Technology

• Google Chrome is testing automatic password changes that would follow data breaches
• Google is calling this an ‘AI innovation’, but I’m not convinced
• Chrome already generates and stores passwords, plus checks databases for compromise passwords, and this would put those all together… using an algorithm… maybe?

Google Chrome could be about to implement AI tools to identify passwords found in data breaches, as well as being able to generate and store stronger alternatives.

That’s according to Twitter user Leopeva64, who found the feature in a Chrome Canary test build (via Ars Technica), writing, ‘Another AI-powered feature is coming to Chrome, “Automated password change,” the description mentions that “when Chrome finds one of your passwords in a data breach, it can offer to change your password for you when you sign in’.

That sounds nifty on paper, though it’s worth noting that the best password managers such as Bitwarden and Nordpass have implemented similar features before; so it’s reasonable to suggest that ‘AI’, whatever that umbrella term means here, isn’t actually living up to what Google is calling an ‘innovation’ here.

Chrome’s password ‘AI innovation’

Leaked password databases like ‘Have I Been Pwned’ have previously fulfilled this function, and aGoogle Chrome already uses this repository to inform users when their passwords have been compromised without resorting to ‘AI’.

Password generation is also a feature common to essentially every password manager under the sun, and storing those passwords for easy access (which Google Chrome has also done for some time) is literally the point of having a password manager; they do what they say on the tin!

It’s entirely possible that Chrome’s process of generating passwords is different - and, perhaps, more secure - using some kind of algorithm, but until security researchers explore this, the change amounts to Chrome offering to change a user’s password immediately following a breach. It’s convenient, but I’m also thinking - this is nothing new, and, truthfully, neither is putting ‘AI’ in the feature description.

In case you missed it, Google recently announced that the Enhanced Protection mode in its Chrome web browser’s Safe Browsing settings is protecting 1 billion users (via 9To5Google) from phishing and malware attacks.

You may also like

• Businesses want to turn to AI for more specialized tasks, but aren't always sure how
• Hackers offer 20 million OpenAI credentials for sale, but it says there's no evidence of a breach
• Google's stronghold on search is loosening ever so lightly, report finds, but don't expect it to crumble down overnight

Aiken has several internet service providers offering service in the city. These are the best internet plans, including options from AT&T Fiber.

• Netflix has announced a new epic samurai series as part of its 2025 Japanese content slate.
• The show is a cross between Shōgun and Squid Game, according to the Netflix Japan content head.
• Last Samurai Standing revolves around samurai warriors battling it out for a huge cash prize.

Thanks to Netflix, I've found my next binge-watch after Squid Game season 3 ends with its new upcoming Japanese samurai battle series Last Samurai Standing.

Last Samurai Standing was announced as part of Netflix's Japanese content slate for 2025 and follows samurai warriors who compete for a huge cash prize. The new historical epic is gearing up to become one of the best Netflix shows, as Netflix Japan content head Kaata Sakamoto compared Last Samurai Standing to one of the best Disney Plus shows, Shōgun, and the streamer's biggest TV show ever Squid Game.

Sakamoto said to Variety: “When most people think about samurai, they think about this very glamorous period in Japanese history. But what a lot of people don’t realize is that, towards the end of the Edo period, the samurai lost a lot of their glamour and their power. Last Samurai Standing is about what would happen if these warriors – the toughest and best in Japan – all of a sudden became common people and had to fight for their lives. Think Shōgun meets Squid Game.”

What is Last Samurai Standing about?

Based on the novel Ikusagami by Shogo Imamura, Last Samurai Standing already has a resemblance to Shōgun and Squid Game based on its similar premise. The official logline reads: "In the Meiji period, at Tenryuji Temple in Kyoto, 292 skilled warriors gathered at nightfall, lured by the promise of a huge cash prize. The rules are simple: whoever can steal the wooden tags distributed to each of them and reach Tokyo will win the prize money. Among them is Shujiro Saga, who enters the deadly game with one goal: to save his ailing wife and child."

Going off the plot, Last Samurai Standing has similarities to the record-breaking FX series Shōgun and Squid Game, which broke all kinds of records that even Stranger Things couldn't beat.

The series, which is set to premiere on the best streaming service in November 2025, also features Junichi Okada in the lead role as well as action choreographer and producer. That means it will debut after the next installment Squid Game is released.

Indeed, Squid Game season 3 is set to be released on June 27, and its first clip teases a new mini-game that'll have a big impact on Gi-hun and his fellow contestants. As for Shōgun season 2, the critically-acclaimed series is currently in development at FX studios, so there's plenty to binge-watch between Netflix and Hulu (or Disney Plus for international readers).

You might also like

• Netflix's #1 movie Kinda Pregnant is kinda bad – here are 3 better comedies with over 95% on Rotten Tomatoes to watch instead
• Netflix's new crime drama Apple Cider Vinegar is the latest female fraudster series that I can't get enough of
• 3 movies leaving Netflix with over 85% on Rotten Tomatoes that you won't want to miss

• OpenAI Deep Research is coming to Plus and free tier users
• CEO Sam Altman has all-but-confirmed Plus users will get 10 uses a month and free tier users will get two uses a month
• The release date of Deep Research outside of ChatGPT Pro is still TBD

OpenAI's new AI agent, Deep Research, will be coming to ChatGPT Plus and Free users shortly, and now we know just how many times a month you'll be able to use the research analyst.

Sam Altman, OpenAI's CEO, shared the info on X, all-but-confirming that Plus users ($20/month) will get to use Deep Research 10 times per month while free users will be able to use it two times per month.

Not only that, Altman also hinted at a future increase in the usage on Plus and free tiers by saying they have the "intent to scale these up over time."

ChatGPT Deep Research launched at the start of February and is currently only available to ChatGPT Pro users to the tune of $200/month (100 queries per month). The AI agent is an incredible tool that OpenAI says can "accomplish in tens of minutes what would take a human many hours."

With the ability to do work for you, by simply asking the AI tool to conduct research and report its findings, Deep Research is an absolute game-changer for productivity and it's incredible that we're going to be getting access to its power for free.

i think we are going to initially offer 10 uses per month for chatgpt plus and 2 per month in the free tier, with the intent to scale these up over time.it probably is worth $1000 a month to some users but i'm excited to see what everyone does with it! https://t.co/YBICvzodPFFebruary 12, 2025

$1000 a month in value?

Altman's announcement was in response to an X user who said, "ok, OpenAI Deep Research is worth probably $1000 a month to me. This is utterly transformative to how my brain engages with the world. I'm beyond in love and a little in awe."

To which Altman responded, "It probably is worth $1000 a month to some users but I'm excited to see what everyone does with it!"

We're yet to try Deep Research ourselves, but considering the examples shown on OpenAI's website, this new AI agent could completely revolutionize the way we interact with AI. OpenAI's website says, "Deep research is OpenAI's next agent that can do work for you independently – you give it a prompt, and ChatGPT will find, analyze, and synthesize hundreds of online sources to create a comprehensive report at the level of a research analyst." That may sound like a professional tool, just for professional use, but some of the examples which you can read about in my article, 3 reasons why I can’t wait to use OpenAI Deep Research, showcase just how useful this tool could be for everyone, not just those using it for work.

While Altman has confirmed Deep Research will come to Plus and free tiers, we still don't have a timeframe on when that will be. Here's hoping it's sooner rather than later.

You may also like

• I tried ChatGPT Search and now I might never Google again
• If you want to know who will win the AI wars, just watch these two Super Bowl ads from Google and ChatGPT
• OpenAI's Deep Research smashes record for the world's hardest AI exam

Vision Pro and M4 MacBook Air announcements are also expected in the coming weeks.

Just 30% of EU member states have nationally implemented The Network and Information Security Directive (NIS2). While this is not unusual given that attention is divided among NIS2, the Digital Operations Resilience Act (DORA), and the European Cyber Resilience Act (CRA), target deadlines are being missed. This puts businesses on the back foot and risks falling even further behind the adoption curve.

NIS2 is the biggest cybersecurity legislation in over a decade, setting the stage for secure digital infrastructure across the EU. The directive's deadline has passed, as of 17 October 2024, but like many other EU regulations, NIS2 is taking a phased approach across Europe — much like when GDPR was introduced in 2016.

Although businesses in the UK are not obligated to comply with these regulations, those with EU-based clients or those that make up part of EU supply chains will come under its thumb.

A view of NIS2

The driver for this legislation is clear. Recent high-profile cyber incidents have caused major disruption — like the cyberattack on the NHS, the CrowdStrike outage, and the SolarWinds fiasco. As society increases its dependency on software, software developers increasingly rely on open source to ship new products quickly. Open source makes up 90% of modern software, with consumption reaching a staggering 6.6 trillion downloads in 2024. Cyber resilience must improve commensurately to meet software consumption.

NIS2 introduces stringent cybersecurity reporting standards across banking, manufacturing, and public administration. UK businesses working with EU partners are already being asked to meet these standards and contracts are being rewritten to demand adherence to NIS2 standards to protect shared systems and data. Remember that the rising tide lifts all boats.

Under the NIS2 directive, companies have 24 hours to report major cybersecurity incidents, with updates due within 72 hours and a final report needed in 30 days. They must also implement and document policies, from vulnerability monitoring to information security training, which exceeds previous requirements.

Vulnerabilities in software supply chains spread rapidly. Even though NIS2 isn’t law in the UK, the underlying principles of the directive are good cybersecurity best practices to observe. Non-compliance not only puts UK companies at greater risk of security incidents, but as with GDPR, we see cybersecurity legislation trending in this direction. Adapting early to these changes avoids unnecessary operational strain, especially for those at risk of being pulled into scope by upcoming EU legislation such as the CRA and Product Liability Directive.

The observability conundrum

Software supply chain security has become a major target for bad actors, with open source malware, or malicious packages living in open source repositories, surging 156% in 2024. Bad actors often target third-party vendors as a gateway into larger organizations, proverbially killing several birds with one stone. Recent high-profile incidents show that their methods are becoming more sophisticated.

One such example is the April 2024 attack on the open source XZ Utils compression tool, where a bad actor posed as a legitimate contributor for two years, embedding malicious code that, if undiscovered, could have caused widespread global disruption.

NIS2 has been introduced specifically to combat these occurrences by imparting personal responsibility for cybersecurity to businesses themselves. Rather than being able to skirt accountability by blaming incidents on suppliers, companies are now personally liable for their cybersecurity. This development is sorely needed, as the rate of software consumption dwarfs the rate at which components are observed and tracked, and that gap is only widening. In the past 12 months, only 60,000 Software Bills of Materials (SBOMs) were published versus 6.9 million new open source components released. This gulf is a ticking time bomb, with businesses facing no-fault liability and harsh financial penalties.

What would Anne Robinson say?

When we think of software, we think of it as one system. The reality is that all software is an interdependent mesh of open and closed source components, and one weak link can be disastrous. On average, each application contains 180 components, with repeated updates, patches, and multiple versions. It's impossible to track every single change manually. Automation and proper tooling are needed to keep up with the pace of consumption that modern software development demands.

If just one component is compromised, the ripple effect can be devastating. This is precisely why regulations are coming thick and fast to identify and remedy the weakest links in software supply chains.

UK businesses, while not directly under NIS2, should be mindful of its implications. Even though the EU is the first to draw a line in the sand on this issue, it will not be the last. Businesses stand to benefit by being proactive rather than reactive. Preparing for the upcoming CRA Bill helps companies stay competitive amid the changing legislative landscape.

The risk of being the weakest link far outweighs the compliance challenges, as protecting partners from costly cybersecurity incidents, compliance failures, and reputational damage safeguards a company's long-term resilience and, ultimately, its bottom line.

We've featured the best business VPN.

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

• Bishop Fox found a way to abuse a SonicWall VPN flaw
• It allows threat actors to bypass authentication and hijack sessions
• There are thousands of vulnerable endpoints

A major vulnerability in the SonicWall VPN which can be exploited to hijack sessions and access the target network has now seen its first proof-of-concept (PoC) attack, meaning it’s only a matter of time before cybercriminals start exploiting it in the wild.

In early January 2025, SonicWall raised the alarm on a vulnerability in SonicOS and urged its users to apply the fix immediately. The flaw is tracked as CVE-2024-53704, and described as an Improper Authentication bug in the SSLVPN authentication mechanism. It was given a severity score of 9.8/10 (critical) and was said it could be abused to allow a remote attacker to bypass authentication.

It impacted SonicOS versions 7.1.x (up to 7.1.1-7058), 7.1.2-7019, and 8.0.0-8035. SonicWall released versions SonicOS 8.0.0-8037 and later, 7.0.1-5165 and higher, 7.1.3-7015 and higher, and 6.5.5.1-6n and higher, to address the bug. At the time, there were more than 4,500 internet-exposed endpoints.

Protect yourself from identity theft online

Go Incogni and get 55% off using code TECHRADAR. Incogni erases you and your family from the sites that expose your personal information to identity thieves and robocalls.View Deal

Proof of Concept

Now, since SonicWall users were given enough time to patch, security researchers from Bishop Fox came forward with more details about the vulnerability, as well as a PoC. After a “significant” reverse-engineering effort, Bishop Fox said the vulnerability could be exploited by sending a custom-built session cookie containing a base64-encoded string of null bytes to the SSLVPN authentication endpoint.

This results in the endpoint assuming the request was associated with an active VPN session and incorrectly validates it. As a result, the target is logged out, while the attacker gets access to the session, including the ability to read the victim’s Virtual Office bookmarks, access VPN client configuration settings, open a VPN tunnel, and more.

"With that, we were able to identify the username and domain of the hijacked session, along with private routes the user was able to access through the SSL VPN," the researchers said.

Via BleepingComputer

You might also like

• This major firmware flaw is affecting Intel-powered PCs across the world
• We've rounded up the best password managers
• Take a look at our guide to the best authenticator app

• Apple has unveiled a major new health study
• The company wants to better understand how technology can improve physical health, mental health, and overall wellbeing
• You can take part using your iPhone, Apple Watch, or AirPods

Apple has unveiled a major new medical study that could completely change the way we think about technology, health, research, and medicine.

The Apple Health Study "aims to further understand how technology can play a role in advancing and improving physical health, mental health, and overall wellbeing." That includes Apple's own best iPhones, Apple Watches, and AirPods. The new study is available in Apple's Research app, and "will also explore relationships between various areas of health, such as mental health’s impact on heart rate, or how sleep can influence exercise."

The study is being conducted in collaboration with leading research outfit and Harvard teaching affiliate Brigham and Women's Hospital.

Apple says it hopes to help tackle one of the biggest issues with medical research – limited numbers of participants and data. The company says that more than 350,000 people in the US have taken part in previous studies for women's health, hearing, and heart health and movement. Here are the details and how you can get involved.

The most ambitious Apple Health Study yet?

The Apple Watch Series 10's humble heart rate monitor could power medical studies for years to come. (Image credit: Future)

Apple says the new study "aims to understand how data from technology – including Apple and third-party devices – can be used to predict, detect, monitor, and manage changes in participants’ health," and will also explore the connections between different areas of health.

It spans a multitude of different health and disease areas, including activity, aging, cardiovascular health, circulatory health, cognition, hearing, menstrual health, mental health, metabolic health, mobility, neurologic health, respiratory health, and sleep. Previously, Apple has focused its studies on one such area, like hearing, making this new study the most ambitious yet by some margin.

Apple says the new study "is designed to explore changes in health and how technology can help identify important insights for future product development," noting that when an aspect of health changes the body can emit a physical or emotional signal. Identifying these changes can help to offer a more proactive approach to health, with monitoring performed using health features like sleep apnea detection on the Apple Watch Series 10.

The best bit? If you live in the US you can take part. If you meet the minimum age requirements, you can sign up for enrolment through Apple's Research app. Participants can perform tasks and ask questions, as well as read research. As you'd expect with Apple, you can chose exactly what data you do and don't share, and you can withdraw at any time.

You may also like

• The Apple AI health coach sounds like it's no longer stuck in development hell
• I've spent 10 years with Apple Watch, and after 10 minutes with the Galaxy Watch Ultra, I can't believe Samsung skipped this feature
• Three Apple Watch features you're probably not using, but should be – according to Apple's VP of Fitness Tech

Netflix's hit reality show is nearly back.

• Ivanti patches four bugs found in Connect Secure, Policy Secure, and Cloud Services Applications
• All four could be used in RCE attacks
• Patches are available, and users are advised to apply them ASAP

Ivanti has released patches for four critical-severity vulnerabilities discovered in a number of its products.

The vulnerabilities are tracked as CVE-2024-38657, CVE-2025-22467, CVE-2024-10644, and CVE-2024-47908. All of these were given a severity score of 9.1/10 (critical). Different bugs impact different solutions, including Connect Secure (ICS), Policy Secure (IPS), and Cloud Services Applications (CSA).

Since they can be used for highly disruptive cyberattacks, users are advised to apply the patches without hesitation - and a security advisory containing more details about the above-mentioned flaws can be found on this link.

Protect yourself from identity theft online

Go Incogni and get 55% off using code TECHRADAR. Incogni erases you and your family from the sites that expose your personal information to identity thieves and robocalls.View Deal

Stepping stones

All of the bugs can be used to execute arbitrary code, remotely. The first clean versions are Ivanti Connect Secure 22.7R2.6, Ivanti Policy Secure 22.7R1.3, and Ivanti CSA 5.0.5, and users are advised to upgrade to these versions straight away.

Ivanti said there is no evidence of in-the-wild abuse. However, the company’s products are quite popular in both enterprises and small and medium businesses (SMB), and as such are often targeted and used as an initial entry point.

"While these products are not the ultimate target, they are increasingly the route that well-resourced nation state groups are focusing their effort on to attempt espionage campaigns against extremely high-value organizations," Daniel Spicer, Ivanti CSO, said.

"We have enhanced internal scanning, manual exploitation and testing capabilities, increased collaboration and information sharing with the security ecosystem, and further enhanced our responsible disclosure process, including becoming a CVE Numbering Authority."

In late January 2025, the US Cybersecurity and Infrastructure Security Agency (CISA) added four Ivanti vulnerabilities to its Known Exploited Vulnerabilities (KEV) catalog, suggesting they were being abused in the wild. The bugs, found in Ivanti Cloud Service Appliance (CSA) and patched in September and October 2024, are being used in two attack chains to gain initial access, conduct RCE, obtain credentials, and impact webshells.

Via The Hacker News

You might also like

• Hackers are still using old Ivanti bugs to break into networks
• We've rounded up the best password managers
• Take a look at our guide to the best authenticator app

• Leica's next M-series camera could be the first of its kind with an EVF
• It could be the smallest, lightest and most versatile Leica M-series camera yet
• Rumors suggest it'll launch later this year

Now here's a Leica rumor with plenty of potential avenues to navigate – Leica is apparently working on a new M-series camera that will ditch the rangefinder experience that the series is famed for, in favor of an electronic viewfinder (EVF).

The details recently shared by Leica Rumors are few and far between, leaving me with plenty of questions – so let's explore where Leica could be going with this potential new model.

Leica's M-series, both analog and digital models, utilize an optical rangefinder to compose and manually focus shots, using a hard-to-master split-image guide to nail sharp focus on your subject. It's not for everyone, but Leica's rangefinder is the soul of the series, a shooting experience you can't find elsewhere (besides the Pixii rangefinder).

Practically speaking, an EVF will be much more useful than a rangefinder – you would get exposure preview and the supremely helpful 'focus peaking' manual focus assist tool. Such a design change should make this mystery model Leica's smallest and lightest digital M-series camera to date too, and also the cheapest.

Rumors suggest this is not a next-gen Leica M12, but a new series within Leica's M-mount ecosystem. An M-series camera with EVF sounds to me a lot like a Leica Q3 / Q3 43 premium compact (see above), but with interchangeable lens mount. I loved testing the Leica Q3, but it is an altogether different and modern camera experience versus the M-series.

The M11 is one of a legendary series of digital Leica cameras that utilizes a rangefinder for composing shots and manual focusing, whereas the rumored new model could be the first-ever M-series camera with electronic viewfinder. Could the mystery model pave the way for autofocus too? (Image credit: Leica) A new dawn for Leica's M-series?

Swapping out the rangefinder for an electronic viewfinder might sound like a small change, but it would mark the single biggest design overhaul that Leica's M-series has ever seen. It would also take away Leica's unique selling point – the rangefinder experience.

Some feel that rangefinders are stuck in the past – a split-image manual-focus only shooting experience that is hard to master, while other mirrorless camera systems by the likes of Sony and Canon enjoy autofocus that's so effective that you, the photographer, barely need focusing skills anymore. For others, the rangefinder experience is the exactly the challenge they enjoy, a charm they can't get elsewhere (notwithstanding the Pixii, which is the only other digital rangefinder in the market).

Let's consider the possibilities of an M-series camera with an EVF. If it is indeed an M-series camera, it will use Leica's M-mount, for which there are many manual-focus lenses available, both pricey Leica options and cheaper third-party alternatives. The lenses are tiny, and would suit what could potentially be Leica's smallest ever M-series camera.

When I picture the mystery M-series camera as a Q3 with Leica M-mount, my mind turns to one feature the Q3 has that the M-series lacks: autofocus. Could a new Leica M camera ever get autofocus? The lens mount would need a fresh redesign, and new autofocus lenses would need to be made, but it seems plausible, even if it would mark a new direction for Leica.

Then there's the Leica SL-series to consider, the latest generation being the Leica SL3. It's a DSLR-style full-frame camera with L-mount, complete with autofocus skills. There's an enticing case for a compact, M-series style camera with an EVF and the autofocus-equipped L-mount.

However, the rumors point to an M-series model with an EVF, a particularly large touchscreen and no rear buttons, and I don't expect any of the seismic changes dreamed up above. This will likely also be a manual focus-only camera, making the most of the assist tools an EVF delivers, and presumably Leica's latest 61MP sensor. In-body image stabilization would be helpful addition, but again, doubtful.

What is the most exciting direction that you think Leica could take its M-series with its next launch? Let us know in the comments below.

You might also like

• Leica Q3 43 review: a stunning premium compact with a unique focal length
• Street photography with the Leica M11-D is the closest experience I’ve had to analog using a digital camera
• Leica's new iPhone camera grip could have been great, but has 3 frustrating drawbacks

• All current AirPods have got a lick of paint from ColorWare
• 'The Blended Collection' offers neutral skin-tone colors
• But you'll be paying a pretty penny for the makeover

Bored of your plain, ceramic-white AirPods? Wish you could get a splash of color into your ears? These new custom-painted AirPod models from ColorWare might be what you need. ColorWare is a long-running company that takes new products, gives them a high-quality new paint-job, and sells them to folks with a mark-up.

The new ColorWare line gives Apple’s current headphone range a new lick of paint, with the AirPods 4, AirPods 4 with ANC, AirPods Pro 2 and AirPods Max all benefiting from the makeover.

Each model now comes in a variety of neutral options designed to match a few varieties of skin tones, including ‘Glaze’ (light pink/cream), ‘Toasty’ (light brown/beige), and ‘Java’ (a deepr brown).

It’s a far cry from the stark white coloring of most AirPods off the shelf, which have changed in shape and design over the years but otherwise stuck with Apple’s minimalist tone.

Some Apple ranges do push the envelope a little with their color palette, like the vibrant iMac M4 line, or ‘desert titanium’ colored iPhones. And the AirPods Max do break with AirPod tradition, with colorways including silver, sky blue, green, and pink.

But generally, Apple is known for sticking to a core set of tones for brand consistency, and to make sure its Apple-ness is visible from a quick glance.

The AirPods and AirPods Pro have stuck to a monochrome design over the years. (Image credit: Apple) A new era?

We previously reported how Apple had planned to release AirPods in a mix of different colors (pink, red, purple, black, blonde) in order to match the iPhone 7 that was on the market at the time, but the tech giant ultimately decided to keep things boring – I mean, simple.

Headphones are key consumer accessories these days, and shoppers are often forced to buy elsewhere in order to get the aesthetic they’re after – possibly sacrificing the audio experience in the process. These repainted (upcycled?) AirPods offer the best of both worlds, with Apple hardware in a fun, new finish.

Of course, other Apple-made headphones manage not to be so plain. The new Beats Powerbeats Pro 2 will come in some fun colors, and the ColorWare 'Blended' range of skin-aping tones is very similar to Beats' collection with Kim Kardashian.

And yes, you will be paying a premium for these. The custom AirPods 4 retail for $329, roughly doubling the original price, while the custom AirPods Pro 2 now cost $439, with a similar markup. The AirPods Max has seen a 50% price increase, bringing its $549 official price up to $849.

Colorway’s website says the painted AirPods are “only available for a limited time”, though, so if you’re not put off by the price increase, now might be the time to buy.

You might also like

• Apple finally gives instructions on how to update AirPods… sort of
• Apple's banking on camera-equipped AirPods to trounce Meta smart glasses, but I'm not sure it's enough
• More of the best wireless earbuds in 2025, all real-world tested

French AI developer Mistral AI has brought its AI chatbot, aptly named Le Chat, to mobile devices. Le Chat is a European alternative to U.S. offerings like ChatGPT and Google Gemini and Chinese-based tools like DeepSeek.

Though Mistral has established a presence among AI developers, this is their first real attempt at a consumer-facing chatbot. Naturally, I wanted to test it out and see how well it worked. Although the abilities of these AI assistants overlap, it is still worth seeing if Le Chat could keep up.

I didn't want to just test Le Chat in isolation, though. I decided to put it head-to-head against ChatGPT, the current default for many people when it comes to AI chatbots. I figured if Le Chat could hold its own in direct comparison, it was worth paying attention to.

I tested both chatbots with prompts the average person might submit to an AI chatbot to help them out. I wrote prompts to solicit both AI chatbots for advice on improving someone's social life, solving a riddle, explaining complex issues to novices, and producing an image. Here’s what happened.

Friendly help

(Image credit: Screenshots from Le Chat/ChatGPT)

I started with a prompt that someone might ask an AI when in a new place and trying to figure out their social life. I asked both AI chatbots the following: “I just moved to a new city and don’t know anyone. What are some practical ways to make new friends as an adult?”

Both chatbots had some solid tips. Le Chat’s response included ten ideas with accompanying explanations, and they were solid but lacked a lot of detail. ChatGPT went much more specific in coming up with apps and other details to pursue in making friends. Le Chat could do the same with some follow-ups, but at least initially, it kept to a more generic set of advice. That might be its international flavor since some of ChatGPT's advice would only make sense in the U.S. due to the availability of some mobile apps and kinds of activities.

Riddling

(Image credit: Screenshots from Le Chat/ChatGPT)

I always like using wordplay and logic questions to test out AI chatbots. I went for an old classic here, challenging the AIs to answer: “I speak without a mouth and hear without ears. I have no body, but I come alive with wind. What am I?”

Le Chat and ChatGPT were both quick to give the correct answer, though ChatGPT, for some reason, had an enthusiastic exclamation point. The chatbots' explanations were also practically identical, suggesting that the choice between ChatGPT and Le Chat doesn't really matter in this case.

Make sense of mortgages

(Image credit: Screenshots from Le Chat/ChatGPT)

To test how well the AI chatbots could explain something complex, I went with: “Can you explain how a mortgage works in simple terms for someone who has never owned a home before?”

As can be seen above, both chatbots had no problem breaking down the definitions and functions of different aspects of a mortgage. If anything, they seemed to be pulling from similar sources. I noted that Le Chat's European origin didn't prevent it from using American dollars as an example.

Otherwise, the only significant difference was that Le Chat's tone was a bit more formal than ChatGPT, which offered a more conversational phrasing when describing how mortgages work.

Picture This

(Image credit: Screenshots from Le Chat)

(Image credit: Screenshots from ChatGPT)

Le Chat has multimodal capabilities, including a model for creating images. To wrap up the test, I devised a somewhat complex prompt for an image to see how the two models would compare on a visually creative task. I asked each chatbot: "Create a vibrant, high-fantasy illustration of a fearless medieval knight battling a colossal emerald-green dragon atop a mountain. The knight, in gleaming silver armor with gold engravings, wields a rune-glowing greatsword and raises his shield. The dragon coils around ancient stone ruins, smoke rising from its nostrils. A fiery sunset casts dramatic hues over swirling clouds and distant mountains."

Both pictures look great despite notable differences in the details. The cinematic lighting and style are shared, but otherwise, the mountains, knights, and dragons are very different. Le Chat's dragon is more of a giant snake, though the knight's sword does have glowing runes, which ChatGPT's image lacks.

That said, if you look closely, you will see that both images have classic AI flaws. ChatGPT's dragon's wings don't seem to connect to its body, and the only visible leg is oddly placed. Le Chat's knight has a shield that just sort of floats next to him, and the dragon's coils are arranged more like an Escher painting than a terrifying mythical beast. And the less said about the shape of the rune-carved gravestones in ChatGPT's image, the better.

Vive Le Chat

As expected, there's no clear “winner” between the two chatbots so much as two solid AI assistants. Le Chat did surprise me with just how fast and efficient it is, but at least from my tests, it's a little brusque and somewhat broad in its answers. That's not a bad thing if you want generalized answers and advice in a hurry.

ChatGPT's answers sometimes seemed more like what a human would say in response to the prompts, with emotion-tinged language that appeared to match the energy of the requests, even if just to seem enthusiastic about a riddle. Further, while both AI chatbots produced exciting, if very flawed images, at least ChatGPT understood that the dragon fighting a medieval knight usually has wings, even if they aren't explicitly mentioned in the prompt.

If I were asked to pick just one, I'd probably lean toward ChatGPT, but that may be as much about familiarity as anything else. If I lived in Europe, I might opt for Le Chat simply because it is more likely to avoid any regulatory pitfalls faced by OpenAI because it is based in the region.

Otherwise, both AI chatbots can handle anything a casual AI chatbot user might toss their way. That should probably worry OpenAI as it strives to maintain a position of superiority, or at least the perception of such, against rivals like Mistral that may have the wind at their backs.

You might also like

• Yes, Google's new Gemini 2.0 Flash is much better than the old 1.5 model – here's what I found
• Google is bragging about Gemini 2.0 Flash, but how does it compare to ChatGPT for casual AI chatbot users?
• ChatGPT Tasks can start taking over your calendar and remind you to finish your to-do list

The chief information officers of at least three major government agencies have been replaced by Silicon Valley executives, including from Palantir and Elon Musk's SpaceX.

• There has been a surge in AI-related romance scams
• 17% of people believe somebody they know has been approached by an AI chatbot on dating apps or social media
• Almost half of British people believe it is possible to develop feelings for an AI chatbot

Valentine's Day may be approaching fast (it’s this Friday, in case you were wondering), but for singletons, the already hellish landscape of the online dating world has an added tech-fueled nightmare they have to look out for. Rather than simply being a timewaster, it’s quite possible that the person you've been flirting with online might not even be human.

According to new research by McAfee, there has been an alarming rise in AI-driven romance scams. Nearly 17% of people (roughly one in five) say that somebody they know has been approached by an AI chatbot posing as a real person on a dating app or social media.

"Technology has revolutionised communication by enabling instant connections across the globe, fostering deeper relationships and collaboration regardless of physical distance – but unfortunately, it's also transforming the way scammers operate," said Vonny Gamot, Head of EMEA at McAfee.

Love is a battlefield

It seems that even people who work in cybersecurity for a living can fall victim to AI romance scams. Take Rob S, a cybersecurity professional, who thought he had made a connection with somebody: “We met in a group chat and really hit it off. After exchanging numbers, we started texting and getting to know each other.”

His suspicions were raised when his potential romantic partner started asking for money: “When she asked for more money, I started to notice the red flags. Her messages were repetitive, her number didn’t make sense for her location, and I realised I might have been speaking with a bot – or, at the very least, a scammer.”

“Looking back, a lot of her messages were generic and didn’t feel personal. That’s when it hit me – there was a real chance I wasn’t even talking to a person.”

One in two British people could fall for an AI

According to the McAfee survey, almost half (46%) of British people believe it's possible to develop romantic feelings for an AI chatbot.

With this in mind, McAfee has a few tips on how to stay safe when looking for love online this Valentine’s Day:

Firstly, if it’s too good to be true, then it probably is. Red fags are: people avoiding video calls, moving conversations off-platform quickly, or declaring love too soon.

Make sure you verify the person’s identity as soon as you can. Do a reverse image search on their profile picture and check social media activity to confirm they are real.

Remember, scammers will use small details like your birthday or pet name to guess passwords, so be vigilant. Never send money or gifts.

Finally, if you have an AI PC, then you can take advantage of McAfee Deepfake Detector to alert you of deepfake audio within videos.

You may also like

• I asked Bing's ChatGPT about love. The results broke (and mended) my heart
• Hinge's AI can help make your dating profile shine – and maybe win at the game of love
• Can Grindr's AI Cupid help you find love?

• Google and Samsung's Atmos rival, Eclipsa, is coming to LG and Google TVs
• The 3D audio format will be baked into Android 16 too
• No support from any streaming services except YouTube so far

There’s a new 3D audio format in town, but when will TV owners get to hear it?

While informed listeners may know their way around Dolby Atmos, DTS:X or even the Auro-3D format coming to Nvidia Shield devices, Eclipsa is a new open-source 3D audio format developed by Google and Samsung.

Imagined as a competitor to Dolby Atmos – much like how Samsung’s HDR10+ picture format rivals Dolby Vision – Eclipsa was formally unveiled at CES 2025 and confirmed to launch in all 2025 Samsung TVs and soundbars.

Speaking to FlatpanelsHD, Google representatives confirmed that Eclipsa will be baked into the Android 16 operating system, meaning it will get support on Google TVs and other smart TV makers that use the Android OS, such as Hisense, TCL, Sony and Philips TV. Beyond that, the aim is to get Eclipsa Audio running on mobile devices and web browsers, alongside AV receivers. FlatpanelsHD also reports that support will come to webOS, used on LG TVs.

But the question remains: where and how will anyone access Eclipsa-coded audio?

YouTube should be the first main beneficiary of Eclipsa, but will creators use it? (Image credit: Shutterstock / Hamara) The waiting game

Eclipsa is channel-based, meaning it ties sound sources to specific channels (like in a 2.0 stereo or 5.1 surround-sound setup) rather than individuating each sound source (as in the object-based Dolby Atmos format). However, Eclipsa supports 'up to 28 input channels' according to FlatpanelsHD, so there's more than enough information to create a convincing 3D soundspace for home listening. While it trades a little flexibility, and specificity of location, it’s also somewhat easier to deploy and recreate consistently.

Given Google’s involvement, Eclipsa Audio is being angled as a multipurpose 3D audio format that can roll out across its massive network of Android devices for any audio source, even humdrum YouTube videos, rather than focusing primarily on high-end entertainment output like movies, TV shows, music and games. YouTube creators will apparently be able to upload Eclipsa-coded content later in 2025.

In a blog post detailing the new format, Google says that “spatial audio technologies were first developed over 50 years ago, and playback has been available to consumers for over a decade, but creating spatial audio has been mostly limited to professionals in the movie or music industries. That’s why Google and Samsung are releasing Eclipsa Audio, an open source spatial audio format for everyone.”

However, 3D audio formats generally have been pushed on high-end TV models and major TV streaming services, with interest driven by cinephiles/audiophiles eager to improve their home theater experience with more immersive sound. It's not clear, either, whether YouTube creators will feel incentivised to start creating content in the new format en masse.

As FlatpanelsHD notes, the Atmos-minded Netflix has yet to voice its support, despite its involvement with Google and Samsung in the Alliance for Open Media (AOM), and Google reps couldn’t comment on Apple’s plans for its tvOS or iOS platforms either.

The notion of open-source 3D audio is very appealing, especially if it does improve the kind of everyday audio experienced through Android smartphones and tablets, or through the YouTube app. But it’s going to be hard to shout about Eclipsa until some major players in the field – and content creators on the ground – actually decide to get involved. A format on the best TVs and best soundbars does no good if there's nothing to watch.

You might also like

• Sonos' streaming box sounds like a dream product, apart from the 2 horrendous flaws
• Is this designer OLED TV worth its explosive price tag?
• Philips' new elite OLED TVs let you choose between extra image quality or better built-in sound

• A dataset containing 14 million details has been discovered
• Leaked information looks to belong to international shipping platform Hipshipper
• Victims are at risk of identity theft and phishing attacks

Nobody is safe from data breaches, and something as simple as ordering a parcel from a reputable company can put you at risk. This is exactly the case for 14 million unlucky shoppers, as an open instance was discovered unsecured online.

Researchers at CyberNews found the instance originated from an unprotected AWS bucket which belonged to Hipshipper - an international logistic and shipping company that works with sellers on both eBay and Amazon, offering delivery and returns to over 150 countries.

The researchers discovered the open instance in December 2024, and the leak was only closed in January 2025, so was open for at least a month - here’s what we know.

Protect yourself from identity theft online

Go Incogni and get 55% off using code TECHRADAR. Incogni erases you and your family from the sites that expose your personal information to identity thieves and robocalls.View Deal

Personal Information exposed

It's pretty easy to imagine how an attacker could use your shipping details to cause harm, and the leaked information included buyer’s personal information like full names, home addresses, phone numbers, and order details.

“Cybercriminals can exploit leaked data to orchestrate advanced scams and phishing attacks,” the researchers explained.

“For example, crooks may impersonate trusted businesses and distribute fraudulent messages that leverage specific order details to demand urgent verification of personal or financial information.”

There’s ‘no indication’ that cybercriminals accessed the exposed dataset, but criminals very often have ways to scan the internet for open instances such as these.

Retail firms are one of hackers most targeted industries, and unfortunately, only using large, reputable companies doesn’t protect your information from leaks - as retail firms like GrubHub, Mizuno, and Hot Topic have all suffered significant breaches in the last few months.

In fact, since 2004, over 17 billion accounts have been breached. Of course, that statistic is a little misleading, as some people will have had many accounts exposed whilst others remain untouched - but it does illustrate the scale of the problem, and reminds us that anyone could be at risk.

But whether your account has been breached once or a hundred times, the dangers are the same.

Protecting yourself

If you’re affected by a data breach, you should be very wary of identity theft - and the software listed can provide dark web monitoring, credit monitoring, even insurance if you do fall victim.

If you want to stay safe on your own, the key is staying vigilant. Keep a close eye on your accounts, statements, and transactions - report any suspicious activity to your bank immediately.

There’s also a risk of phishing attacks when your data is exposed - as criminals can use the information to craft personal and specific emails in order to trick victims into believing the attacker is a friend, colleague, or family. But that’s not all, CyberNews researchers explained, as “revealing personal details may even pose risks to physical safety.”

“Criminals could use this information for stalking, harassment, or planning burglaries. Furthermore, attackers may compile and use leaked data for financial or personal gain, often subjecting victims to harassment, reputational damage, or other harmful actions.”

Be extra careful if you receive unexpected communications, especially from someone you don’t know. Be sure to thoroughly look into every email address messages are sent from, and don’t click any links you don’t 100% trust.

We’ve written a full guide on how to avoid online phishing to better protect yourself if you need more information.

You might also like

• Check out our list of the best firewall software around today
• Foh&Boh data leak leaves millions of CVs exposed - KFS, Taco Bell, Nordstrom applicants at risk
• We've also rounded up the best malware removal software on offer right now

Get a laptop that will help you get your business done with Windows laptops or MacBooks, tested and reviewed by CNET's laptop experts.

Here's how FromSoft is bottling its tough combat magic into bite-size play loops.

I got the inside scoop on Apple Health's new five-year research study. All Apple users can participate starting today.