If you're an Apple Music subscriber who's never thought to get into its dedicated and available-for-no-extra-cost Apple Music Classical app, now would be a good time to hit the download button.
Why? Because Apple Music Classical, the place where Apple Music has housed its massive classical music catalog since its launch in March 2023, (much of it in Dolby Atmos), just added three excellent new features: Listening Guides, Personalized Recommendations and – my favorite – Editorial Stations.
The trio of new perks are exclusive to the app (which is a shame, since just this month Apple Music Classical became available on the web), but Listening Guides is available in over 100 works from launch. It's described by Apple as "a groundbreaking new feature that takes users inside a notable work of music as they listen, highlighting details and explaining a work in real time as it unfolds phrase by musical phrase."
Every day is a learning dayI tried it while streaming Wolfie M's as no one calls the hugely prolific composer, who died tragically young) Clarinet Concerto in A Major, and it's like getting CliffsNotes in real time as you stream. Text appears happily on my iPhone, mentioning florid arpeggio runs, intense high trills, and smooth interchanges from the string section to guide us back to the major scale.
It's delightful – I even now know that in 1791 the clarinet was still a relatively new instrument, and that Mozart actually wrote this concerto for the bass clarinet. Every day is a learning day.
Right now, Listening Guides is available in English, French, German, Japanese, Korean, and Simplified Chinese. But Apple says more languages and works will be added – and won't those writers have their good work cut out for them? I look forward to it.
Personalized Recommendations, meanwhile, does what the name suggests, using your listening history to suggest related offerings, as well as new recordings of works you previously listened to.
Both of these perks are hugely welcome in a genre that has felt a little stoic to me thus far (it often feels like discovery across streaming sites is a little stilted when it comes to classical – you're expected to know what you like), but Editorial Stations is by far my favorite.
As you've probably guessed, it's a selection of continually playing playlists (like radio stations but without the DJ or ads) arranged by instrument, composer, period, or genre, curated by Apple Music Classical’s editors. I'm currently listening to one titled Opera, and Bizet's Carmen is all around me in Dolby Atmos.
Added to yesterday's launch of DJ in Apple Music to help budding disc jockeys build epic sets, plus the unavoidable fact that Apple Music Replay walked all over Spotify Wrapped at the end of last year, you can say this for Apple Music and its free-to-users Apple Music Classical sibling: it's good. It's one of the best music streaming services out there, and then some.
You may also likeOne of the downsides of the best earbuds is that over time, they get dirty: in addition to external debris such as dust and dirt, there's also ear wax to get in the way of your listening pleasure. And it seems that the design of the Pixel Buds Pro 2 makes them not only prone to getting gummed up with unwanted additions, but these can affect the sound more than you might expect.
That's why Google has published a published a guide to safely cleaning your Pixel Buds Pro 2 (via 9to5Google for spotting it). According to Google, the rear vents and small openings on the outside of the buds can get blocked, and that can "greatly affect" the audio quality.
The good news is that it's simple to solve. All you need is micellar water – you'll usually find it with the skincare and beauty products in a store – plus a paper towel and an extra soft toothbrush. Google advises against using detergent, sprays or alcohol.
The advice here is specifically for Google's earbuds. We've got our own guide to earbud cleaning, including AirPods, right here.
These vents need to be kept clear. (Image credit: Google) How to safely clean Pixel Buds Pro 2The star here is the micellar water, which is used by many of us to get rid of makeup without feeling like we've sanded half our face off. It's very good at shifting things, and that makes it ideal for your Pixel Buds too.
Google recommends pouring normal water into a cup and micellar water into another, and then swirling the toothbrush in the micellar water. Once you've done that, hold your bunged-up bud in one hand and use the toothbrush to brush the mesh in "deep circular motions" for 15 seconds.
Google recommends blotting the earbud on the paper towel and repeating the brushing two more times before rinsing the brush in the regular water. Once you've done that, it's time for one more brush – this time without the micellar water.
Once you've done both buds you'll need to let them air dry. Google recommends at least two hours, or longer if they need it.
You might also likeThe latest update for strategy game Sid Meier's Civilization 7, update 1.1.1, is here and includes a healthy mix of both free and paid content for players.
In my eyes, the most significant addition here is the reintroduction of Quick Move. This setting disables movement animations for your units and has been a staple of Civilization games for the last few years. I personally love the detailed animations in Civilization 7 but, when you're playing lots of back to back matches then turning them off is a massive time saver that increases the pace noticeably.
Map generation has been tweaked, with the introduction of a whole new default start position option aptly known as 'Standard'. This will feel familiar to Civilization 6 veterans, as it offers randomization and a lesser focus on more predictable continents.
The previous default setting has been renamed 'Balanced' and can be selected in Advanced Options. It will also remain the default option in multiplayer. Developer Firaxis Games also has reassured fans that more map variety, generation, and new, larger maps, are still in the works.
You are now able to rename both settlements and commander units freely too, allowing for more freedom and customization.
Some general UI improvements have been implemented, mainly focusing on increasing the visibility of important information like your current age progress or unit health. You will also find a new 'Restart' button on the main menu, which lets you restart your current game with the same settings in the first few turns.
Finally, a new natural wonder, Mount Everest, has been added. It offers useful tile bonuses and reveals all mountain tiles on your continent when it's first discovered.
Moving over to the realm of paid content, update 1.1.1 rounds off the premium Crossroads of the World Collection with the addition of a new leader and two new Civs. The new leader is Simón Bolívar, who specializes in Expansion, Militaristic play with multiple buffs geared towards maintaining war support and rewarding successful conflicts.
The new Civs are Bulgaria and Nepal, which both come with their own unique ability, civilian unit, and military unit. You learn more about them by visiting the full patch notes on the game's official website.
Although the update is available now for Sid Meier's Civilization 7 on PC, PlayStation 5, PlayStation 4, Xbox Series X, Xbox Series S, and Xbox One, it will come to Nintendo Switch at a later date.
You might also like...Thanks to a recent Monster Hunter Wilds Showcase event, many fans have been left clamoring for more information on a certain sea monster. But is Lagiacrus coming to Monster Hunter Wilds later this year?
Monster Hunter Wilds is another step forward for the long-running co-op action franchise. For one, its huge open world is more seamless, and more alive than ever before, with distinct seasons that promise to mix up hunts in interesting ways. Now that we've published our Monster Hunter Wilds review, the game has landed on our list of best multiplayer PC games.
Furthermore, Capcom has now detailed the next few months of post-launch updates, which you can read about in our Monster Hunter Wilds Roadmap. New monsters are the headline additions set to arrive, including Mizutsune and a new Ray Dau variant.
Here's everything you need to know about Lagiacrus being added to Monster Hunter Wilds. As more info is revealed, we'll be sure to update this page. For now, let's break down the recent teaser trailer, and look at when the fan-favorite monster last appeared in the series.
Is Lagiacrus coming to Monster Hunter Wilds?Capcom has teased the return of Lagiacrus, showing the sea serpent at the end of a recent showcase event for Monster Hunter Wilds. Given this, I'm confident in saying that yes, Lagiacrus will come to Monster Hunter Wilds in the future.
The only caveat here is that while Lagiacrus is indeed shown in the teaser, Capcom is yet to actually confirm that the monster shown is Lagiacrus, or the details on when they could arrive. As such, keep an eye on this page for updates once we get them.
The reason I'm confident that Lagiacrus is coming to Monster Hunter Wilds is that the brief tease showed a green-scaled monster, with plenty of underwater sound effects. Lagiacrus debuted in Monster Hunter Tri, notably featured in a hunt that made use of the game's underwater combat. Hopefully this style of hunt returns for Wilds.
This is all great news for Monster Hunter fans, who voted Lagiacrus the overall third favorite monster from the series in an official 2023 poll.
When could Lagiacrus arrive in Monster Hunter Wilds? (Image credit: Capcom)We don't know exactly when Lagiacrus will launch in Monster Hunter Wilds. What we do know is that the Lagiacrus teaser mentioned previously was paired with the text "Second Title Update".
This is likely Title Update 2, which is set for Summer 2025. As you can see in the image above, the roadmap for Monster Hunter Wilds extends into the end of May, before things get a bit murky. I expect Capcom to air another Monster Hunter Showcase in June that'll detail Title Update 2, and hopefully give some more info on what to expect from Lagiacrus.
Monster Hunter Wilds Lagiacrus FAQ (Image credit: Monster Hunter Wilds) Is Lagiacrus in Monster Hunter World?Lagiacrus never appeared in Monster Hunter World. They were also absent from the Monster Hunter World: Iceborne expansion.
What games is Lagiacrus in?Lagiacrus first appeared in Monster Hunter Tri, as the game's flagship monster. It then came to Monster Hunter Generations, though only on land, providing a very different fight when compared to the underwater action in Tri. In terms of spin-offs, Lagiacrus also notably featured in Monster Hunter Stories and Monster Hunter Stories 2.
Is Lagiacrus in Monster Hunter Wilds?No, Lagiacrus is not currently in Monster Hunter Wilds. It has been teased however, and will likely arrive in Summer 2025, though we're still waiting on confirmation.
Is Lagiacrus in Monster Hunter Rise?Lagiacrus was never added to Monster Hunter Rise, despite being much-requested by fans. In an interview with IGN, Monster Hunter producer Ryozo Tsujimoto explained why:
"Unfortunately Lagiacrus is just a very technological…difficult monster to implement. It's very snake-like, its parts are all kind of touching the ground at different moments,” Tsujimoto explained. “So it's very like a writhing sort of ground-based monster.”
You Might Also Like...Google just announced Gemini 2.5, and it's a major upgrade to Gemini that the company is calling its "most intelligent AI model" yet.
Announced on the company's blog, Google revealed the experimental version of 2.5 Pro, which is available today for all Gemini Advanced subscribers. More 2.5 models will arrive in the future.
Google's Gemini 2.5 models are a new generation of thinking models that are able to reach "a new level of performance by combining a significantly enhanced base model with improved post-training."
2.5's thinking capabilities will be implemented into all future Google AI models, which the company says will allow them to "handle more complex problems and support even more capable, context-aware agents."
So what does this mean? Well, Google is doubling down on its impressively frequent AI updates, and this time we're getting better reasoning capabilities than ever before.
Available right now, you can access Gemini 2.5 Pro Experimental simply by selecting the model in the Gemini app or directly in Google's AI Studio. You'll need a Gemini Advanced subscription to see this as an option.
Pricing for the improved model (for those who want to use it for scaled production use) will be announced in the coming weeks, and more 2.5 models are expected to launch in due course.
Gemini 2.5 Pro is impressive (Image credit: Google)Google shared some benchmark results for Gemini 2.5 Pro Experimental and the results are seriously impressive.
The new AI model scores 18.8% on Humanity's Last Exam compared to 14% for ChatGPT's o3-mini and 8.6% for DeepSeek R1. Humanity's Last Exam is the most thorough and difficult AI benchmark, so to score substantially higher than its competitors is no mean feat.
18.8% is the highest score we've ever seen on Humanity's Last Exam (without tool use). Google is calling Gemini 2.5 Pro's reasoning capabilities "state-of-the-art" and it's clear to see why.
Google continues to drive forward with its AI development at a rapid pace. Just last week the company made Gemini Deep Research free for all and followed that up with improvements to its impressive AI podcasting tool, NotebookLM.
We'll be testing Gemini 2.5 Pro and putting the new Experimental model through its paces, so stay tuned to TechRadar for further Google AI coverage.
You might also likeGoogle has finally added humans to the long list of things Pixel Studio on its Pixel phones can generate AI images of. The technology, however, still seems to be in its infancy because the results are, shall we say… variable, verging on terrifying?
For instance, I asked Pixel Studio to generate an image of “an athlete” and it showed me an image of a man without any skin at all! (See the image above.) Not only was the image quite shocking, it wasn’t anywhere near what I’d asked for.
To be fair to Google, it does say “This feature is being fine-tuned, so you might get unexpected results of blocked images” when you open Pixel Studio, but I didn't think that would mean showing me somebody without any skin!
Pixel Studio also did a far better job at generating images of people in other professions than “athlete”, although it does seem to default to males when I asked to see things like a doctor, a paramedic, or a police officer.
Here's what I got from Pixel Studio when I asked for "a doctor". Why the long face? (Image credit: Google) Pixel 9-onlyPixel Studio is Google's on-board AI image generation app included on Pixel 9 phones - it won’t work on earlier models or other phones. The announcement about people being added to Pixel Studio was made a couple of weeks ago but has only just rolled out.
As you’d expect for an AI image generator, it works from a text prompt. You simply type in what you want to see and Pixel Studio generates an image for you. For comparison, Apple’s Image Playground does a similar job, but on the iPhone.
Compared to Image Playground, which seems to only be able to produce cutesy, cartoon-like images, Pixel Studio can produce much more realistic images.
By enabling proper photo-realistic human AI character generation, Google is treating the Pixel Studio app more like the best AI image generators, like say Midjourney or Imagen 3.
Of course, Pixel Studio can make cartoon images too, if you ask it to. It defaults to photo-realistic for images, but you can ask for things to be generated “in a cartoon style” or “in a watercolor painting style”, etc.
Here's what Pixel Studio showed for "a paramedic". (Image credit: Google) Rolling outThings have improved for AI image generation recently. Google has just added native image generation and editing to its Gemini AI, however, it’s currently only available inside Google AI Studio, its developer tool, which is free to access, but not as easy to use as the main Gemini app.
ChatGPT has also just added native image generation to its 4o model, and Grok 3 can now also edit images. We'll be testing all of these new image generation upgrades and pitting the tools against each other. Stay tuned to TechRadar to make sure you don't miss our future comparisons.
Windows 11 looks like it’ll get its basic search functionality seriously bolstered, with a natural language searching feature progressing nicely through testing – but it’s only for those with Copilot+ PCs.
These ‘local semantic search’ powers have arrived in the latest preview release in the Beta channel (build 26120.3585, as noticed by Neowin), for Copilot+ laptops with AMD or Intel processors. Furthermore, they’ve also turned up in Release Preview for Snapdragon (Arm-powered) Copilot+ PCs.
The move means you can use natural language for a search query in Windows, such as “find photos of me with my dog” or “find that document which is my holiday packing checklist,” rather than having to remember any exact file names.
This doesn’t just work in terms of searching your files and folders (meaning File Explorer), but also with searches in the Settings app – so you can perform queries such as “show me the Bluetooth devices connected to my PC” to pick out another example.
All of this leverages the power of the NPU of the Copilot+ PC. All the processing is done locally, with no data sent to the cloud, which obviously means that you don’t have to be connected to the internet.
Also worked into this particular piece of functionality is the ability to use this AI-enhanced search to find photos in the cloud, should you wish.
Microsoft explains: “In addition to photos stored locally on your Copilot+ PC, photos from the cloud will now show up in the search results together. In addition to searching for photos, exact matches for your keywords within the text of your cloud files will show in the search results.”
This is for OneDrive only for now, but Microsoft says it’s working to bring support to third-party cloud storage services.
As for caveats, right now, searching for Windows settings will only work within the Settings app itself, but the eventual aim is to have these results flagged from the search box on the desktop taskbar (as is the case with the normal search function).
It’s worth noting that if you are a Windows tester in the Beta channel, this feature is only gradually rolling out, so you may not see it for a while yet (and you may need a couple of reboots of your Copilot+ PC to fully trigger the AI-bolstered search when it does turn up).
(Image credit: Future) Analysis: A golden gift for Copilot+ PCsA natural language search is a nifty ability for Windows 11 search, and a good use of that NPU. Windows 11’s search powers have always been rather sluggish and lacking, often proving not just slow, but failing to find anything useful, and flagging up weird results (or pointless web content). It’s been a long-complained-about area of Windows (the same is true of Windows 10), so hopefully this will go some way towards pepping up the overall experience, as well as making the functionality a lot more convenient.
Of course, with semantic indexing, Microsoft’s AI is effectively cataloguing (read: rifling through) all your files in order to have the search work in a more timely and responsive manner. Hence the reason why the company clarifies that all processing and data is stored locally, and doesn’t leave your PC – due to the potential privacy implications otherwise. This is especially important because as Microsoft notes elsewhere: “Semantic indexing is enabled by default on Copilot+ PCs.”
You can turn it off, mind, or you can selectively exclude certain files or folders (or drives). All these options are housed in the Settings app, in Privacy & Security > Searching Windows > Advanced indexing options.
This AI-driven search feature was seen in the Dev channel a while ago, so the fact that it has progressed to Beta (and Release Preview for Snapdragon-powered Copilot+ PCs) suggests it’s close to arriving in the finished version of Windows 11 for these devices.
Still, we can never be sure any feature in testing will see the light of day, but it seems very likely in this case. As it’s a complex piece of functionality, though, Microsoft could still have some tweaking and debugging on its plate. This is something Microsoft really needs to nail for release, as it’ll show off a considerable advantage of a Copilot+ PC if it turns out well – which will be a much-needed addition to the list of selling points for these computers.
You may also like...Consumers might not realize it, but their internet experience is highly dependent on networks of servers called CDNs, or Content Delivery Networks. CDN servers consist of a network connected to different Points of Presence (PoPs) around the world, and—as their name suggests—are designed to deliver content to end users as quickly as possible. They function by minimizing the distance that data needs to travel to best distribute a user’s traffic quickly and efficiently, according to their request.
Historically, the best CDN providers have been focused on specific geographic regions, like Western Europe or the United States. Today, however, more companies are beginning to take a multi-CDN approach.
At a basic level, multi-CDNs are two or more CDNs aggregated into a single system. When a request comes in from an end user, a multi-CDN system can automatically determine which of its CDNs provides the fastest route—broadening coverage areas while reducing overhead. The result is a more robust, better-optimized content delivery system, which many businesses are finding indispensable for their increasingly globalized customer base.
Single-CDN vs. Multi-CDNFor businesses contemplating whether to choose a single-CDN vs. multi-CDN approach, their decision will potentially come down to the type of content they are trying to deliver. The key concept at play here is static versus dynamic.
Dynamic content—for instance, fluctuating ecommerce or live-streaming platform offerings—can typically be effectively delivered through single-CDN approaches. Some multi-CDN providers have a switching layer that continuously switches between the CDNs in their mix. Decisions are then made based on performance to change content delivery providers for a better result for the client. Although this can occasionally create issues for an end user watching a live stream, in most cases it does not.
Nevertheless, as the digital landscape shifts toward more dynamic content, vast swathes of internet content remain static. Static content (e.g. images, CSS files, data objects, JavaScript, and prerecorded videos) change infrequently, making it perfectly suited for high-performance multi-CDN distribution. For example, the AdTech and MarTech industries often depend on multi-CDNs for banners and other pre-rendered content that remains consistent for all end users.
Since the content doesn’t change, it simplifies delivery and allows it to load faster from any geographic caching point. Additionally, many streaming platforms offer static, “on-demand videos”, which can also benefit from multi-CDNs. By caching content across multiple CDNs in a coordinated way, multi-CDNs reduce the number of requests that need to be made to the media’s origin server, reducing traffic costs from both an ingress and egress perspective.
In all these cases, multi-CDNs can route traffic to the best-performing CDN, based on their region and request, to reduce latency and improve the user experience. In comparison to a single-CDN, global providers who offer multi-CDN solutions can all but guarantee uptime, availability, and reliability—if one CDN has an outage or has performance issues, another can take over to ensure continuous availability.
For organizations with a global customer base, multi-CDNs offer a better experience for end users, delivering content quickly from the geographically closest node, providing the most efficient path – especially when compared to a region with some hyperscalers’ – irrelevant of their location. Moreover, for organizations aiming to futureproof and remain competitive, prioritizing speed and efficiency has become essential—particularly as the world becomes driven by AI technologies.
Multi-CDNs in an AI WorldAI has been integrated into almost every application. This has, inevitably, created a demand for high availability—one that multi-CDN strategies are uniquely well-equipped to meet.
AI solutions depend on high-quality data and content to succeed. These foundational elements provide the context necessary for AI to generate accurate and relevant outputs. While not dependent on multi-CDNs, AI models leverage them to retrieve data more quickly from the nearest server, delivering dynamic content with high speeds and low latency.
It is important to note that AI has also raised users’ digital expectations. Consumers now require faster, more accurate, and more relevant results across their digital experience. Thus, any solution within one's portfolio must match the speed and capabilities of AI technology. Multi-CDNs are also able to rise to this challenge by delivering dynamic content with higher speeds and lower latency, ultimately expanding global reach, and enhancing the user experience.
In addition to being delivered faster, content is also delivered more reliably. Most organizations that deploy single-CDN technology, of course, have a back-up provider in place in case the first one goes down. However, spinning up that second provider can take a long time—potentially frustrating customers and damaging business operations. However, there are various types of Multi-CDNs, including configurations of multiple single CDNs.
In some cases, a valuable switching layer is added, enabling real-time delivery optimization by dynamically switching between CDNs every second based on performance. This approach is particularly beneficial in environments where it eliminates the spin-up time required for a secondary CDN, while ensuring seamless and uninterrupted content delivery. Meanwhile, additional content caching can be accomplished, further reducing the cost of traffic both to and from its origin.
While AI does not rely directly on using multiple CDNs, the infrastructure that supports AI services—such as content delivery, data storage, and real-time inference—works better with a multi-CDN approach. By pooling the strengths of multiple individual CDN providers into one optimized system, it ensures a level of redundancy and reliability that no single CDN could possibly offer on its own.
Choosing a Multi-CDN ProviderOver the years, many organizations have turned to hyperscalers for their computing and storage needs. While these companies offer a wide array of resources, they have a dominant status in the market; and therefore, have been able to charge what some would argue are excessively high fees.
These providers will usually offer credits to welcome new buyers to their platforms, but they often do not extend a tiered or competitive price point once those credits expire, leaving organizations in a tight financial bind. The result is that companies will jump from provider to provider using up free credits, before seeking out a more cost-effective option. Additionally, hyperscalers often deploy multiple single-CDNs instead of a multi-CDN, leaving their users vulnerable to down-time and lost revenue.
As strategies shifted, enterprise to SMB organizations have begun to implement their own solutions involving multiple single-CDN providers. However, the challenges involved with this are often significant, thus creating a need for multi-CDN providers. Leading multi-CDN providers emerge as significant disruptors in this space by offering CDN services that compete with the hyperscalers’ – at a lower cost and with improved performance and reliability.
Where to Go from HereCDNs are designed to deliver content as quickly as possible. With the modern internet user, there are more dynamic expectations of the digital experience that multi-CDNs are better equipped to handle, especially with AI-based services and performance.
Organizations should strongly consider integrating a multi-CDN approach with a provider who can increase efficiency, uptime, availability, and speed with a global network. While a multi-CDN approach will not fit every scenario, more businesses are finding that the ease, flexibility, and scale vastly exceed those of traditional single-CDNs for a wide variety of use cases and workloads.
We've compiled an extensive list of the best AI tools.
This article was produced as part of TechRadarPro's Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro
Google Maps is one of the apps that gets the most love from Google in terms of regular updates, and we've spotted another couple of upgrades rolling out to the mapping app in recent days – upgrades to add some extra convenience to your travels.
Firstly, the app is changing up how it displays route overviews. As Android Police has noted, the info card you see when you select any place on the map on Android now shows you more details about parking availability and any applicable toll fees.
This information was previously available in Google Maps, but you had to dig around a little bit to find it – now it's included on the main route overview pop-up, alongside the estimated travel time and the current traffic conditions.
It's a small tweak but one that's potentially very useful. It brings together more of the data Google Maps has in the same place, which means you're more clued up about your journeys and you can get moving more quickly. It doesn't look as though the update is live for everyone yet, but it should be eventually.
More AI in Google MapsGemini is starting to let you ask about places you view on Google MapsGemini's 'Ask about place' chip for Google Maps has started showing up.✅ Details - https://t.co/wAU6mY4VJQ pic.twitter.com/q6ne9kbxbwMarch 25, 2025
Our second new feature concerns Google's favorite tech of the moment: AI. As spotted by Android Authority, if you select a place on the map in Google Maps for Android, then activate Gemini (with a long press on the power button for example), you get a new Ask about place chip you can tap.
Gemini being able to tell you about places isn't new, but the addition of a specific shortcut means you can get some AI-generated information about a location much more quickly from Google Maps. Just watch out for hallucinations, as always.
For example, you could ask when a certain place opens or closes, or ask if it's suitable for kids. The responses will then be displayed in a Gemini overlay on top of the screen, as usual, and you can ask follow-up questions if needed.
We can expect much more in the way of Gemini integration in Google Maps and Google's other apps in the future, too. Google has been busy pushing the AI bot into as many apps as it can, while it continues to work on retiring Google Assistant.
You might also likeGoogle is now pushing out Android Auto 14.0 to phones, and while the update doesn't feature anything major in terms of the user-facing interface, it does prepare the way for a big change further down the line: replacing Google Assistant with Gemini.
As spotted by 9to5Google, there are hidden references in the code for the latest Android Auto app that mention Gemini and Gemini Live, suggesting you'll soon be able to talk to Google's more modern AI model from your vehicle.
There are text strings here too, telling you that "Gemini is now the personal AI assistant in your car" and asking for feedback on the "Gemini experience" – clear signs that the switch is coming sooner rather than later.
We also get a mention of "message summary content", so it sounds as though Gemini might be able to summarize your messages as they come in, so you can keep your eyes on the road. Let's hope Google can make sure those summaries are accurate, and sensitive to bad news.
Out with the old, in with the new Google Assistant had a good innings (Image credit: Shutterstock)None of this is any surprise of course. Google has been pushing Gemini as a Google Assistant replacement for a while now, and earlier this month announced that Google Assistant would be retired this year as far as smartphones go.
Getting other devices on Gemini – smart speakers, wearables, car dashboards – is going to take longer. Gemini is more demanding than Google Assistant in terms of the processing power and memory it needs, so older devices may miss out.
In terms of timescales, a lot of this is still up in the air. Google, understandably, doesn't want to commit to a schedule – but it would seem Android Auto (which is just an extension of Android, after all) is next in line for a Gemini upgrade.
Google Assistant had a good run, first appearing in 2016, but Gemini is now the future. You may have already noticed certain minor features being removed from Google Assistant on smart devices as Google prepares to discontinue its development.
You might also likeOura has announced a major upgrade to its activity tracking algorithms and platform that it says will more accurately detect steps and active calories burned during exercise.
"Today, we’re excited to announce all-new updates to Oura’s activity tracking platform and algorithms, empowering our members to more accurately track all the ways they move—from yard work to yoga," the company said in a press release.
Oura says it has developed a "new and improved" algorithm that will improve the accuracy of step counts and active calories burned. Dubbed "Real Steps", the company says the change is the next iteration of its step-counting algorithm.
Rather than estimating step count using generic movement, the new update "enables Oura to act more like a pedometer, using a more advanced machine-learning model to more accurately determine when ring movement is a step."
This should mean more accurate step counting, but I've got some bad news. Oura says you will likely see a drop in your step count compared to what you're used to, with members seeing an average decrease of 20%.
So if you've been using your Oura Ring to walk 10,000 steps a day, the chances are you might only have been walking 8,000.
Oura Active Calories (Image credit: Future)Oura's Active Calorie burn metrics now include the intensity of your movement during exercise, as measured by your heart rate. As such, if you're doing more intensive workouts like high-intensity interval training, you might see an increase in calories burned.
However, lower-intensity exercises like walking and yoga will likely yield a decrease in what you're used to seeing.
These changes aren't specific to any one model, so they should be available on the company's best smart ring, the Oura Ring 4, as well as the older Oura Ring 3.
Oura didn't specify how the change is rolling out, so just make sure to keep your Oura app and your Ring's firmware up-to-date.
While getting more accurate fitness tracking on a health and fitness device is always welcome, it sounds like some users might be in for a bit of a shock when they get their new step counts for the first time.
You may also like...Sorry, but The Witcher 4 is probably another two years away, at least according to CD Projekt.
As spotted by PC Gamer, a snippet of information from CD Projekt's latest earnings presentation has clued us in as to when we can expect The Witcher 4 to release. Or rather, that it's probably not going to hit shelves and digital storefronts until 2027 at the very earliest.
"Even though we do not plan to release The Witcher 4 by the end of 2026, we are still driven by this financial goal," says CD Projekt, commenting on its share-based incentive programs.
"We are not going to announce the precise launch date for the game yet," added CD Projekt CFO Piotr Nielubowicz. "All we could share now to give more visibility to investors is that the game will not be launched within the time frame of the first target for the incentive program, which ends December 31, 2026."
We at least know that The Witcher 4 has entered full-scale production. According to the presentation, the upcoming game has 411 staff members working on it as of February 28, 2025.
A remaining 296 developers are working on other projects at CD Projekt, including Orion (the Cyberpunk 2077 sequel), Sirius, Hadar and others.
You might also like...Just as we think we’re getting one step ahead of cybercriminals, they find a new way to evade our defenses.
The latest method causing trouble for security teams is that of device code phishing, a technique that tricks users into granting access to sensitive accounts without attackers needing to steal a password.
Microsoft recently issued a warning about a particular device code phishing campaign being conducted by Storm-2372, where a supposed Russian-backed threat actor was wreaking havoc by hijacking user sessions through legitimate authentication flows. These attacks are trickier to detect than usual given that they exploit real login pages (rather than the spoofed versions that traditional phishing techniques relied on) and are capable of bypassing multi-factor authentication (MFA).
The recent warning from Microsoft will most likely be the first of many. Various other platforms follow the same style of authentication flows and attackers will most likely replicate the technique elsewhere. It is down to security teams once again to identify the warning signs of this new breed of phishing, and implement the best cybersecurity practices to get ahead of the curve.
Understanding device code phishingUnlike traditional credential phishing attacks, device code phishing is unique in that there is no need to directly steal a password. Instead, attackers manipulate victims into handing over access to their accounts by exploiting authentication methods designed to make logging in easier.
They start the same way as most email attacks do: through social engineering. By impersonating a trusted colleague or IT administrator, the attackers send an email invitation to an online meeting (often a Microsoft Teams meeting) that looks legitimate. The email is designed to appear normal – for instance, it might look like a genuine Teams meeting invite.
When the victim clicks the link in the fake invite, they are prompted to log in using a special code (the “device code”), which is provided by the attacker. And because the website they land on is a real Microsoft login page, the user doesn’t suspect anything phishy.
What makes this technique especially dangerous is that it exploits legitimate authentication systems without creating counterfeit ones. This removes the need for attackers to steal passwords. Instead, they can gain access by capturing session tokens which allow them to operate without triggering additional authentication prompts. And because the tokens are already verified, attackers can often bypass MFA.
At first glance, nothing seems unusual. Suspicion is reduced due to the official Microsoft website, and therefore, victims won’t hesitate to enter a device code to authenticate the session. However, instead of linking their own device, they are unknowingly authorizing the attacker's session. Once access is granted, the attacker has the keys to the kingdom and is free to operate within the victim’s account, access sensitive information, and launch lateral attacks.
How users can recognize and avoid these attacksDevice code phishing has created a minefield where legitimate tools are utilized for malicious purposes. Organizations must be proactive in recognizing these attacks and be sure to have effective authentication security measures in place.
Users should always treat unexpected meeting invites with suspicion, especially if they contain login prompts that require immediate action. Before entering any device code, users should verify the legitimacy of the request through a separate communication channel, such as a direct phone call or an internal messaging platform. If a login request appears out of the blue, it’s always best to avoid proceeding until its authenticity is confirmed.
Device codes are particularly impactful as they are designed to be entered on trusted devices. As a result, users should never share a login code with another person or enter a code they receive via email or chat unless they personally initiated the request. Legitimate services will never email a device code and then ask a user to input it on a separate website. If workforces can get to grips with this fundamental security principle, it can prevent many device code phishing attempts from succeeding.
Organizational steps to mitigate riskProtecting against these attacks can’t rely solely on the user and organizations must take steps to reduce the risk of device code phishing.
One of the most effective measures is to disable any unnecessary device code authentication flows. If it isn’t essential for business operations, then it should be removed to eliminate a significant attack vector. Security teams should regularly review authentication policies and restrict device code logins to only trusted devices.
Conditional access policies go one step further, as they can restrict authentication attempts based on user behavior, device type, geographic location, and risk level. If a login attempt occurs from an unfamiliar location or outside of approved business hours, access can be blocked or require additional verification.
This is why it’s key to embrace behavioral AI measures which can establish baseline “normal” behaviors within an organization's IT environment, and in turn question anything that seems out of the ordinary. Behavioral AI systems analyze characteristics like login patterns to detect anomalies, such as multiple authentication attempts from different locations or unusual device code submissions. By comparing these activities to known-good user behaviors, deviations from the norm can be flagged as suspicious.
And since device code phishing hinges on meeting invites to spread the attack, these should also be monitored. Security teams should regularly audit and flag unusual meeting request patterns, particularly those originating from compromised accounts.
Lastly, security awareness programs should be an ongoing feature of any cybersecurity strategy. Cyber threats evolve constantly, so training should also be continuous. Employees must be trained to recognize the warning signs of device code phishing and understand the risks of entering authentication codes without verification. Creating a culture where security is front of mind when handling unexpected requests is vital.
The time to act is nowAs this latest technique continues to prove effective, cybercriminals will no doubt expand their use of device code phishing. Organizations must act now to defend against this emerging threat. A combination of user awareness and strong security policies which are strengthened by advanced threat detection can help organizations to stay ahead.
The sooner organizations implement these measures, the sooner they can reduce their exposure to device code phishing and protect their employees, data, and systems from this growing cyber threat.
We've listed the best identity management 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
Operational technology (OT) has long struggled with modern cybersecurity demands, but operators now face an increasingly dire cyber threat from nation-state actors. OT is essential for managing cyber-physical systems across fields, including manufacturing, transport, and energy, putting it in the sights of hostile actors backed by China, Russia, Iran, and more.
Yet many OT environments are profoundly unprepared for the threat, often struggling with essential vulnerability management activity that should be a baseline to reliable security.
OT security teams should consider a shift to a policy of exposure management, a smarter approach that prioritizes the most exploitable, high-risk vulnerabilities first. Organizations relying on OT must move to reduce operational strain while closing the gaps that leave their systems open to hostile state-backed actors.
Why OT is a prime target for nation-state cyberattacksOT environments are prime targets for nation-state actors and cybercriminals attacking critical national infrastructure (CNI). Adversaries have a range of objectives, from stealing classified data and conducting corporate espionage to disrupting economic stability.
In the last few years, multiple high-profile incidents have been linked to known threat groups. For example, Volt Typhoon and Salt Typhoon are two prolific groups linked to China that have conducted several attacks on U.S. infrastructure.
Volt Typhoon has infiltrated critical infrastructure, including communications, energy, and water, and is known for using stealthy, low-and-slow tactics to exploit native tools and systems. Salt Typhoon, meanwhile, is believed to be involved in exfiltrating data from ISPs for use by Chinese intelligence operations.
Sandworm, closely linked to Russia’s military intelligence, is another long-running APT targeting critical infrastructure. The group is believed to be behind several attacks on Ukraine’s power grid over the last decade, creating the Industroyer and Industroyer 2 malware designed for industrial equipment using specific protocols. Sandworm also unleashed the notorious NotPetya ransomware.
Iran has also proven itself to be a major player in international cyberattacks. The CyberAv3ngers group has attacked U.S. water facilities using compromised PLCs and HMIs. The group has also targeted civilian infrastructure with IOCONTROL, a Linux-based backdoor designed for multiple standard OT control systems.
While high-level APTs like these have the resources and expertise for advanced tools and tactics, many OT attacks begin with unsecured devices connected to the internet, providing a clear attack path to establish footholds in critical systems.
After assessing nearly one million OT devices across 270 organizations in multiple fields, we found persistent evidence of malware in OT systems. Sample companies in manufacturing, natural resources, and logistics and transportation all had more than 10% of their OT devices communicating with malicious domains.
The problem with traditional vulnerability managementVulnerability management is a persistent issue across most sectors but can be particularly difficult when dealing with OT environments. In addition to the large and continuously increasing number of vulnerabilities to address, OT security teams must also deal with complex networks that include many disparate assets, often using their own proprietary operating systems. OT assets are seldom compatible with scanning and IT management tools designed for standard IT networks.
As a result, teams often struggle to implement the prioritized, ordered approach to vulnerabilities needed to keep ahead of hostile attacks.
Of the 270 organizations we assessed, 70% had at least one known exploitable vulnerability (KEV) in their OT systems. Twelve percent of the nearly one million devices included in the study contained a KEV that had yet to be patched. Worse, 40% of organizations have OT assets insecurely connected to the internet, creating a direct pathway for cyberattacks.
Security teams are often stuck pursuing slow and inefficient patch management programs that lack clear direction. Prioritization is usually based extensively on CVSS scores, which fail to consider the context within the company and, thus, the vulnerability's real-world exploitability and impact. More dangerous vulnerabilities may be overlooked while less important issues drain resources.
The case for exposure managementDealing with vulnerable OT assets requires a more dynamic approach, prioritized by the real risk to the organization and its infrastructure. Exposure management has emerged as one of the most effective strategies, enabling teams to identify and focus on vulnerabilities with the most significant potential for real-world exploitation.
Exposure management weighs priorities based on multiple risk factors, including identifying which KEVs are actively exploited in the wild and whether assets are affected by insecure remote access or misconfigurations that increase risk. The assessment also considers a device's criticality to business operations, for example, prioritizing those that would disrupt production or cause safety issues in the event of a breach.
The result is a drastically reduced and more focused to-do list for security teams. For example, our research found roughly 111,000 devices with KEVS. But filtering the list by vulnerabilities linked to ransomware and devices with insecure connectivity immediately reduces the total number to 3,800. Suddenly, the task has shrunk by a factor of 30, even before applying more context for specific organizations.
How to start implementing exposure management in OT securityExposure management follows a five-step process to identify, assess and resolve OT vulnerabilities.
1. Scoping
The first step is identifying those OT assets most critical to operations, such as production lines in manufacturing or scheduling control systems in maritime transport. This is especially important for asset-intensive companies with a large volume of devices to manage. The aim is to reduce the number of assets that need continuous security inspection.
2. Discovery
Next, this initial list of assets is built into a detailed inventory, focusing on the highest-risk devices. This needs to be a highly data-driven method, while more extensive and complex operations will need an automated approach to make discovery manageable.
3. Prioritization
The high-risk inventory can now be prioritized based on severity. As discussed, this process needs to move beyond basic CVSS scores to consider the actual risk posed by KEVs, the asset’s connectivity status, and the potential impact of a breach. Exploit prediction scoring and business impact assessments provide more data points to inform these decisions.
4. Validation
Before taking any action, it’s crucial to ensure vulnerabilities are exploitable and not blocked by elements like closed ports or firewalls. This avoids wasting resources on patching vulnerabilities that look severe on paper but are low risk in reality.
5. Mobilization
With all that preparation done, it’s time to get moving. It’s best to integrate exposure management into existing security workflows like patching and access control wherever possible to keep things efficient. Organizations should also look to establish cross-team collaboration between IT, security, and operations, as OT often becomes heavily siloed from standard IT practices.
Hardening OT against advanced adversariesTraditional vulnerability management is failing OT security teams by focusing on attempting to patch everything rather than addressing real threats. In the face of increasingly aggressive state-backed actors, this inefficient approach leaves critical infrastructure vulnerable to severe security incidents.
Identifying and prioritizing high-risk vulnerabilities through an exposure management approach will enable these organizations to manage vulnerabilities quickly and efficiently, drastically improving defenses against nation-state threats, ransomware, and cybercriminals.
We feature the best network monitoring 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
While federal guardrails for AI tools in the U.S. are being dismantled—even as global coalitions gather to adopt new standards—tech leaders face a critical choice: exploit the regulatory void for short-term gains, or step up to shape what responsible innovation looks like. Your move matters more than you think.
This isn't just another think piece lecturing you about "tech responsibility"—though responsible tech practices absolutely matter. Instead, it's about building resilient organizations that thrive because of ethical governance, not in spite of it. Let me show you how.
Leading in the VoidWhen regulations retreat, two things happen at once: responsibility shifts from government oversight directly to your organization, and your decisions take on amplified significance. This vacuum presents a crucial leadership opportunity. When government oversight steps back, the most innovative companies step forward, not to exploit the gaps, but to demonstrate what good governance looks like in practice.
Consider the evolution of privacy practices - companies that proactively developed comprehensive data protection frameworks before GDPR and CCPA became law had a significant advantage. Those that waited found themselves rushing through expensive, disruptive compliance programs. The same dynamic is playing out with AI governance today.
The Now/Next ImperativeSuccess in this environment requires mastering what I call the Now/Next Continuum—a strategic framework that helps organizations navigate immediate pressures while building toward better futures. It's not about artificially balancing short-term versus long-term thinking—it's about recognizing the natural throughline between them. Today's decisions actively shape tomorrow's possibilities.
This means asking questions like: how will this AI deployment decision affect our ability to adapt to future regulatory changes? What precedent are we setting for our industry? How does this choice align with our vision for technology's role in society?
Building Future-Ready GovernanceYour governance framework shouldn't depend on regulatory stability to function. Instead, build systems that:
The most successful organizations aren't waiting for regulatory clarity—they're creating it. They understand that ethical governance isn't a burden—it's a catalyst that drives deeper, more meaningful innovation.
The Leadership ImperativeIn an era where tech leaders occupy the highest echelons of policy making and can reshape federal agency positions with a single decision—and where AI continuously finds novel pathways to familiar harms—the argument that "existing rules are sufficient" simply doesn't hold water. Responsibility means far more than mere compliance. Even smaller companies must ask harder questions. We must move beyond "Can we?" to "Should we?" Beyond "Is it legal?" to "Is it right?"
This shift requires developing new muscles: strategic foresight, ethical reasoning, and the ability to balance competing interests while maintaining a clear moral compass. It means building teams that understand both the technical and human implications of AI deployment.
Human-Centered StabilityBuilding resilient policies in volatile times isn't about predicting every possible regulatory shift. It's about anchoring your governance in something more fundamental: human thriving. This might sound abstract, but when you consistently prioritize human outcomes over regulatory minimums, your policies become naturally resilient to political shifts.
What does this look like in practice? It means designing AI systems that enhance rather than replace human capability, implementing robust testing frameworks that assess societal impact, and creating clear escalation paths for ethical concerns.
The Ethics AdvantageWhich all adds up to the plot twist: companies that treat ethics as their operational foundation rather than their compliance checklist aren't just doing good—they do better. They're more innovative, more trusted, and more resilient to market shifts.
These organizations understand that ethical AI isn't about restriction—it's about direction. It's about channeling innovation toward outcomes that create sustainable value for both the business and society. They recognize that the strongest competitive moats aren't built with technology alone, but with technology guided by strong ethical principles.
The future belongs to organizations that understand this fundamental truth: In an era of constant change, where AI advances faster than our ability to regulate it, ethical governance isn't just a responsibility—it's a competitive advantage. The question isn't whether to lead on governance, but how quickly you'll step up to do it.
Your organization's approach to AI governance in this regulatory void won't just determine your short-term success—it will define your legacy in shaping the future of human-centered technology. Choose wisely.
We've featured the best todo list app.
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