and Safari claim over 90% market share globally. The subscription price alone might be the biggest deterrent. Perplexity’s Comet is $200 a month. Compared to the price of 'free,' it would take a lot more than an occasional paper summary to make people pay up.

And while Comet shows how it might be useful as a way for Perplexity to meld its AI with a web browser, it's hardly alone in pursuing the idea of augmenting web browsers with AI. OpenAI is building a ChatGPT-native Operator browser. Currently, The Browser Company, Opera's Aria, Microsoft's Edge with Copilot, and others are all providing similar services. And Google is continuously adding AI to Chrome, offering quick overviews, summaries, and image explanations.

AI browsing future

Perplexity, OpenAI, and any other contender face the same challenge of getting people to switch. And while it's possible that Perplexity breaks through to a new group of users with the idea of an AI browser, they will then have to face off against each other as well as the AI-enhanced versions of Chrome, Safari, and the others on the rise from Brave to Firefox, each with their own pitch for a better AI, more privacy, or another appealing feature.

Or, maybe the concerns about technical issues and user privacy will keep the AI browser a niche product with limited appeal for hardcore users, a bit like Linux. AI might be the future of browsing, a brief fad, or something in between.

The question of what makes it worth the effort will need to be answered. If Comet and others can streamline the online experience and save us a lot of time, they'll be popping up everywhere, but for now, they’re novelties priced at a premium. We'll have to see if the AI-enhanced browser can find the users it needs to last.

You might also like

• Perplexity AI's Comet browser will streak across the web this month
• Perplexity's voice assistant offers a Siri alternative for iPhones
• AI comes to the URL with a new web browser that answers you back