Feed aggregator

Gisèle Pélicot detailed the horror of discovering that her former spouse had systematically filmed the dozens of suspected rapes — storing thousands of images that police investigators later found.

(Image credit: Lewis Joly)

A VPN can help you maximize your NFL streaming experience. Here’s how.

US intelligence services have charged a group of five Russian hackers linked to Unit 29155 of Russia’s Main Directorate of the General Staff of the Armed Forces (GRU), as well as one civilian hacker working for the Russian Government. This puts the Russian military intelligence agency at the center of critical infrastructure attacks on NATO member states.

The five GRU members have been charged in connection with a series of attacks known as ‘Whisper Gate’ which were carried out in January of 2022. The operations targeted Ukraine’s government agencies and infrastructure, and are largely understood as a precursor to the Russian invasion of Ukraine.

The US State Department has offered a $10 million bounty for information on each of the six suspects through its Rewards for Justice program. The GRU is also reported to have targeted the computer systems of 26 other NATO countries which were providing support to Ukraine, a strategy we have seen deployed many times before.

First shot of the war

The attacks were countered by US cyber command and private cyber security firms like Microsoft and Google, and so were unsuccessful in bringing down satellite communications or crippling Ukrainian government operations as intended.

US officials said that the GRU members posed as criminal hackers by leaving ransom notes, but the attacks were designed to destroy any data acquired. According to the joint advisory, “Unit 29155 cyber actors' objectives appear to include the collection of information for espionage purposes, reputational harm caused by the theft and leakage of sensitive information, and systematic sabotage caused by the destruction of data”

Since none of the hackers are on US soil, it's unlikely they will face the consequences of the charges anytime soon, but the US have reportedly partnered with Interpol to “bring this indictment to fruition”.

Via BleepingComputer

More from TechRadar Pro

• Take a look at our pick of the best firewall software
• Notorious Russian hackers target government officials
• Check out our pick of the best malware removal software around

Apple is pretty good at, well, not exactly keeping secrets, but at least it has a spotless record for never confirming a rumor before it's real. That means we know almost everything the Cupertino tech giant plans to unveil at its Glowtime Event on Monday. New iPhone 16 phones, fresh Apple Watches, AirPods, and Apple Intelligence.

The leaks have been so steady and, I'm assuming, accurate, that Monday's Keynote by Apple CEO Tim Cook and company is nothing more than a confirmation event, it's the graduation from rumor to reality that probably required little more than a flip of a tassel from one cap corner to another.

(Image credit: Future / Lance Ulanoff)

I'll be there, hopefully as close to the front row as possible, and jostling among analysts, fellow journalists, and others in the demo room, angling for our first hands-on experience with the new gadgets. There'll be excitement and breathless commentary. Some of it will be authentic, and some of it might be a bit forced because the upgrades are minor. We'll struggle to capture what might be a set of half-measure Apple Intelligence features (while we wait for the rest to show up later and even next year). But the empty seat at the keynote and space in the demo room will be for that one more thing that seldom arrives.

Invented by Apple's Steve Jobs, "One More Thing" is the surprise feature or product that, one would hope, nobody saw coming. FaceTime has shined in this light, as did the MacBook Air, Apple Watch, Apple Music, and, of course, Apple Vision Pro. Granted, most of these products were not a complete surprise. Leaks and rumors about the work Apple is doing in any tech sector or vertical are inevitable but at least details about these items tend to be far fuzzier before they come into focus when Tim Cook utters that iconic phrase.

With Glowtime, though, there is not even a hint of something special in Apple's back pocket. The only promise is predictability and, if we're being honest, playing catchup.

Post by @lanceulanoff View on Threads

The star of this Apple Event will not be the iPhone 16, iOS 18, a plastic Apple Watch SE or some new AirPods. It will be Apple Intelligence, Apple's rather late entry in the AI game. I am excited about it and can already see the potential on the iOS 18 Dev beta. 

When I polled people on Threads and X about the Apple Event and asked which products and technologies they were most excited about, a surprising number put Apple Intelligence at the top of the list. On Threads, it handily beat iPhone 16 Pro Max, a cheaper Apple Watch, and AirPods by a wide margin. It came in second on X.

Of course, we know pretty much all there is to know about Apple Intelligence; Apple detailed it at WWDC 2024. This won't be a true product introduction, just a partial delivery on a promise and one that is well behind what we're already experiencing from Google's Gemini AI on Google Pixel 9 phones and Galaxy AI on Samsung Galaxy Z Fold 6.

Monday's Keynote...is nothing more than a confirmation event, it's the graduation from rumor to reality that probably required little more than a flip of a tassel from one cap corner to another.

The iPhone 16 lineup appears unlikely to bring any surprises. Faster chips and subtle design changes are expected. Even the rumored camera button (which I do not think will materialize) won't be a shock.

I don't think it matters if the one more thing is a breakout success. Apple's track record is quite good in this regard. It's about reminding people that Apple can surprise and innovate in ways we didn't anticipate and in ways that make competitors blanche.

I could be wrong. Tim Cook and other Apple execs could finish up what I expect to be an hour-long presentation and Cook will turn to leave the stage. Then he might pause, turn back to the audience, and bring his hands together in his signature "prayer" pose, raise one hand, and say, "Hold on... there is One...More...Thing."

If he does that, I and probably everyone watching in person and online will be truly surprised for the first time in a long time. I don't care what comes next, I just know that I wasn't expecting it and that is enough.

You might also like

• Apple Event: get ready for 'Glowtime'
• Apple's iPhone 16 launch event – 5 things we expect and 3 things ...
• Apple just sent out invites for the iPhone 16 event – 'It's Glowtime.' is ...
• How to watch the iPhone 16 launch event live
• Apple builds hype for next week's iPhone event with mega-deals on ...
• Apple hints at smart ring plans with its unsubtle Apple Watch survey ...

Full spoilers follow for Arcane's first season. Potential spoilers follow for Arcane season 2, too.

Arcane season 2 is only two months away from gracing our screens and, after a near-three-year wait for its return, we'll finally get answers to the biggest questions we've had since the season 1 finale.

But, while I'm as desperate as you likely are to find out if any of Piltover's Council members survived Jinx's terrorist attack on their chambers in the series' second and final installment, I'm also eager to see if a new fan theory has any truth to it. You see, Arcane's loyal fanbase – I include myself among their number – have waited a looooong time for the hit Netflix show to return. it's understandable, then, that many of my fellow devotees have come up with all sorts of weird, wacky, and wonderful theories since season 1 ended in November 2021.

Think Jinx is going to be season 2's primary big bad? Think again (Image credit: Netflix/Riot Games/Fortiche)

The arrival of Arcane season 2's official trailer, however, has only heightened the intrigue around one particular hypothesis surrounding the animated League of Legends (LoL) show's actual main villain. Indeed, given her assault on Piltover's dignitaries in season 1 episode 9, the footage that was part of Arcane season 2's first teaser, and a bundle of first-look clips over the past 10 months, it would be safe to assume that Jinx is the central antagonist.

But what, I ask, if she isn't? What if there's not only someone else pulling the strings behind the scenes, but what if they're also a character we haven't seen before? I suspect you're equal parts intrigued over who this individual could be, and frustrated over me delaying the reveal of their identity. So, I won't waste any more of your time: fans think that the mastermind of events before and during Arcane's two seasons is someone named LeBlanc.

Who is LeBlanc? And what role could they play in Arcane season 2?

Say hello to LeBlanc, everyone (Image credit: League of Legends universe website)

I can see your puzzled look through the screen, dear reader. Let me elaborate: in LoL lore, LeBlanc, who also goes by such jovial aliases like The Deceiver, Mordekasier, and Matron of the Black Rose, is a mysterious but powerful sorceress who's manipulated people and events since the expansionist nation of Noxus was founded. Thanks to her dark magic abilities, she's able to shapeshift, can appear in multiple places at once, and can control people from the shadows.

Okay, so why do Arcane fans think that LeBlanc is the big bad of one of the best Netflix shows? On the surface, it seems like a bit of a stretch, especially when there are other obvious candidates. There's compelling evidence, though, to suggest LeBlanc is eventually going to be unmasked as the award-winning series' villain-in-chief.

Ambessa, Noxus' most famous warlord, could have ties to LeBlanc in Arcane (Image credit: Netflix)

Let's start with her ties to Noxus. Arcane and LoL diehards won't need me to tell them that LeBlanc's association with Noxus means she could be Ambessa's puppet master. Remember, Ambessa is not only a Noxian warlord, but also Piltover council member Mel's mother. If Mel died during Jinx's attack, a grief-stricken Ambessa will be far more vulnerable to the machinations of a powerful sorceress who wants to further sow division between the progressive, idealistic city of Piltover and its poorer, undercity cousin known as Zaun, which is where Jinx is originally from.

Next, there's Hextech, the potent, magic-infused technology created by Viktor and Jayce in the critically-acclaimed show's debut season. Why wouldn't a formidable dark magic wielder be interested in this newfound technology, especially given its dynamic qualities and potential to supercharge whoever commands it? 

We've already seen its effects on creating powerful new weapons, fast-paced methods of transport, and augmenting people's bodies in season 1. In the hands of a truly malicious individual, then, it could be used by said person for nefarious means. And, considering Arcane co-creator Christian Linke has already teased the possibility that "any foreign nation", Noxus included, "would find [Hextech] very interesting" (see the X/Twitter post below), I suspect Ambessa and LeBlanc will want to get their hands on it.

#Arcane co-creator @Praeco answers fan questions about what's coming in Season 2. pic.twitter.com/AaSqLC3Uy1August 10, 2024

Then there's the possibility that LeBlanc can be seen in season 2's latest trailer. Seen amid the quick shots that flash up around the 1:54 mark, an inky, jet-black character who looks like a lot like Silco (Jinx's deceased surrogate dad), can be briefly glimpsed. The identity of this character isn't confirmed, but I and many others (as you'll see in this Arcane Reddit thread) believe it could be LeBlanc disguising herself as an even more terrifying vision of Silco.

Why do we think that? For one, this individual's all-black look ties into the 'Black Rose' aspect of one of LeBlanc's alter-egos. Look more closely, however, and the yellow, tear-like marks that sit beneath this character's glowing eyes are identical to the black markings that appear underneath LeBlanc's eyes (see the image further up this page for proof of that). Oh, and let's not overlook that, on LeBlanc's official LoL character profile page, the show's title is even referenced in a line that includes the word "arcane" in it. Oh, and the same page also mentions LeBlanc's involvement in LoL's infamous Rune Wars, which Linke also refers to in the tweet above.

Admittedly, those final two pieces of evidence are tentative links at best, but there's enough here to indicate that LeBlanc could play a role in Arcane before season 2 brings the curtain down on Netflix's most successful animated TV project. Here's hoping many of us weren't completely off-track when Arcane returns sometime in November.

You might also like

• Netflix 'aggressively taking action' after huge data breach sees full episodes of Arcane season 2, Heartstopper season 3, and more leak online
• Netflix releases new suspense-filled, Jinx-starring clip for Arcane season 2 as fallout from full-episode leak continues
• New Netflix teaser for Arcane season 2 shows Heimerdinger and Ekko summon their inner Batman for an amusing stealth-based mission

AI translation is fast becoming a business that's essential, turning global interactions local.

And at IFA 2024, it goes well beyond translation software. There are no lack of gadgets, apps, and tools designed to ease multi-lingual conversations. The Honor Magic V3, for example, whose foldable design means you and others can all see translations in real-time.

One new release we got at Europe's biggest tech event is the Timekettle W4 Pro earbuds. We've reviewed plenty of Timekettle translators in our time, and they've usually left us very impressed. Language is no longer the barrier it used to be, and so I was keen to lay my hands on the company's new translation device.

Making global interactions local

Looking like a pair of early-2000s Bluetooth earpieces designed for business-business, both units have a soft, rubbery feel where it rests on the ear, with a hardened plastic mouthpiece that doesn't interfere with speech. Each piece sat in its own moulded box where, like any other earbuds, the device will charge.

Translations are made and displayed via what's effectively a speech-to-text app. While we didn't get any hands-on time with the app, we observed the process, and it's exactly as you'd expect: select the original language and the output language.

Image 1 of 5

(Image credit: Timekettle)Image 2 of 5

(Image credit: Timekettle)Image 3 of 5

(Image credit: Timekettle)Image 4 of 5

(Image credit: Timekettle)Image 5 of 5

(Image credit: Timekettle)

In our case, this was Chinese to English. We popped on the headset, our host did the same, and we conducted a light conversation. I made no attempt to moderate my speech or speak loudly. I simply spoke in my normal register. But even in the busy event hall, the Timekettle device easily picked up my every word, and that of my host, displaying them on the app.

It took around 2.5 seconds while the app processed the speech before the translation appeared on the screen. If you've used Google Translate, the effect is more or less the same.

And in my ear. A soft human-sounding voice audibly translating my host's fluent Chinese into English. Not too loud to overpower the conversation, but easily understandable even as others spoke around me.

Quite by accident, I also experienced Chinese to German translation. The initial set-up was for this, and that also worked very well (with what little German I understand).

An effective bit of kit, then, for businesses working with global partners. I'd like to see the translations coming faster, to allow for a more natural flow. However, in practice, I found this an impressive experience for tearing down linguistic barriers.

Read more from TechRadar Pro

• We tested the best free text-to-speech software
• AI could boost UK job market by 610,000
• AI hype is real in manufacturing – successful adoption journeys are scarce

Fujifilm Instax mini Link 3: two-minute review

One could spend hours pondering why people might want to make small instant prints out of their smartphone snaps – photos they could easily share with each other via WhatsApp or post on Instagram. Personally, I think it’s all about the fun, tactility and immediacy that can be derived from occasionally ‘going analogue’ in this increasingly digital world in which we live. Vive la difference, and all that.

 Fujifilm’s Instax mini Link 3 is the latest in the company’s line of portable battery-powered printers which, after linking up with a smartphone or tablet via Bluetooth, can spit out fast-developing instant photos. These photos, which are roughly the size of a credit card, can be given to friends and family, pinned on your corkboard or stuck on your fridge door or slipped in your wallet to look at whenever you like. It’s a simple device that does what it’s supposed to and does it well.

Image 1 of 4

(Image credit: Sam Kieldsen | Future)Image 2 of 4

(Image credit: Sam Kieldsen | Future)Image 3 of 4

(Image credit: Sam Kieldsen | Future)Image 4 of 4

(Image credit: Sam Kieldsen | Future) Fujifilm Instax mini Link 3 key specs

Film type: Instax mini film
Print size: 62mm x 46mm
Print time: 15 sec. approx..
Developing time: 90 sec. approx.
Connectivity: USB-C, Bluetooth
Dimensions: 90 x 37.3 x 125mm
Weight: 210g

I’ll talk about the prints themselves first. This won’t be news to anyone who’s used Instax mini film before, but these Polaroid-style photos look really impressive, with beautifully rich and vivid colours and an attractive glossy finish. That’s once they develop, of course: you’re looking at about three minutes in total from the time you starting printing until the finished photo is fully developed. Detail is pretty sharp, but the small print size puts a limit on that. The prints are hardy too: don’t expect them to fade quickly or fall to bits in your wallet.

They are also fairly expensive by portable printer standards however, with a standard 20-pack costing around $20/£15, which means about $1 or 75p a per print. Some rival printers using Zink or dye-sublimation methods offer a smaller per-print cost. I would say the Instax mini photos are better looking and more durable than Zink prints at least.

Part of the selling point with the mini Link 3 is the added functionality that comes from the mini Link companion app, and to be clear the printer is essentially useless without a smartphone to pair it with. The app offers light editing tools and a range of frames, effects and other adornments with which to customise and personalise your photos. These are all a bit cheesy and basic (I suspect few graphic design aficionados will be impressed) but if the point is to have fun, they hit the spot.

You can import photos from your camera roll or take new shots from within the app. In fact, the app’s Instax camera mode turns the printer itself into a Bluetooth remote control, allowing you to set your phone up and take a shot from afar by pressing the printer’s function button.

The app also features an augmented reality (AR) mode that lets you take photos on your phone complete with added-in objects, but to be honest I quickly gave up trying to use it due to the confusing setup and underwhelming effects. With the rest of the app functions being straightforward and easy to immediately grasp, this one feels out of place; I doubt it’ll prove popular with users.

Image 1 of 4

(Image credit: Sam Kieldsen | Future)Image 2 of 4

(Image credit: Sam Kieldsen | Future)Image 3 of 4

(Image credit: Sam Kieldsen | Future)Image 4 of 4

(Image credit: Sam Kieldsen | Future)

The mini Link 3 printer itself is an attractive object with its curvy lines, ridged surfaces and three-LED lighting system, which pulses and glows different colours depending on what you’re doing; it usually functions as a battery life indicator. There are three colour finishes available, too: Sage Green, Rose Pink and the Clay White model I was sent to review.

The mini Link 3 is lightweight and compact enough to slip into a large jacket pocket. I wouldn’t have any qualms about tossing it into a backpack either. The plastic body is sturdy and there are few buttons or moving parts that could get damaged: just one button for power, one function button that fulfils a number of functions (such as initiating Bluetooth pairing), a small sliding button to open the rear door when you need to load in a new film cartridge and a tiny flap that opens up to reveal the USB-C port used for recharging the internal lithium ion battery.

Fujifilm Instax mini Link 3 price and availability

The Instax mini Link 3 is available now, priced at £114.99 in the UK (where it’s available at a range of online stores including Amazon, Jessops and My Fujifilm). In the US it’s cheaper at $99.95 and available from the Instax site.  

Battery life is excellent. Fujifilm claims you can print around 100 shots from a full charge and while I haven’t personally tested that, being sent only three 10-shot cartridges of Instax mini film with my review sample, I didn’t need reach for the included USB-A to USB-C recharging cable once during my week with the printer (note that there’s no AC adapter in the box). If you do deplete the battery, you can fully recharge it in about two hours.

Image 1 of 4

(Image credit: Sam Kieldsen | Future)Image 2 of 4

(Image credit: Sam Kieldsen | Future)Image 3 of 4

(Image credit: Sam Kieldsen | Future)Image 4 of 4

(Image credit: Sam Kieldsen | Future) Fujifilm Instax mini Link 3: also consider

Fujifilm Instax mini Link 2
The predecessor to the mini Link 3, this model’s advancing age and recent replacement means you may be able to pick it up at a significant discount. It uses the same Instax mini film and works with the same mobile app, so functionality and image quality is very similar.

Canon Selphy Square QX10
This dye-sub portable printer produces 2.7in square photos, with each print taking about 40 seconds to emerge and, according to Canon, are durable enough to last for 100 years. The prints have a peel-and-stick backing means you can mount them quickly too.

Should I buy the Fujifilm Instax mini Link 3?

(Image credit: Sam Kieldsen | Future) Buy it if...

You want fast, vibrant prints
Instax mini prints develop in a matter of seconds and boast vivid, bold colours. They’re robust too, and won’t fade or rub off easily.

Don't buy it if... How I tested the Fujifilm Instax mini Link 3

• One week of use
• Tested with iPhone 13

I spent a week using the mini Link 3 in conjunction with my iPhone 13, testing each mode of the mini Link companion app with both freshly taken photos and older images from the smartphone’s camera roll. I also took the printer out ‘in the field’ with me, taking advantage of its battery and compact size to use it on a night out with friends.

First reviewed September 2024

Honor has revealed a new Snapdragon-powered business laptop is on the way. 

After releasing the Intel version of the Honor MagicBook Art 14 in China, and officially launching in Europe yesterday at IFA 2024, the mobile company confirmed what internet whispers had suspected: a version of Honor MagicBook Art 14 is coming with a Snapdragon X Elite processor. 

The company has yet to set a release date or price for the slim-line laptop - but we’ll be watching this one like a hawk. 

What’s new from Honor?

Honor has been busy at this year’s top tech event, showcasing the Honor Magic V3 (read our full review right here), alongside the MagicPad 2, and the MagicBook Art 14 for a European audience. 

Now, just twenty-fours later, the company has confirmed it’s squeezing Qualcomm’s mobile processor into its latest laptop, and creating an entire ecosystem of devices and apps that work seamlessly together. With the Snapdragon processor, Honor expects the device to offer Wi-Fi 7 support, improved power efficiency, and better performance. 

From what we’ve seen here in Berlin, there doesn’t appear to be any other significant changes to the laptop. So, expect the same specs found in the Intel version running Windows 11. That includes 1TB storage, a choice of 16GB or 32GB RAM, and a sharp 14.6in OLED screen. Better still, in contrast to modern Apple devices, the device features a host of ports that offer more flexibility on the go.

Image 1 of 11

(Image credit: Honor / Future)Image 2 of 11

(Image credit: Honor / Future)Image 3 of 11

(Image credit: Honor / Future)Image 4 of 11

(Image credit: Honor / Future)Image 5 of 11

(Image credit: Honor / Future)Image 6 of 11

(Image credit: Honor / Future)Image 7 of 11

(Image credit: Honor / Future)Image 8 of 11

(Image credit: Honor / Future)Image 9 of 11

(Image credit: Honor / Future)Image 10 of 11

(Image credit: Honor / Future)Image 11 of 11

(Image credit: Honor / Future)

We had the chance to try out the MagicBook Art 14 powered by the Intel Core Ultra 7 Processor 155H at this year’s IFA. And while we weren't able to push it to the limit, try all the AI tools, and open a thousand tabs in Chrome, what we did experience was a quick and efficient laptop with a pleasing design and build. Expect a real performance boost if it’s anything like the last MagicBook 14 we reviewed (find out more here).

Built from magnesium, with a keyboard constructed from titanium, helps ensure the machine stays incredibly lightweight - weighing just over 1kg and measuring 11.5mm at its thickest point, this is about as portable as it gets. And yes, it’s slightly lighter than a MacBook Air, which will cheer the company’s social media marketing team, who seem to exist to troll its po-faced rivals. 

There’s a lot to praise about the screen, too. At 14.6in, it’s a good size for business users who need a broad touchscreen display without being hefty to carry around, boasting 700nits luminance, and 100% DCI-P3, making it ideal for watching movies, although we’d need to test it properly to see how it stacks up in the field of best video editing laptops. However, we did find the screen a little too reflective in an admittedly well-lit exhibition hall, despite an abundance of eye-care technology built-in. 

The ultrabook is also impressively soft-to-the-touch. It might not be, as Honor CEO George Zhao claimed, “as soft as a baby's skin”, but it’s not far off, feeling very comfortable in the hand, with a near-matte feel. 

Perhaps the real eye-catcher, though, is the detachable magnetic camera, which locks into place and feels sturdy once in place. It's a really rather novel way of dealing with built-in webcams that traditionally use sliders, covers, or just a spot of Blu-Tack over the lens. It also has the benefit of being reversible, if you need to show others on a video call what you're looking at. While we haven't had confirmation yet, expect Honor to start selling these separately for when the originals inevitably tumble under the desk never to be seen again. 

Read more from TechRadar Pro

• We reviewed the best business PCs
• Asus ExpertBook P5 AI laptop boasts military-grade durability, and enhanced performance
• Microsoft unveils a plethora of next-generation Surface laptops with support for 5G and CoPilot+

Husam Abukhedeir, the chief neurosurgeon at Al-Shifa Hospital, helped the injured, watched many died, including his sister, then knew what he had to do to protect his family. How is he faring today?

(Image credit: Christopher Pike for NPR)

When researchers at software management company, JFrog, routinely scanned AI/ML models uploaded to Hugging Face earlier this year, the discovery of a hundred malicious models put the spotlight on an underrated category of cybersecurity woes: data poisoning and manipulation.

The problem with data poisoning, which targets the training data used to build Artificial Intelligence(AI)/Machine Learning(ML) models, is that it's unorthodox as far as cyberattacks go, and in some cases, can be impossible to detect or stop. Attacking AI this way is relatively easy and no hacking in the traditional sense is even required to poison or manipulate training data that popular large language models (LLMs) like ChatGPT rely on.

Data poisoning can be used to make the AI model do your bidding. Or AI models can be convinced to give erroneous output by modifying the data sent into a trained model. These are two different types of attack–one that is done before the AI model deploys, the other done post-deployment. Both are incredibly difficult to ferret out and guard against.

In its analysis, JFrog noted the “intriguing“ payload embedded within the model looked like something researchers would upload to demonstrate vulnerabilities or showcase proofs-of-concept. That was not the case with the nefarious models uploaded to Hugging Face’s AI collaboration repository. Researchers may have been behind it because the payloads had links to IP addresses from KREOnet, or the Korea Research Environment Open Network.

Built-in AI problems exacerbate detection while fertilizing exploits

Examples of training data manipulation can be traced to the origins of machine learning, with researchers demonstrating subtle adversarial attacks on input results in a model outputting an incorrect answer with high confidence a decade ago.

It’s even possible that generative AI models scrapping the internet could eventually “poison” themselves as their outputs become inputs for future training sets, in a process known as “degenerative model collapse.”

What muddles the waters further is that AI model reproducibility is in itself a challenge as there are vast pools of data used to train models, and researchers and data scientists may not even understand exactly what went into a model and what is coming out, exacerbating the detection and traceability of malicious code.

Inconvenient as all of this sounds in the AI gold rush, turning a blind eye to data poisoning and data manipulation can embolden attackers to focus on stealth backdoor exploits of AI software. The results can be malicious code execution, as in the case of Hugging Face, new vectors to successfully carry out phishing attacks, and misclassified model outputs that lead to unexpected behaviors, depending on the goals of the attacker.

In a world increasingly blanketed with an ecosystem of interconnected AI, GenAI, LLMs, and APIs, the global cybersecurity industry should release a collective shudder and take action to protect against the rise of attacks on AI models.

Protecting against the “indefensible”

Experts advise several techniques to protect AI-driven systems from data poisoning or manipulation campaigns. Most focus on the data training stage and the algorithms themselves.

In its “Top 10 for LLM Applications” list, the Open Source Foundation for Application Security (OWASP) recommends steps to prevent training data poisoning, starting with paying attention to the supply chain of internally and externally sourced training data, with continuous verification of data sources across pre-training, fine-tuning, and embedding stages and flagging of any biases or anomalies.

OWASP also recommends “sanitizing” the data with statistical outlier and anomaly detection methods to hunt down any adversarial data from potentially being fed into the fine-tuning process.

If training data is corrupted, alternate AI algorithms can be used to deploy the impacted model. More than one algorithm can be used to compare results, and fallback to pre-defined or averaged outputs when all else fails. Developers should closely examine AI/ML algorithms that interact or feed into others, as it can lead to a cascade of unexpected predictions.

Industry experts also suggest that cybersecurity teams check the robustness and resilience of their AI systems by pentesting and simulating a data poisoning attack.

A 100% cybersecure AI model can be built and poisoned using training data. There is no defense other than validating all the predictive output, which is very expensive computationally.

Building a resilient future for AI

Without trust and dependability, the greatest innovation in tech may hit the brakes.

Organizations need to prevent backdoor threats in AI code generation by treating the entire ecosystem and supply chains that underpin GenAI, LLMs, etc. as part of the overall threat universe.

By monitoring the inputs and outputs of these systems and detecting anomalies with threat intelligence, findings and data from these efforts can help developers promote and use controls and protections in the AI software development lifecycle.

Overall, examining the risks of AI systems within the broader business processes, including checking the entire data governance lifecycle, and monitoring how AI behaves in specific applications, you can stay one step ahead of one of the most challenging issues facing cybersecurity.

We've featured the best database 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

Hisense has unveiled its own 360 surround speaker system - named the HT Saturn - as an alternative to the Sony Bravia Theater Quad (as reported by FlatpanelsHD). 

The system, revealed at IFA 2024, consists of four wireless speakers, a subwoofer and a control box - similar to the Sony Bravia Theater Quad, although the Quad doesn't come with a subwoofer in the box (it can be paired with the SA-SW3 and SA-SW5 wireless subwoofers).  

Hisense claims 720W of power, which far exceeds the Quad's claimed 504W. The HT Saturn will support both Dolby Atmos and DTS:X audio formats. 

If the HT Saturn follows the same processes as the Theater Quad, the control box will be where connections are made (such as HDMI for your TV) and it will then wirelessly transmit the audio signal to the four speaker units and subwoofer. 

This system will be an expansion on Hisense's range of soundbars, which includes the Hisense AX5125H - one of the best soundbars available - which is currently the closest you can get to an immersive surround sound experience in Hisense's soundbar lineup thanks to its four separate units, consisting of a soundbar, subwoofer and two satellite speakers. 

There are currently no further specs, pricing or release date information available, but we'll be sure to keep an eye on it. 

Price vs performance

We were impressed by the Sony Bravia Theater Quad, but there's no getting away from its high price point. (Image credit: Future)

The HT Saturn's main objective will to provide a noticeable upgrade over the fantastic Hisense AX5125H, which we complimented in our 4.5 out of 5 star review for delivering a powerful, dynamic and 'real' Dolby Atmos surround sound at such a budget price (the AX5125H costs roughly $299/£299/AU$499).

The HT Saturn will probably cost significantly more than this, but given Hisense's usual pricing, it's likely to be extremely competitive with the Sony Bravia Theater Quad, which is $2,499/£2,499/AU$3,699. That in itself is a sizeable price hike over its flagship soundbar, the Sony Bravia Theater 9, which currently retails for $1,199/£1,199/AU$1,799. 

The prospect of a budget alternative to the Quad is exciting however, and when we heard the Quad in action, we were suitably impressed with its surround sound prowess. We didn't feel it beat the Bravia Theater 9 by enough to justify the significant price jump, but when we paired it with the Sony Bravia 9 (Sony's 2024 flagship TV) and used it as the center speaker in the Quad setup, the Quad took on another level. 

It's unclear whether Hisense's HT Saturn will do something similar with its own TVs, such as the Hisense U8N, one of the best TVs released in 2024 so far, but if it can do this at a more affordable price then sign us up. We've got high hopes after seeing what Hisense could do with the AX5125H. 

You might also like

• The best Dolby Atmos soundbars
• The best wireless speakers
• Hisense announces new OLED TV with the same panel as the LG C4 for way cheaper – but it’s not launching everywhere yet

Understanding the cybercrime ecosystem has never been more important than in 2024. Ransomware attacks and data breaches continue to proliferate, and it’s imperative that executives understand what’s driving the increase and more importantly how to prevent oneself from becoming a victim.

The nature of threat intelligence is changing

Cyber threat intelligence has historically focused on identifying major cybercrime groups and state actors (dubbed “Advanced Persistent Threats” or “APTs”). However the vast majority of breaches are not caused by these highly sophisticated actors.

Instead, organizations are breached by relatively low-sophistication actors that often operate on social media sites or dark web forums. For this reason, cyber threat intelligence has gotten a bad reputation for “low actionability.” If you concentrate all of your efforts on actors and tactics that only perpetrate a tiny proportion of attacks, it isn’t going to be all that useful to the average company.

It’s not that cyber threat intelligence can’t be useful. However, it has been focused on the wrong things. Take for example, the often underlooked trend that drives thousands of major breaches annually: breached identities compromised through infostealer malware. Steering more attention toward these less sophisticated actors will make a huge impact.

Breached identities and the value of intelligence

Right now, more than 70 million breached identities are being traded in the cybercrime ecosystem. This is largely due to infostealer malware, which steals all the credentials, session cookies, browser history, and even crypto wallet data from a single use.

Breached identities went from a minor nuisance in the 2010s, to one of the leading causes of data breaches in 2024. Yet many organizations are still not aware, and not properly responding to this threat. It is much more likely that an employee accidentally downloads infostealer malware onto their computer than an APT attempting a mass cyber attack on your organization.

With a single misclick or bad download, cybercriminals can sell incredibly sensitive data from a user for just a few dollars. That’s one of the reasons that the 2024 IBM-Xforce report states that “in this era, the focus has shifted to logging in rather than hacking in”.

Imagine you have an employee who falls victim to this malware, and dozens of credential pairs to third-party applications, internal systems, and networks, are leaked onto the dark web. Ransomware groups and threat actors will then purchase these credential sets and use them to stage an attack.

By our estimation, around 10% of breached identities contain credentials to corporate IT systems, amounting to millions of corporate credentials currently in the hands of cybercriminals. Back in 2023, Flare conducted a study that showed hundreds of thousands of breached identities with corporate access, including credentials to corporate VPNs, Single Sign-On environments, CRMs and accounting software have been sold by criminals in the past two years.

The future of continuous threat exposure management

Breached identities are just one threat vector that exposes thousands of companies to breaches. Legacy CTI providers are failing to provide adequate depth to their findings. That’s why the future of cybersecurity intelligence is using Continuous Threat Exposure Management (CTEM). CTEM products focus on two major priorities:

Identifying High-Risk Exposure: The first priority is to identify specific high-risk data exposure that leaves an organization at risk of suffering a major incident, ransomware attack, or data breach. Exposure could be anything from a breached identity to exposed data in a misconfigured S3 bucket.

Identifying Key Trends that Enable Strategic Decision Making: An effective CTEM solution should also present the latest trends in cyberattacks and data breaches in a way that is both actionable and enables effective strategic decision making.

By leveling up from cyber threat intelligence to CTEM, organizations can substantially reduce their risk of data breaches and ransomware attacks while focusing their security spend on intelligence that drives actionability.

We've featured the best encryption software.

This article was produced as part of TechRadarPro's Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here:  https://www.techradar.com/news/submit-your-story-to-techradar-pro

Ali Truwit's lower leg was amputated after a shark bit her during a post-graduation trip last May. This week, she won Paralympic silver and set a new American record in the women's 400 meter free S10.

(Image credit: Julia Nikhinson)

Who's ready for another month full of new movies and TV shows to stream? The answer, of course, is you – and you better be a Netflix subscriber this week, otherwise your streaming offerings will be comparatively piecemeal. For one, the usually reliable Prime Video doesn't have anything new to offer this weekend, so it's up to other top streaming services to fill the void.

And fill the void they have. Whether it's the return of critically-acclaimed Apple TV Plus and Hulu shows, or the arrival of Studio Ghibli's latest masterpiece on Max, there's still plenty to wrap your eyeballs around. Here, then, is what's worth watching on the world's best streaming services this weekend.

Rebel Ridge (Netflix)

When I first saw the official trailer for Netflix's Rebel Ridge, I couldn’t help but think it sounded like a modern take on the classic action flick Rambo: First Blood, which spawned an entire franchise built around its PTSD-laced lead character. Once you learn one of September's new Netflix movies' plot is about an ex-Marine passing through a small town who ends up having a run-in with corrupt local police that escalates with every passing minute, you'll understand why it sounds so familiar, too.

Rebel Ridge is Jeremy Saulnier’s first new project since he directed two of True Detective's first season. And, after reading an interview with him on Netflix Tudum, I can understand why he wrote the screenplay given that he’s a fan of "80s and ’90s action films that not only deliver on spectacle, but succeed in tying on-screen mayhem to a real and true emotional component". At the time of publication, Rebel Ridge sounds like one of the best Netflix movies of 2024, with critics giving it a stupendous 98% score. To say it seems like a promising watch is something of an understatment!

Amelia Schwanke, senior entertainment editor 

Slow Horses season 4 (Apple TV Plus)

It boggles my mind that Slow Horses hasn't enjoyed more mainstream success over the past few years. It's easily one of the best Apple TV Plus shows around, yet it continues to pass many people by.

Well, no more, I say! With its perfect 100% critical rating on Rotten Tomatoes, Slow Horses season 4 is the acclaimed spy thriller series' best entry yet. Simply put, you don't have a valid reason not to check out this Gary Oldman, Jack Lowden, and Kristin Scott Thomas-led TV show anymore. It's action-packed, dramatic, and comes equipped with that quintessential British humor that I love. Okay, you'll need to stream the other three seasons first, but they're only six episodes apiece, so you'll gallop your way through them, plus season 4, in a couple of days. You won't regret sticking this Apple TV Original on, trust me.

Tom Power, senior entertainment reporter

The Boy and the Heron (Max)

Hayao Miyazaki and Studio Ghibli, you've done it again. Just when I think the legendary director and his animation studio cannot outdo themselves again, they go and prove me wrong with The Boy and the Heron. 

The latest cinematic masterpiece from Miyazaki-san is based on Genzaburō Yoshino's 1937 novel of the same name, albeit it doesn't adapt the story that unfolds in its pages. Nonetheless, Ghibli's take on The Boy and the Heron, which sees its young protagonist Mahito Maki transported to a typically magical land (typical for a Ghibli film, anyway) with a peculiar grey heron's help, is a visually gorgeous, thematically rich, and wonderfully realized movie that deserves to be seen. If it doesn't join our best Max movies guide with immediate effect, I'll be shocked.

Tom Power, senior entertainment reporter

Tell Me Lies season 2 (Hulu/Disney Plus)

If you’ve been eagerly awaiting the next installment of Tell Me Lies, it's out now on Hulu (US) and Disney Plus (internationally). The Emma Roberts-produced series follows the turbulent relationship between Lucy Albright (Grace Van Patten) and Stephen DeMarco (Jackson White) over the course of eight years.

Season 2 picks up with Lucy beginning her sophomore year and, if the trailer is anything to go by, it’s off to an explosive start. Here, we see her trying to just get on with her all important second year of college, only to have plenty of spanners thrown in the works, including trying to avoid Stephen despite him being around every corner. It’s the messiest relationship on TV right now so, even if it's unlikely to make our best Hulu shows and best Disney Plus shows lists, you’ve still got front row seats to all the drama this weekend.

Lucy Buglass, senior entertainment writer 

The Perfect Couple (Netflix)

I’ve been raving about The Perfect Couple for a while now (you may have seen me praising Nicole Kidman’s shady matriarchial character). I do love a good whodunnit, especially when you’ve got an ensemble of truly awful characters that could all be capable of committing whatever the crime is at the heart of its story. Trying to piece together the clues and figure out who the culprit is before the show does is always satisfying, too, although I’ve had mixed results so far.

If you’re in the mood for playing detective this weekend, get stuck into this six-episode series focusing on Kidman's Greer Garrison Winbury, who makes visitors sign an NDA before spending time there. She’s meant to be begrudgingly watching her son marry his fiancée, but a dead body on the beach takes precedent. 

The Perfect Couple hasn't got off to the best start, with its mixed critical reception seeming like a turn off. But hey, when have we ever listened to them? Try it out for yourself and see if it's worthy of a spot on our best Netflix shows guide.

Lucy Buglass, senior entertainment writer 

Wise Guy: David Chase and The Sopranos (Max)

Oscar-winning documentary filmmaker Alex Gibney sits down with The Sopranos creator David Chase in this must-see two-part documentary. With talking heads from some of the cast as well as archive footage, including clips of the late, great James Gandolfini, this is the perfect companion piece for those wanting to know more about not only one of HBO’s most successful TV shows, but one of the best series ever made.

And, honestly, who could blame you? The Sopranos is still so ingrained in pop culture almost two decades (!) after its final season aired, it frequently appears on best HBO Max shows lists and is very quotable – “all this from a slice of gabagool?".

Lucy Buglass, senior entertainment writer 

Selling Sunset season 8 (Netflix)

With eight seasons now under its belt, Selling Sunset is one of a few Netflix reality shows that just keep on giving. It seems there's no shortage of expensive, luxuriously designed homes in Los Angeles so, if you like your architectural digest to be paired with celebrities and cutthroat real estate brokers, this is the series for you.

I can say this with authority, too, because it's one of TechRadar entertainment writer Grace Morris' all-time favorite reality shows to stream on Netflix. Indeed, she's a fan of the agents from The Oppenheimer Group who are "style icons with their jaw-dropping outfits in every episode" – and she's not alone, with critics giving season 5 a perfect 100% rating. As one of the highest-rated new Netflix shows arriving in September, I'm sure there are plenty more fans that can't wait to stream the 11 new episodes, which are all now available. So, what are you waiting for? 

Amelia Schwanke, senior entertainment editor

For more streaming coverage, check out our guides on the best Prime Video movies, best Disney Plus movies, best Paramount Plus shows, and best Hulu movies.

Applied Digital, a Texas-based firm that describes itself as a “designer, builder, and operator of next-generation digital infrastructure” for HPC applications, has confirmed a multimillion-dollar investment from industry leader Nvidia.

The company announced that it had entered into an agreement with Nvidia worth $160 million.

With the capital, Applied Digital will bring its data center and GPU cloud solutions to market “at scale” in order to deliver a “proprietary, purpose-built, hyper-efficient platform for the world’s most advanced HPC and AI workloads.”

Nvidia invests $160m in Applied Digital

The Texas-based company is known for building water-cooled data centers optimized for AI workloads as well as for other cloud computing purposes.

Last quarter, Applied Digital reported a net loss of $64.8 million on a revenue of $43.7 million – it attributed much of the loss to expenses related to equipment and facilities that were not yet operational or generating revenue.

Applied Digital CEO Wes Cummins commented: “We’re proud of our strong relationship with Nvidia, and the confidence that both they and Related Companies, along with institutional investors, have placed in us. Our team is eager to bring to market the vital capacity and contiguous, liquid-cooled IT environments that are tailor-made for AI, HPC, and other accelerated compute workloads.”

The $160 million deal will see 49.4 million shares transfer to Nvidia ownership at a price of $3.24 per share – the last closing price on September 4. In the days following the announcement, share prices climbed to $5.59.

Nvidia has not commented on the deal.

However, the company’s investment may later form part of an investigation if the US DOJ decides to go ahead with an antitrust probe into the company’s alleged habits of stifling competition.

Previously, Nvidia affirmed: “Nvidia wins on merit, as reflected in our benchmark results and value to customers, and customers can choose whatever solution is best for them.”

More from TechRadar Pro

• These are the best AI writers
• We’ve rounded up a list of the best cloud hosting providers
• More subpoenas issued to Nvidia as US continues antitrust case

Check your receipt. Grow your own veggies. Ask about senior discounts. NPR readers share their top tips on cutting costs at the grocery store.

(Image credit: urbazon)

If you like the idea of using a smart speaker to control devices like your thermostat, lights, and switches but aren't enthusiastic about the idea of a device constantly listening in your home, smart home company Aqara may have a solution. 

Rather than passively waiting for you to speak a wake phrase (like 'Hey Google' or 'Alexa'), Aqara's button-like device with a microphone will not begin recording until you pick it up.

Smart speakers don't work by constantly recording everything in your home. Instead, they use software that spots the distinctive acoustic patterns of their wake phrase and ignore everything else. However, not everyone feels comfortable with having a hot microphone in their room.

The AI-enabled Voice Mate H1 gets around that problem by requiring a physical interaction. That does mean you can't use it hands-free, which is one of the main appeals of smart speakers, but it should help ease the privacy worries that continue to surround even the best smart speakers and put many people off trying one.

Although the wireless puck-style design is unusual, it makes the device simpler to operate than a button and allows you to easily carry it between rooms throughout the day.

Compatibility check

The Voice Mate H1 works with Aqara's voice assistant, Copilot, which launched earlier this year. The button requires the Aqara Hub-M3 to work, and according to The Verge, voice commands are cleared from the hub once they're completed.

Of course, all of this is only useful if the Voice Mate H1 works with the rest of your smart home tech. The Hub-M3 supports Aqara's own Matter-compatible Zigbee devices and works as a Matter bridge, but according to the company, it's currently only compatible with third-party sensors (presence, light, and contact), light bulbs, switches, thermostats, and plugs for the time being.

The Voice Mate H1 should work just fine if you stick with Aqara's own range of devices. However, if you want to mix and match your smart home tech, you'll need to make sure they're all compatible with the Hub-M3 first. Cautious homeowners will also want to check Aqara's full privacy policy to see what information the company collects and how it's used.

You might also like

• How to keep your smart home secure
• People are using their Alexa speakers as glorified timers, and it's cost Amazon $25 billion – but is AI the answer?
• Siri forgot how to tell the time on the HomePod – and that shows how much it needs an AI upgrade

Check out these September releases, including some brand-new originals on Tubi and Oscar-winning flicks.

The middle school winner of NPR’s Student Podcast Challenge tells how her golden retriever, Lopez, helped restore her independence.

Today's jobs report further boosts predictions of a Fed rate cut this month.