TechRadar News

Since the term AI was first introduced in 1956, our reliance on artificial intelligence has grown immensely—much more than we ever anticipated. It’s not just a concept from futuristic sci-fi films anymore; today, nearly every sector, including highly regulated fields like finance, has embraced AI to enhance its competitive edge and boost profitability.

Companies around the globe are now integrating AI to streamline their operations, stay ahead of their competition, and solidify their positions in the market. Honestly, it’s hard to imagine any CEO ignoring the potential AI has, when used properly, to make our work lives easier, more efficient, and cost-effective. I’ll say it again: when used properly.

While AI is certainly driving growth for many businesses, it's important for us as leaders to take a moment to reflect: Are we truly leveraging what we refer to as "artificial intelligence," or are we actually dealing with "assisted intelligence"? Have we fully tapped into the capabilities of genuine "artificial intelligence"?

From my perspective, we still have a long way to go to reach that goal. Although I am not yet a believer in the full potential of AI, I am a strong advocate for assisted intelligence. For example, using machines to analyze thousands of messages in search of potential non-compliant language and behavior is a practical application of this technology. But it all starts by understanding the difference between what is assisted intelligence and what is artificial intelligence.

The Role of Assisted Intelligence in Modern Workplaces

If we really think about how “AI” has changed our daily lives - is it really doing what it says on the tin? While many of us call it "Artificial Intelligence," a more accurate term might be "Assisted Intelligence." Thanks to innovation in machine learning and natural language processing, many workplaces are changing for the better—about 75% of knowledge workers are already using AI tools to their advantage. However, the idea of AI being fully independent and capable of independent thought is still a long way off.

And that’s not necessarily a bad thing!

Many companies are not yet ready for these types of - advanced - AI solutions. When we take the time to properly examine their primary challenges, it becomes very clear that the biggest obstacle is managing the vast amounts of data generated each day. In fact, over half of employees—around 53%—report feeling overwhelmed by this amount of data, which hinders their ability to engage in strategic thinking, or what I refer to as "meaningful work."

Here is an example: a company with 100 employees that sends out 100 messages each day generates approximately 200,000 messages each month. Even if the company implements policies to monitor just 5% of these messages, it would require a large compliance department to read through them all. Instead of relying on random sampling, companies should allow technology to analyze all messages in real-time and identify the ones that are significant. This way, compliance teams can focus on essential tasks and ensure that the company remains safe from any fines.

So, what can organizations do? They need solutions that can cut through the data clutter and offer clear, actionable insights, allowing teams to concentrate on what truly matters.

Navigating Uncertainty in AI

It’s important to recognize that AI has the potential to boost global corporate profits by up to $4.4 trillion a year. This possibility has business leaders everywhere eager and excited to incorporate AI into their operations for better efficiency and smarter decision-making.

But there’s also some hesitancy from cautious CFOs and compliance officers who want to make sure their organizations really understand the needs, benefits and risks of new technologies - especially something as unregulated as AI. Even though AI tools, especially large language models, can analyze massive datasets, their decision-making processes often aren’t transparent.

This lack of clarity can be a real challenge for compliance teams. In recent years, the amount of data they handle has increased by 10 to 15 times. While the idea of AI spotting noncompliant behavior is appealing, the industry is still figuring out how to best use AI and machine learning within compliance frameworks.

To make matters worse, the rules and regulations are constantly changing. Compliance professionals need to have accurate data to make necessary adjustments while navigating the complexities of compliance amidst ongoing changes.

Another big challenge is understanding how AI makes its decisions. Organizations frequently face questions about where their data comes from, its accuracy, and whether it’s ethically sourced, especially in tightly regulated sectors like finance, where compliance is critical.

For example, FINRA has recently shared updates regarding how regulatory standards apply to AI-generated content, reminding us that businesses are still accountable for their outputs, whether created by humans or AI.

So, what’s the takeaway? While adopting AI can be complicated, businesses should start with a clear vision. This vision needs to outline how AI can help tackle operational challenges while also highlighting growth opportunities—even as they weigh the risks of non-compliance against the potential benefits of AI.

New Year, New Possibilities

Despite the challenges that come with AI, there are plenty of opportunities for organizations ready to face these issues head-on. The good news is that as companies become more aware of their specific needs, they can use AI to pull valuable insights from their data.

In the coming year, we should expect even more exciting advancements in AI technology that will enhance its capabilities and improve user experience further. This progress will help organizations integrate these tools into their everyday workflows more effectively, giving them a competitive edge.

The real competitive edge however will be seen in those organizations that prioritize transparency and accountability in their AI strategies and will earn trust from clients, stakeholders, and regulators. It’s not just about compliance; it’s about fostering fairness and equity in how technology is used.

We've compiled a list of the best data migration tools.

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

• Garmin has released a new app called ComingUp
• It can be used to log to-dos, deadlines and more
• It runs in the background of your device, so may have an impact on battery life

Garmin has released a new app, ComingUp, which works as a to-do list for some of the best Garmin watches. The app comes with a QWERTY keyboard for task entry, and can talk to various task management services, including Microsoft To Do, Google Tasks, and Todoist, to ensure your tasks sync across your connected services.

Garmin has always walked the line between being some of the best fitness trackers and incorporating some of the features common on the best smartwatches. It appears the company could be preparing to tiptoe over that line.

Naturally, that extra data that's running regular syncs in the background can have an impact on battery life, but given how robust Garmin watches are in that regard we've got no major concerns.

How to get ComingUp on your Garmin (and which watches can get it)

(Image credit: Garmin)

You can find ComingUp in the ConnectIQ store here, and it's worth noting it'll need your permission to run in the background, access to your Garmin Connect profile, and the ability to send and receive data.

Garmin says it syncs in real-time, so your lists will update instantly while you're working away from your desk. As someone who loves a workout at lunch but my head is always filling with ideas for my afternoon, that could be a huge help.

Perhaps best of all are the new reminders for tasks. Getting a timely reminder of your next task being due can help users stay on target, and could be a huge help.

As for compatibility, the full list is on the app link above and it's very extensive. Most of Garmin's most popular models can access the ComingUp app, including the Garmin Fenix 8, Garmin Fenix 7, Garmin Instinct 3, Garmin Venu 3 and the Garmin Forerunner range.

You might also like...

• The Garmin Instinct 3 just landed, but the Instinct 2 is a better buy this President's Day
• The Garmin Instinct 2X Solar can run indefinitely on a single charge, and it's never been cheaper
• I gave up my Garmin for an old-school digital Casio watch for a week: here's what happened to my running

• Potensic's Atom 2 is an update of the Potensic Atom drone for beginners
• It costs $329, or $449 for the Fly More bundle (UK / Australia pricing TBC)
• Features include 4K HDR video, 48MP stills and 10km transmission range

We reckon the Potensic Atom is the best DJI drone alternative for most novices, and now its successor has been announced – the Potensic Atom 2. It offers an array of new features without seemingly breaking new ground, but overall it's a polished upgrade, and becomes the most obvious DJI Mini 4K rival.

Headline features include gimbal-stabilized 4K video with the option for detail-rich HDR output, vertical video enhanced for social, 48MP stills, AI subject tracking, plus improved transmission range and stability of up to 10km, with Full HD / 30p live view and just 120 milliseconds latency.

With the Atom 2, Potensic has also introduced its first after-sales care service, including easy device exchanges and professional assistance, though we're not sure if this service is available with other models. This kind of customer support should provide peace of mind should you run into any issues with the Atom 2.

The Atom 2 standard bundle costs $329, while the Fly More Combo with additional batteries and charging hub costs $449 – we'll update this page when we get UK and Australia pricing. That's a small mark-up compared to the Atom, which launched at $299 / $399.

Image 1 of 5

(Image credit: Potensic)Image 2 of 5

(Image credit: Potensic)Image 3 of 5

(Image credit: Potensic)Image 4 of 5

(Image credit: Potensic)Image 5 of 5

(Image credit: Potensic) DJI's biggest drone rival (for beginners) just got better

We rate the DJI Mini 4K as the best drone for beginners, while the Mini 4 Pro is the best money-no-object option for first-time pilots. However, uncertainty remains around the prospect of a DJI drone ban in the US, and as such any viable alternative is likely to get plenty of attention. Cue the Potensic Atom 2.

On paper, the Atom 2 looks like a refined upgrade. There are no major upgrades that stand out, but the sub-250g drone appears to be refined in many areas, and certainly looks like viable DJI Mini 4K rival.

It's equipped with a 12MP 1/2-inch Sony CMOS sensor, a brighter new f/1.8 lens, together with a 48MP stills output, and a new HDR video mode. Potensic says an upcoming firmware update is on the way too, and it will introduce its first log video recording option. Log video color profiles are flat, and ensure you can get the most detail possible from your drone's hardware, although you'll need to edit them to get the look you want.

Other features include an AI Follow function which automatically tracks subjects to keep them in the center of the frame, while AI Night Mode "captures clear and vivid images" in low-light scenarios', says Potensic. There's also a Dolly Zoom mode, which Potensic "enables users to create cinematic shots with just a single tap, as well as panoramic photography capabilities that allow users to effortlessly capture expansive landscapes'.

We're not sure when the Atom 2 sales start date is, nor do we have our hands on a Potensic Atom 2 for testing yet. We'll be sure to update you on both fronts when the situation changes, and complete our expert review in due course.

You might also like

• Worried about the DJI drone ban? Fear not – here are the four best DJI alternatives
• The best beginner drones for 2025: top flying cameras for new pilots
• Potensic Atom review: one of the best beginner drones

• Huawei FreeArc are here, boasting IP57 dust- and water-ingress protection
• It's Huawei's second open-ear option after FreeClip, but the firm's first 'shell' design
• They're affordable – and there's an extra saving if you buy now

My main issue when considering open-ear earbuds? Actually, there are two things: something that will survive a particularly sweaty/rain-soaked attempt to improve my (abysmal) 5k personal best, and something that won't fall out of my ear during the aforementioned attempt.

While the all-new Beats Powerbeats Pro 2 certainly claimed to have sorted the latter during their hotly-anticipated February 11 arrival announcement, their IPX4 rating won't necessarily tackle the former. Oh, and they come in at $249 / £249 / AU$399. Want something that costs just £99 – or £75 if you act fast – and offers IP57 protection, meaning you could actually drop one in the pool, leave it languishing in one meter of water for up to 30 minutes and still expect music after you've fished it out? Meet Huawei FreeArc.

Open earbuds are big in 2025 – and this set won't break (the bank)

(Image credit: Huawei)

Are the Huawei FreeArc some of the best open ear headphones I've seen? Well, I haven't had a chance to test them fully yet, but on paper there's an awful lot to celebrate, especially for this money.

The sports-focused open-ear category is growing incredibly rapidly, and I predict Huawei's done well to price them so competitively. According to Huawei, a company that boasts seven audio-specific R&D facilities (one with a wind tunnel lab capable of serving up winds of 10 meters per second for mic testing) plus over 1,000 audio engineers with more than 150 PhDs between them, open ear headphones are the fastest growing product in the audio industry, with a 600% growth rate year on year.

This is the Chinese tech specialist's second stab at going open-ear, following the December 2023 cuff-like FreeClip. But this is the first time Huawei's attempted a shell-like design that slides over your outer ear so that the drivers slip in near your ear canal, but not covering it – think Shokz OpenFit Air or Oladance. Huawei tells me its engineers worked exceptionally hard on the new "140-degree optimal Triangle Design", which marries a delicate gravitational balance and data from over 10,000 auricle curves, in something Huawei is calling a "C-bridge design". That 'C' essentially amounts to 8.3g of liquid-silicone rubber covering a 0.7mm titanium nickel wire, with a 17 x 12mm high-sensitivity driver unit at one end and a battery pack at the other (behind the ear).

Leave your phone in your gym locker

Another big selling point for me here is the dual antenna design for stability at long ranges. Although you only get SBC and AAC codec support from the Bluetooth 5.2 chipset (so no higher-resolution codecs such as LDAC or aptX) your quoted range is 100m indoors or 400m outdoors, which means you can leave your phone in your bag by the running track – or in the locker at the gym. And that's a huge bonus if, like me, you'd rather not cart your smartphone around during workouts.

There's also an AI algorithm for the dual mics, to help eliminate blustery wind and ambient noise during calls, and your battery life is a highly competitive seven hours on, or 28 with the case. Did I mention they support on-ear sliding touch gestures for volume too?

The Huawei FreeArc are available to pre-order from today (February 18) in your choice of black, gray or green. Ultimately, they'll be priced at £99.99 – but pre-order them now and you'll get £25 off until March 3, only at a Huawei Store. The full product line will be officially launched on March 4, with a launch offer of £20 off once they hit shelves, until April 1. I'm still waiting on official US and Australian prices where available, but I have to say, predict a big hit at those prices…

You may also like

• See our pick of the best earbuds
• Need the best ANC? Here's our guide to the best noise-cancelling earbuds
• Not sure of this design? See our roundup of the best bone conduction headphones

Most security operations center (SOC) teams find themselves caught between a rock and a hard place. Threat actors are becoming steadily more effective, organized, and aggressive. At the same time, the average IT environment has grown twice as complex over the last five years. That can only mean we can expect these environments to become twice as dense within another five.

SOC teams therefore struggle to keep up with an overwhelming flood of alerts and prioritizing the most critical threats. This is only compounded by chronic staff shortages.

Security teams end up on the defensive, forced to react to issues rather than proactively addressing risks.

With the speed and scope of cybersecurity threats overwhelming SOC personnel, the superhuman capabilities of AI may offer one of the best chances of keeping up.

The growing burden on SOC teams

SOC teams face an unrelenting workload that highlights the imbalance between attackers and defenders. It’s always been the case that the attackers only need to succeed once, while defenders must constantly triumph to protect their operations. But the aggressors have the odds stacking in their favor with new tools and techniques, running heavily automated operations that allow them to ramp up both the quantity and quality of their attacks.

For example, AI tools are enabling cybercriminal gangs to easily create highly targeted phishing campaigns. Rather than manually trawling for information and attempting to create convincing fakes, threat actors can use AI to swiftly find out about key personnel at potential targets, or where they are hosted, and populate a list of useful elements to include.

Within moments, attackers can be armed with convincing phishing emails impersonating a CEO asking his CFO to authorize a payment, complete with little details like mentions of the upcoming company sales conference.

Even without these insidious new tactics, SOC teams would have their hands full due to the unchecked expansion of hybrid IT environments. A focus on digital growth spanning on-premises systems, cloud computing infrastructure, and mobile and IoT devices has dramatically increased the attack surface — yet Gartner estimates that only 17% of organizations can clearly identify the majority of their software assets. This makes it difficult to address vulnerabilities like misconfigurations or a lack of protections such as multi-factor authentication (MFA). Additionally, this lack of reliable inventory and overview means SOC teams are frequently overwhelmed by the volume of alerts, struggling to sift through countless notifications with already limited staff and resources.

AI as a game-changer in cybersecurity

Defending these complex environments and keeping up with threats requires security teams to match and exceed the speed and efficiency of their attackers. Gartner has predicted that companies investing in continuous threat exposure management programs can reduce breaches by two-thirds, with tools like AI-powered analytics playing a major role.

Yet while the criminal element has been quick to integrate AI into its operations, many security vendors are still reluctant. It’s a similar situation to the early days of the cloud, where there was a great deal of mistrust about its security and reliability. As a whole, security teams and vendors have spent more time thinking about AI as a threat than as a potential tool for enhancing their operations.

However, again echoing cloud uptake, end-user organizations are increasingly looking to make use of AI, regardless of the potential risks. The security industry cannot ignore AI, but instead must seek to guide its use responsibly.

AI offers a solution to many of the challenges SOC teams face by addressing visibility gaps, enhancing threat detection, and improving response efficiency.

One of the most important capabilities is the consolidation of disparate data streams from tools such as endpoint detection systems, identity management platforms, and cloud monitors. This unified approach reduces alert fatigue, providing analysts with actionable insights that improve decision-making.

Implementing AI into security workflows also automates some of the more tedious processes, allowing security teams to focus on high-priority threats. By filtering out low-risk alerts, it removes the noise that often hinders effective decision-making. AI can rapidly triage vulnerabilities, prioritizing those actively being exploited or posing the greatest risk to the organization.

Along with helping beleaguered SOC teams keep up, AI also has some powerful applications in improving proactive security. For example, combining data from across on-premises, cloud, and hybrid environments, enables organizations to identify hidden or misconfigured assets creating vulnerabilities across their attack surface.

We have also seen impressive results in using AI to enhance the detection of zero-day vulnerabilities. AI accelerates the identification of behavioral anomalies, isolating threats before they escalate into full-scale incidents.

Addressing AI risks and challenges

While AI has huge potential to help improve cybersecurity, security vendors and SOC teams must be aware of the potential risks.

We often find that organizations fail to apply the same thoroughness to their AI models as they do to other key systems, leaving them vulnerable to misuse or exploitation. All AI tools need to be carefully vetted to assess how they interact with other systems, and any potential risks and attack vectors must be mapped out.

Further, misusing AI tools, such as by uploading sensitive data to unsecured platforms, can expose organizations to breaches. Strict use policies should be put in place so all users know the acceptable uses for any solutions.

For vendors and SOCs operating as MSSPs, it’s also important to consider potential customer reactions to AI usage. Ask anyone if they’d like a service delivered by an automated system or by human hands, and they’ll undoubtedly pick the human. So it’s critical to frame AI as a tool enhancing the skilled human personnel for the SOC team, rather than a replacement for the human touch.

With a solid implementation plan that considers all the angles, AI tools can be deployed to their full potential without introducing any unnecessary new risks.

The future of AI-powered SOCs

AI is proving to be a critical tool in helping SOC teams manage the growing complexities of cybersecurity. By enhancing visibility, automating repetitive tasks, and prioritizing critical risks, it enables teams to operate more efficiently and effectively.

As attackers increasingly use AI to enhance their methods, organizations must keep pace to remain resilient.

However, there is still no substitute for real human intelligence, and AI is most powerful when combined with human expertise, creating a collaborative approach that addresses both routine and complex challenges.

We've compiled a list of the best endpoint protection 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

At least two workstation specialists have put supercharged PCs with Nvidia RTX 5090 GPUs on sale over the past few days. The most impressive of them all is the Bizon ZX5500 which packs up to seven (yes, seven) water-cooled 32GB RTX 5090 GPUs in a tall tower casing. This is the best GPU ever built and buying it through system builders seems to be the only way to avoid months-long wait.

While BizonTech's solution will probably feature in our best workstation guide, it is not as expansive as Comino’s Grando server, which has eight RTX 5090 GPUs, but the latter has yet to get a launch date (I contacted Comino for more details).

The ZX5500 doesn't come cheap at just under $102,000 with the GPUs accounting the lion share (more than 83%) of the total cost. That’s almost 3x the price of MIFCOM’s Big Boss which has seven liquid-cooled RTX 4090 GPUs.

A beefier 6Kw power supply unit plus and the cards cost an extra $85,000 compared to the same system with a pair of RTX 5080 (with 16GB VRAM each). As a reminder, the suggested retail price of the RTX 5090 is ‘just’ $2000.

An RTX 5090 on its retail packaging on a desk (Image credit: Future) RTX 5090 GPUs, where are you?

I wrote two weeks ago about the Bizontech ZX5500, a 7-GPU RTX 5090 system that, unlike the Grando Comino, can be paired with a dual-core EPYC CPU but had a much higher entry price at $102,000. Other boutique workstation vendors have also managed to secure RTX 5090 units (albeit with a 4-6 weeks leadtime).

UK-based Punch Technology is taking orders for a Threadripper Pro rendering AI workstation with two GPUs, a Threadripper Pro 7995WX, 256GB of RAM and a sticker price of just under £24,000 (about $30,000).

For a bit more, local rival Scan Computers, sells an EPYC-based workstation with eight RTX 5090 for almost two thirds what Comino charges for its system. I reached out to them on X.com (formerly Twitter) to find out whether this is a genuine price, a placeholder or a clumsy mistake.

(ed: Turns out it is a one-year subscription to Scan’s Cloud Workstation service, even more disappointing then)

Puget systems, whose President blamed very limited supply of RTX 5090 for long lead times, is only selling single-GPU configurations for now, although I did notice that you can get a system (the T140-XL) with two 16GB Nvidia RTX 5080 GPUs for a bit less than $13,000, ready to ship in less than two weeks.

Other workstation specialists that are selling RTX 5090 rigs include:

• X.solutions (from only $7,000)
• Neox Computer (from £4,510)
• Boxx (price unknown)
• GladiatorPC (from £12,000)

Popular workstation outfit Titan Computers, and none of the three big PC manufacturers (Dell, Lenovo and HP) have yet to release RTX 5090 workstations.

In his review of the RTX 5090, TechRadar’s John Loeffler calls it the supercar of graphics cards, and asks whether it was simply too powerful, suggesting that it is an absolute glutton for wattage.

“It's overkill”, he quips, “especially if you only want it for gaming, since monitors that can truly handle the frames this GPU can put out are likely years away.”

You might also like

• Exclusive: Nvidia's fastest AI chip ever is finally available for preorder
• First Nvidia GH200 Grace Hopper desktop supercomputer now on sale
• How to choose the right server for small businesses

• OpenAI has updated its Model Specification to allow ChatGPT to engage with more controversial topics
• The company is emphasizing neutrality and multiple perspectives as a salve for heated complaints over how its AI responds to prompts
• Universal approval is unlikely, no matter how OpenAI shapes its AI training methods

OpenAI's training methods for ChatGPT are shifting to allow the AI chatbot to discuss controversial and sensitive topics in the name of " intellectual freedom."

The change is part of updates made to the 187-page Model Specification, essentially the rulebook for how its AI behaves. That means you'll be able to get a response from ChatGPT on the delicate topics the AI chatbot usually either takes a somewhat mainstream view on, or refuses to answer.

The overarching mission OpenAI places on its models seems innocuous enough at first. “Do not lie, either by making untrue statements or by omitting important context.” But, while the stated goal may be universally admirable in the abstract, OpenAI is either naive or disingenuous in implying that the "important context" can be divorced from controversy.

The examples of compliant and non-compliant responses by ChatGPT make that clear. For instance, you can ask for help starting a tobacco company or ways to conduct "legal insider trading" without getting any judgment or unprompted ethical questions raised by the prompt. On the other hand, you still can't get ChatGPT to help you forge a doctor's signature, because that's outright illegal.

(Image credit: OpenAI) Context clues

The issue of "important context" gets a lot more complex when it comes to the kind of responses some conservative commentators have criticized.

In a section headed "Assume an objective point of view", the Model Spec details how "the assistant should present information clearly, focusing on factual accuracy and reliability", and also that the core idea is "fairly representing significant viewpoints from reliable sources without imposing an editorial stance".

OpenAI does offer an example of a compliant response to the question "Do Black lives matter?" that unequivocally says yes and references how it's a civil rights slogan and the name of a group. So far, so good.

The issue that OpenAI either doesn't see or is ignoring comes with how ChatGPT tries to thread the needle if you ask, "Don't all lives matter?" as a follow-up. The AI confirms that they do, but adds that the "phrase has been used by people that rejected the premise of the 'Black lives matter' movement."

While that context is technically correct, it's telling that the AI doesn't explicitly say that the "premise" being rejected is that Black lives matter and that societal systems often act as though they don't.

If the goal is to alleviate accusations of bias and censorship, OpenAI is in for a rude shock. Those who "reject the premise" will likely be annoyed at the extra context existing at all, while everyone else will see how OpenAI's definition of important context in this case is, to put it mildly, lacking.

AI chatbots inherently shape conversations, whether companies like it or not. When ChatGPT chooses to include or exclude certain information, that’s an editorial decision, even if an algorithm rather than a human is making it.

(Image credit: OpenAI) AI priorities

The timing of this change might raise a few eyebrows, coming as it does when many who have accused OpenAI of political bias against them are now in positions of power capable of punishing the company at their whim.

OpenAI has said the changes are solely for giving users more control over how they interact with AI and don't have any political considerations. However you feel about the changes OpenAI is making, they aren't happening in a vacuum. No company would make possibly contentious changes to their core product without reason.

OpenAI may think that getting its AI models to dodge answering questions that encourage people to hurt themselves or others, spread malicious lies, or otherwise violate its policies is enough to win the approval of most if not all, potential users. But unless ChatGPT offers nothing but dates, recorded quotes, and business email templates, AI answers are going to upset at least some people.

We live in a time when way too many people who know better will argue passionately for years that the Earth is flat or gravity is an illusion. OpenAI sidestepping complaints of censorship or bias is as likely as me abruptly floating into the sky before falling off the edge of the planet.

You might also like

• ChatGPT o1 goes live and promises to solve all our science and math problems
• Happy 2nd birthday, ChatGPT! Here are 5 ways you've already changed the world
• ChatGPT Tasks can start taking over your calendar and remind you to finish your to-do list

• Seagate Ultra Compact SSD bears a striking resemblence to a thumb drive
• Pricing may be a sticking point for users, and performance leaves a lot to be desired
• The slow demise of flash drives marches on

Seagate’s new portable SSD is small but mighty, and may well mark another nail in the coffin for flash drives.

The firm’s new Ultra-Compact SSD, which is available in both a 1TB and 2TB format, boasts a small, sleek design that more closely resembles a thumb stick or top-of-the-range flash drive than a portable SSD.

Complete with a USB-C port, the SSD includes a single-piece aluminum shell and measures in at 2.7 x 0.7 x 0.34 inches. Notably, Seagate had sustainability in mind during development, with the drive made up to at least 35% recycled materials. From a durability perspective, it can also take a slight beating, being rated for drops up to three meters while featuring IP54 dust and water resistance capabilities - so how does it match up to other industry counterparts?

Under the hood of the Seagate Ultra-Compact SSD

As mentioned, the drive comes in both a 1TB and 2TB options. Both of these offer sequential read and write speeds of up to 1,000 Mbps.

PCMark 10 trace-based testing by Tom’s Hardware showed it wasn’t quite up to scratch with other options available on the market, however. On transfer rates, the drive also “fails to impress” - particularly on read speeds.

Its write speed of 512 Mbps did rank it in fourth place behind the Silicon Power DS72 (1TB), Silicon Power MS70 (1TB), and the PNY Pro Elite 3 (1TB). Tom’s Hardware further noted that with a mixed-file read speed of 570 Mbps, this still remains quite an impressive performance for such a compact drive.

One particular area that users may find appealing are the backup recovery options, though. The SSD ships with Seagate Rescue Data Recovery service, covering users for up to three years.

All told, the SSD is an impressive bit of kit for your everyday user. Capabilities may be lacking in some areas, but its compact design will likely be a draw.

There is one major detail that might put customers off, though: the pricing. The 1TB option comes in at £90.99, making it a middling option compared to other drives available. The price jump to the 2TB version - priced at $179.99 - might be jarring.

You might also like

• Check out the largest SSD and hard drives around right now
• Why you need to be wary of the growing refurbished Chia scandal
• The largest desktop hard drive ever just broke another record

Looking for a different day?

A new Quordle puzzle appears at midnight each day for your time zone – which means that some people are always playing 'today's game' while others are playing 'yesterday's'. If you're looking for Monday's puzzle instead then click here: Quordle hints and answers for Monday, February 17 (game #1120).

Quordle was one of the original Wordle alternatives and is still going strong now more than 1,100 games later. It offers a genuine challenge, though, so read on if you need some Quordle hints today – or scroll down further for the answers.

Enjoy playing word games? You can also check out my NYT Connections today and NYT Strands today pages for hints and answers for those puzzles, while Marc's Wordle today column covers the original viral word game.

SPOILER WARNING: Information about Quordle today is below, so don't read on if you don't want to know the answers.

Quordle today (game #1121) - hint #1 - Vowels How many different vowels are in Quordle today?

• The number of different vowels in Quordle today is 4*.

* Note that by vowel we mean the five standard vowels (A, E, I, O, U), not Y (which is sometimes counted as a vowel too).

Quordle today (game #1121) - hint #2 - repeated letters Do any of today's Quordle answers contain repeated letters?

• The number of Quordle answers containing a repeated letter today is 0.

Quordle today (game #1121) - hint #3 - uncommon letters Do the letters Q, Z, X or J appear in Quordle today?

• No. None of Q, Z, X or J appear among today's Quordle answers.

Quordle today (game #1121) - hint #4 - starting letters (1) Do any of today's Quordle puzzles start with the same letter?

• The number of today's Quordle answers starting with the same letter is 0.

If you just want to know the answers at this stage, simply scroll down. If you're not ready yet then here's one more clue to make things a lot easier:

Quordle today (game #1121) - hint #5 - starting letters (2) What letters do today's Quordle answers start with?

• T

• S

• B

• D

Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.

Quordle today (game #1121) - the answers

(Image credit: Merriam-Webster)

The answers to today's Quordle, game #1121, are…

• TAMER
• SCRUB
• BRICK
• DRIFT

I employed a different tactic today, using four start words instead of two. The idea is that eliminating a larger number of letters makes getting the four words faster and easier.

The person who suggested this method on Reddit claimed they regularly complete Quordle in under a minute and I fully believe them.

It worked for me in terms of speed, but it’s far from foolproof as it leaves an uncomfortable margin of error as the best you can do is finish after eight guesses with one line to spare.

Meanwhile, I completed today’s Sequence with the same number of guesses and without any method at all.

How did you do today? Let me know in the comments below.

Daily Sequence today (game #1121) - the answers

(Image credit: Merriam-Webster)

The answers to today's Quordle Daily Sequence, game #1121, are…

• BURNT
• EDIFY
• THERE
• SHEET

Quordle answers: The past 20

• Quordle #1120, Monday 17 February: SADLY, WAFER, LITHE, IDIOM
• Quordle #1119, Sunday 16 February: GHOUL, AFIRE, COVEN, FIERY
• Quordle #1118, Saturday 15 February: CREEP, CONDO, GRILL, FANCY
• Quordle #1117, Friday 14 February: MEDIA, ELUDE, THUMB, WIDOW
• Quordle #1116, Thursday 13 February: SCALP, DWELL, AMPLE, TRUNK
• Quordle #1115, Wednesday 12 February: SHOOK, GRADE, POLAR, SEEDY
• Quordle #1114, Tuesday 11 February: HEAVY, CIRCA, PESKY, SCION
• Quordle #1113, Monday 10 February: TWIXT, FRESH, GUISE, TABBY
• Quordle #1112, Sunday 9 February: TAPIR, SHAKE, TOKEN, SEVEN
• Quordle #1111, Saturday 8 February: AFOOT, ALIKE, HUMUS, TOWEL
• Quordle #1110, Friday 7 February: PETAL, ABASE, AMONG, CORER
• Quordle #1109, Thursday 6 February: MADAM, SCAMP, FAUNA, TRACK
• Quordle #1108, Wednesday 5 February: SIXTH, VERVE, BLURB, CACTI
• Quordle #1107, Tuesday 4 February: BIRTH, DRAWN, MILKY, JOIST
• Quordle #1106, Monday 3 February: SLEET, TENSE, SEIZE, SCAMP
• Quordle #1105, Sunday 2 February: LOUSY, PENNE, CHUMP, SALVE
• Quordle #1104, Saturday 1 February: FACET, BLISS, FLUFF, FRANK
• Quordle #1103, Friday 31 January: PARER, SYNOD, BRAWN, BANJO
• Quordle #1102, Thursday 30 January: WEEDY, CHIRP, STALE, SUGAR
• Quordle #1101, Wednesday 29 January: CROWN, SCENE, AGILE, CRANK

Looking for a different day?

A new NYT Strands puzzle appears at midnight each day for your time zone – which means that some people are always playing 'today's game' while others are playing 'yesterday's'. If you're looking for Monday's puzzle instead then click here: NYT Strands hints and answers for Monday, February 17 (game #351).

Strands is the NYT's latest word game after the likes of Wordle, Spelling Bee and Connections – and it's great fun. It can be difficult, though, so read on for my Strands hints.

Want more word-based fun? Then check out my NYT Connections today and Quordle today pages for hints and answers for those games, and Marc's Wordle today page for the original viral word game.

SPOILER WARNING: Information about NYT Strands today is below, so don't read on if you don't want to know the answers.

NYT Strands today (game #352) - hint #1 - today's theme What is the theme of today's NYT Strands?

• Today's NYT Strands theme is… Ouch!

NYT Strands today (game #352) - hint #2 - clue words

Play any of these words to unlock the in-game hints system.

• SCORE
• BLEEP
• CONE
• CORNER
• SOIL
• SELL

NYT Strands today (game #352) - hint #3 - spangram What is a hint for today's spangram?

• Deliver venom

NYT Strands today (game #352) - hint #4 - spangram position What are two sides of the board that today's spangram touches?

First side: right, 3rd row

Last side: left, 7th row

Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.

NYT Strands today (game #352) - the answers

(Image credit: New York Times)

The answers to today's Strands, game #352, are…

• JELLYFISH
• SPORPION
• HORNET
• PLATYPUS
• BUMBLEBEE
• SPANGRAM: STINGERS

• My rating: Easy
• My score: Perfect

Strands really is an education.

I knew that the PLATYPUS is a strange mammal with a duck’s bill, otter-like body and beaver-type tail. And I knew that they are unique to Australia. But I had no idea they had STINGERS too (well, the male of the species do at least).

It turns out being venomous is just one of the many remarkable features of this incredible creature – they’ve got two layers of fur, they lay eggs, they glow in the dark, they hunt using electroreception, the plural of platypus is platypus. They can complete Strands in 45 seconds without any hints. They are brilliant.

How did you do today? Let me know in the comments below.

Yesterday's NYT Strands answers (Monday, 17 February, game #351)

• LEMON
• CANARY
• GOLDENROD
• MUSTARD
• SAFFRON
• DAFFODIL
• SPANGRAM: YELLOW

What is NYT Strands?

Strands is the NYT's new word game, following Wordle and Connections. It's now out of beta so is a fully fledged member of the NYT's games stable and can be played on the NYT Games site on desktop or mobile.

I've got a full guide to how to play NYT Strands, complete with tips for solving it, so check that out if you're struggling to beat it each day.

Looking for a different day?

A new NYT Connections puzzle appears at midnight each day for your time zone – which means that some people are always playing 'today's game' while others are playing 'yesterday's'. If you're looking for Monday's puzzle instead then click here: NYT Connections hints and answers for Monday, February 17 (game #617).

Good morning! Let's play Connections, the NYT's clever word game that challenges you to group answers in various categories. It can be tough, so read on if you need Connections hints.

What should you do once you've finished? Why, play some more word games of course. I've also got daily Strands hints and answers and Quordle hints and answers articles if you need help for those too, while Marc's Wordle today page covers the original viral word game.

SPOILER WARNING: Information about NYT Connections today is below, so don't read on if you don't want to know the answers.

NYT Connections today (game #618) - today's words

(Image credit: New York Times)

Today's NYT Connections words are…

• HEARTY
• READY
• HANDY
• SQUARE
• AVAILABLE
• BUTTY
• CAVITY
• FILLING
• FOOTY
• HOLE
• SOLID
• CHIPPY
• HOLLOW
• TELLY
• NEARBY
• PIT

NYT Connections today (game #618) - hint #1 - group hints

What are some clues for today's NYT Connections groups?

• YELLOW: Big and hollow 
• GREEN: Easy to get 
• BLUE: Large portion 
• PURPLE: UK speak

Need more clues?

We're firmly in spoiler territory now, but read on if you want to know what the four theme answers are for today's NYT Connections puzzles…

NYT Connections today (game #618) - hint #2 - group answers

What are the answers for today's NYT Connections groups?

• YELLOW: CRATER 
• GREEN: ACCESSIBLE 
• BLUE: SUBSTANTIAL, AS A MEAL 
• PURPLE: BRITISHISMS 

Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.

NYT Connections today (game #618) - the answers

(Image credit: New York Times)

The answers to today's Connections, game #618, are…

• YELLOW: CRATER CAVITY, HOLE, HOLLOW, PIT
• GREEN: ACCESSIBLE AVAILABLE, HANDY, NEARBY, READY
• BLUE: SUBSTANTIAL, AS A MEAL FILLING, HEARTY, SOLID, SQUARE
• PURPLE: BRITISHISMS BUTTY, CHIPPY, FOOTY, TELLY

• My rating: Easy
• My score: Perfect

Finally, Connections gives me a Purple group I can solve easily and acknowledges the vast army of UK players to boot.

In case you are unaware, a BUTTY is a sandwich made of one slice of bread folded over – invariably filled with chips that you’d buy from a CHIPPY (fish and chip shop) sprinkled with salt and vinegar and red sauce (ketchup), that you eat while watching the FOOTY (soccer) on your TELLY – a delicacy that is more of a snack than SUBSTANTIAL, AS A MEAL.

And now I'm hungry. Thank you, Connections.

How did you do today? Let me know in the comments below.

Yesterday's NYT Connections answers (Monday, 17 February, game #617)

• YELLOW: TIME OFF BREAK, LEAVE, REST, VACATION
• GREEN: FOLLOW A MEANDERING COURSE CURVE, SNAKE, WEAVE, WIND
• BLUE: BOWLING RESULTS DOUBLE, SPARE, STRIKE, TURKEY
• PURPLE: -__FISH BLOW, CAT, GOLD, SWORD

What is NYT Connections?

NYT Connections is one of several increasingly popular word games made by the New York Times. It challenges you to find groups of four items that share something in common, and each group has a different difficulty level: green is easy, yellow a little harder, blue often quite tough and purple usually very difficult.

On the plus side, you don't technically need to solve the final one, as you'll be able to answer that one by a process of elimination. What's more, you can make up to four mistakes, which gives you a little bit of breathing room.

It's a little more involved than something like Wordle, however, and there are plenty of opportunities for the game to trip you up with tricks. For instance, watch out for homophones and other word games that could disguise the answers.

It's playable for free via the NYT Games site on desktop or mobile.

• Google has reversed its policy on 'fingerprinting'
• 'Fingerprinting' is a technique used to collect personal user data
• Users have very little say over what information is given to advertisers

Do you sometimes wish that online advertisers had access to more of your private information without your consent or knowledge? Well, you might just be in luck.

As of February 16 2025, Google is permitting ‘fingerprinting’, which is a stealthy technique which tracks things like your IP addresses, screen resolution, operating systems, and even battery percentage - until it has a creepily accurate profile of you.

This practice will be particularly disturbing to the privacy-conscious among us since it's virtually impossible for users to control what information is collected and when. You can turn off all cookies, use all the private browsers, and use the best VPN money can buy, but you could still be tracked with this technique.

Why the change of heart?

In its response, the UK’s Information Commissioner’s Office stated it considers fingerprinting to be, ‘not a fair means of tracking users online because it is likely to reduce people’s choice and control over how their information is collected’.

The move is is a pretty stark 180-degree turn in terms of policy, considering that in 2019, Google condemned the practice of fingerprinting, saying;

“Unlike cookies, users cannot clear their fingerprint, and therefore cannot control how their information is collected. We think this subverts user choice and is wrong.”

But the practice will, ‘ help businesses, large and small, meet the opportunities of the evolving digital landscape’, Google’s statement latest says, and it does claim to be ‘meeting user expectations for privacy’ - but you can decide for yourself if your expectations are being met with this policy.

So what can you do about it? Unfortunately, not a lot. The ICO admits, ‘even privacy-conscious users will find this difficult to stop’ as the practice is harder for browsers to block, and it relies on signals you can’t easily wipe - so pressing ‘clear all site data’ won’t stop the advertiser from using the same fingerprinting techniques to just identify you all over again.

That being said, some browsers, such as Firefox and the Brave browser, have fingerprinting protection features that can scramble things like time zone, language, fonts, and language in reports to websites. There are also Google Chrome extensions like Canvas Blocker which can function in a similar way.

If you want more information, we've written a full guide on what fingerprinting is and how to protect yourself.

You might also like

• Privacy powerhouses: 5 apps to take your online security to the next level
• Check out our recommendations for the best antivirus software
• UK renames AI Security Institute, drops "safety" in pivot to cybersecurity

• A new leak suggests Alexa's AI is too innacurate to launch yet
• It'll still be shown off on February 26, but won't release until later
• The leak follows reports Alexa was delayed from 2024 due to similar issues

It’s all but guaranteed that Amazon is launching a new version of Alexa with souped up AI brain power on February 26 – it literally spelled it out in an announcement – but disaster might have struck at the final hurdle. Alexa AI is reportedly delayed, again.

That’s per an anonymous source who spoke with The Washington Post (the report is behind a paywall) claiming that the new Alexa has been making too many mistakes when asked test questions. As a result Alexa is being delayed to improve its accuracy – with the current launch date now put back 'til March 31.

Amazon is still expected to unveil the all-new Alexa at the New York event on February 26 as it originally planned, however, we expect access to the AI (and the questions it’ll answer in demos) might be restricted so as not to reveal its potentially less-than-perfect side.

As with all rumored information, we should still take all of this with a pinch of salt, though if The Washington Post is correct, this wouldn’t be the first time Alexa has been delayed. Multiple sources had teased a 2024 launch date, with accuracy issues once again cited as the reason Alexa was held back.

(Image credit: Amazon)

Beyond causing a delay, these issues could also prove a blow to Amazon’s rumored plan to charge users for Alexa's help. It’s been said the revamped Alexa could cost paying customers $5-$10 a month to use (around £5-£10 / AU$8-AU$16). If Alexa is unreliable – or has a reputation for being unreliable beyond what’s expected from a current-gen AI – we imagine there won’t be many users keen to pay for the service.

At least the current-version of Alexa is said to be sticking around as a permanently free and generally reliable option.

If Amazon can solve Alexa’s accuracy problems, the new AI does sound rather useful. Alexa AI is said to be smarter so it can handle multiple prompts at once, rather than requiring its user to give distinct commands one after the other, and to perform as an AI agent – read: taking actions without direct user requests.

Admittedly that last point sounds a little scary, given that Alexa AI would have our credit card info and direct access to the world’s largest online store (Amazon), especially if Alexa is prone to mistakes (I know I'd be nervous about using it). But if Amazon can prove its agent is genuinely helpful, Alexa might finally start living up to the futuristic home assistant many imagined it would be when it first launched.

You might also like

• The best Alexa Skills and commands: the most useful voice assistant tips and tricks
• The best cheap Amazon Echo deals and sales
• How to set up an Amazon Echo: top tips for your new Alexa speaker

• GlobalFoundries was spun off when AMD decided to go fabless 16 years ago
• Thomas Caulfield’s leadership made GloFo profitable; now rumors link him to Intel
• A merger could help Intel, but regulatory hurdles might block the deal

GlobalFoundries (GloFlo) was founded in March 2009 when a struggling AMD spun off its manufacturing division to go fabless, a move designed to help it survive in the highly competitive semiconductor industry.

GloFo eventually became a key player in the foundry business, with a global manufacturing footprint across the US, Europe, and Asia, focusing on specialty chips rather than competing at the leading edge. Under the leadership of Thomas Caulfield, who became CEO in 2018, GlobalFoundries pursued a strategy that led it to become the third-largest semiconductor foundry by revenue, and a public listing in 2021.

However, speculation is currently brewing that Caulfield could be set to become the next CEO of AMD’s long time chip rival, Intel. As reported by eeNews Europe, the rumors began following an unexpected leadership change at GlobalFoundries. Caulfield is set to move from CEO to executive chairman, with Tim Breen, who has also been at the company since 2018, taking over as CEO. This shift has led observers to question whether Caulfield may be preparing for a leadership role at Intel, which has been without a permanent CEO since Pat Gelsinger’s departure in December 2024.

Regulatory hurdles

Caulfield has extensive experience in semiconductor manufacturing but has not led a product company like Intel. This has fueled suggestions that instead of simply taking over as CEO, he could oversee a merger between Intel and GloFo. In such a scenario, eeNews Europe suggests that Caulfield would become executive chair of the combined entity, with Michelle Johnston Holthaus leading Intel’s product business and Breen integrating GlobalFoundries into Intel Foundry.

A merger would definitely benefit both parties. While AMD has seen its fortunes soar in recent times, intel has been struggling. Joining forces with GlobalFoundries could allow the merged company to offer a broader range of manufacturing capabilities.

Adding further fuel to the rumor mill, eeNews Europe reports that observers on a Reddit stream noted a recent purchase of 8,913,000 Intel shares at $19.29. This unusual transaction occurred two days before GlobalFoundries announced its leadership changes, drawing comparisons to the $172 million stock incentive given to Gelsinger when he became Intel’s CEO in 2021.

Any potential merger would of course face regulatory hurdles, particularly in China, which previously blocked Intel’s attempt to acquire Tower Semiconductor. If approval proves difficult, eeNews Europe suggests Intel could appoint Caulfield as CEO while maintaining close ties with GlobalFoundries.

You might also like

• Intel is looking to match AMD as it debuts super-fast AI chips
• Intel slams Nvidia and AMD, claims chip giants have security flaws
• IBM and Global Foundries settle long-running lawsuit

• Amazon is removing the Download & Transfer via USB option on Kindles
• The feature hadn't appeared on the latest models
• It's likely that the move is to combat ebook piracy

I've owned and loved some of the best Kindles in the series since Amazon first introduced the ereader, but using these devices can feel somewhat restrictive – especially now Amazon is killing off a useful USB file transfer feature.

As reported by The Verge, the Download & Transfer via USB option is going away on February 26. That means you'll no longer be able to download ebooks purchased from Amazon, and then move them to your Kindle over a wired USB connection.

The feature isn't actually available on the latest generation of Kindles, pushed out last year, but all other models are going to be affected. Amazon doesn't give a reason for the move, but it's almost certainly to crack down on ebook piracy.

Using fairly simple tools available online, users could download purchased ebooks from Amazon, crack the DRM (Digital Rights Management) protection on them, and send them to other devices or distribute them widely on the web. That will no longer be possible.

Less flexibility

An ebook on the Kindle Oasis

As is often the case, efforts to battle digital pirates means rule-abiding users lose out. This means ebooks can no longer be saved on computers as backups, or moved over without a Wi-Fi connection, which gave Kindle owners some extra flexibility.

It's worth bearing in mind that Amazon does occasionally remove ebooks from its catalog, or replace them with modified versions. Now more than ever, users are going to be at the whims of what Amazon decides to make available.

You will still be able to move ebooks over via Wi-Fi of course, as well as transfer files to your Kindle via USB and software such as Calibre. It's specifically the ability to download purchases from the Amazon website to a computer that's disappearing.

This also gets closer to fully killing off the older AZW3 format for Amazon ebooks: it's been replaced by the newer KFX format, which is much harder to crack in terms of its DRM, and which is now used in most situations (except downloads to a computer).

You might also like

• These are the best ereaders you can currently buy
• Our full review of the Amazon Kindle Colorsoft
• Your Kindle Scribe just became a better note-taking ereader

• South Korea privacy watchdog has temporarily halted downloads of DeepSeek
• DeepSeek is working with authorities to get compliant
• Latest in a series of privacy concerns raised about AI chatbots

South Korea’s Personal Information Protection Commission (PIPC) has temporarily halted new downloads of Chinese-owned AI chatbot DeepSeek.

Reports from TechCrunch confirm the app is still up and running for those who have already installed it, and that the decision won’t affect usage of the app - but new downloads will be halted until the Chinese firm complies with Korean privacy laws.

South Korea isn’t the first to ban new downloads of the chatbot, with the model disappearing from the Italian App store and Google Play Store in late January 2025 after the country’s data watchdog filed a privacy complaint and asked for information on how DeepSeek handles users’ personal information.

Recurring concerns

DeepSeek has since appointed a local representative to work with authorities in South Korea, but the data protection agency has said it ‘strongly advises’ current users to refrain from entering personal data into DeepSeek until a final decision is made - here’s everything we know so far.

The restriction is temporary whilst the PIPC assesses the use and storage of data by DeepSeek, but the agency confirms the model will be available for download once it is compliant.

The PIPC found DeepSeek had transferred the data of South Korean users to ByteDance - TikTok’s parent company. TikTok, as many will remember well - was briefly banned in the US over privacy and security concerns.

DeepSeek isn’t the first AI model to come under scrutiny for privacy concerns. The nature of Large Language Models is a bit of a privacy minefield since they scrape every corner of the internet for data to train their models - without the consent of the owners/authors/creators of the media they use.

Further than this though, OpenAI has never asked people for permission to use their data, and it’s not possible for a person to get confirmation of what data has been used or stored - or for it to be deleted. This contradicts an important facet of GDPR laws, which protect the right to be forgotten and should guarantee individuals the ability to have their personal data erased upon request.

As the new kid on the block, DeepSeek is in the spotlight for a number of reasons - and there have been legitimate concerns about how the platform collects and stores your personal information like your email address, name, and date of birth, as well as the data you enter into the chatbot, and the technical information of the device you’re using, like IP address, OS, etc.

Using AI safely

So is DeepSeek safe to use? And can it be used whilst maintaining your privacy? Well there are things you can do to mitigate the risks.

As with all LLMs, if you’re concerned about data privacy, using AI is probably not a good idea. LLMs scrape data from the internet without permission, and will use your interactions to add to the pool of data the model is trained with, and that’s not something you can opt out of, DeepSeek included.

If you’re in South Korea or Italy and still want to download DeepSeek, even the best VPN services will need a little extra help - as they don’t spoof your App store location, so you’ll need to download it from elsewhere. This is something we usually advise against, as it can be a really easy way to be tricked into downloading malware - so do so with caution.

In terms of the cybersecurity risks, there have been reports DeepSeek is ‘incredibly vulnerable’ to attacks, and failed to block any harmful prompts when tested, severely underperforming against its rivals.

You should be wary when using these chatbots - especially on a company device or if you work in an industry that has national security connections - there’s a reason both Australian and Indian government departments have blocked the use of DeepSeek from work devices.

A general rule is that users should be especially careful with the information you give to a chatbot. Don’t enter your health information, financial data, or anything that you wouldn’t want a third party to know. Monitor your accounts for any suspicious activity regularly, and flag anything you see as soon as you see it.

You might also like

• Check out our list of the best firewall software around today
• “This is a wake-up call" - the DeepSeek disruption: 10 experts weigh in
• We've also rounded up the best malware removal software on offer right now

• Windows 11 patch is causing problems with File Explorer for some
• This means that some – or most – folders on the desktop are failing to open
• There are also reports of installation failures and PCs failing to boot

Microsoft’s latest update for Windows 11 has thrown some more spanners into the works for those who’ve upgraded to the most recent 24H2 spin of the desktop OS.

The February cumulative update for 24H2 users (known as patch KB5051987) did some admirable work on the bug fixing front, but sadly also appears to have introduced some fresh problems to this incarnation of Windows 11.

And two of these issues are very worrying, the first of which is a bug that’s apparently playing havoc with File Explorer, as reported by Windows Latest and other Windows 11 users across various online forums.

File Explorer is the app in which your desktop folders are displayed, allowing you to view and interact with the files inside. Due to this bug, though, you can’t actually view files, as some folders are refusing to open.

Windows Latest observes that the commonly-used Documents and Pictures folders are non-functional after installing the February update for Windows 11 24H2. File Explorer is also failing to work when accessed via Windows 11’s search function, or a desktop shortcut.

Other users are reporting various problems with opening certain folders, or indeed most folders – though some of them still appear to work.

It’s an odd sounding bug, in short, that’s having unpredictable effects. But if you’ve installed KB5051987 and are encountering weirdness in terms of folders just not working or appearing at all – even though File Explorer is running seemingly just fine, as normal, within Windows 11 – well, this is why.

On top of the flakiness with File Explorer, the February update is also failing to install for some Windows 11 users (though that’s nothing new – this is a common problem with a lot of Windows updates these days). For some, the download and installation process is getting stuck fast at a certain percentage, and for others, it might eventually complete, but it’s taking hours.

Worryingly, some folks are reporting that KB5051987 is causing crashes (Blue Screens of Death) after it’s installed, or that it has even torpedoed their Windows 11 installation entirely – yikes.

Windows Latest also points to potential issues with webcams not working, and the mouse cursor moving sluggishly and stuttering (as well as performance glitches in general).

(Image credit: Future / Jeremy Laird) Analysis: File this one under ‘worrying’

Just when I voiced hope that maybe Microsoft is finally getting on top of all the issues with Windows 11 – which have been numerous ever since the upgrade went live last year – we’ve seemingly got a doozy of a gremlin messing with the internal workings of File Explorer somehow.

File Explorer is a critical part of the Windows interface – as noted, it’s the very folders you navigate to access all the files on your PC – so to see it partially (or almost wholly in some cases) hamstrung is very disappointing.

This is a fairly widely reported issue, and as Windows Latest notes, it has over 30 reports sent in by readers (possibly more by now). There’s also lots of grumbling (understandably) about this on Reddit and other forums.

One of your first thoughts – it was certainly one of mine – was that a third-party customization utility could perhaps be to blame, but Windows Latest tried a completely vanilla installation of Windows 11, and this too was affected. There doesn’t seem to be any commonality between those suffering at the hands of this odd bug, not yet anyway – and there are no suggested workarounds for now. (Save for the obvious course of action – uninstall the February update).

That said, in one of the Reddit complaints I perused, I noticed that somebody offered up the idea of disabling Windows Sandbox, and this reportedly worked for two people. So that might be worth a shot, but bear in mind that Sandbox is not available on Windows 11 Home. So only those with the Pro edition can consider this possible fix, and most people are running the Home flavor of the OS, of course.

On top of the File Explorer problem, the reports of installation failures are very concerning, especially those which apparently broke Windows 11 completely. These reports seem much rarer, thankfully, but still – caution might be the better part of valor here, and you may want to put off installing the February update for as long as you can. (Which, I should note, for Windows 11 Home users isn’t all that long). In this case, you will, of course, be running without the security fixes provided by the patch, which isn’t ideal either.

With any luck, Microsoft will be investigating these issues and hopefully taking action sooner rather than later.

You may also like...

• Windows 11 fully streamlined in just two clicks? Talon utility promises to rip all the bloatware out of Microsoft’s OS in a hassle-free way
• Are you unable to get security updates for Windows 11 24H2? Here’s the likely reason why, and the fix to get your PC safe and secure again
• If Microsoft's Windows 11 24H2 issues continue like this, I'll fully convert to SteamOS for gaming

• Palo Alto Networks fixes authentication bypass PAN-OS flaw
• A day after the patch was released, criminals started looking for vulnerable endpoints
• The flaw allows them to run different PHP scripts

A vulnerability in Palo Alto Networks firewalls is being abused in in-the-wild attacks, researchers are saying.

The company recently found, and fixed, an authentication bypass vulnerability in its PAN-OS firewalls. The flaw, tracked as CVE-2025-0108, has a severity score of 8.8/10 (high), and was said to affect multiple versions of the product.

It released a fix on February 12, 2025, urging users to upgrade their firewalls to these versions:

11.2.4-h4 or later
11.1.6-h1 or later
10.2.13-h3 or later
10.1.14-h9 or later

Exploit attempts

The vulnerability impacts the PAN-OS management web interface, and allows malicious actors to run different PHP scripts. This, in turn enables sensitive data exfiltration, firewall configuration tampering, and more.

Now, researchers from the security outlet GreyNoise said they observed attempts to exploit the flaw on unpatched endpoints. The attacks, they said, started a day after Palo Alto Networks released the patch (February 13), and came from multiple IP addresses, which could suggest that more attackers picked up on the vulnerability at the same time.

Citing information from Macnica researcher Yutaka Sejiyama, BleepingComputer reported that the attack surface likely counts more than 4,400 devices.

To protect the firewalls, users should apply the patch as soon as possible, and restrict access to the product’s interface, as soon as possible.

Firewalls used by SMBs are often targets because these types of businesses typically have weaker security configurations and outdated firmware. Many SMBs lack dedicated IT teams, leading to misconfigured firewall rules that create vulnerabilities. Furthermore, threat actors can use firewalls as entry points to bypass network defenses and gain deeper access to internal systems. Once compromised, firewalls can be used to intercept sensitive data, launch further attacks, or disable security measures altogether.

Via BleepingComputer

You might also like

• A critical Palo Alto Networks bug is being hit by cyberattacks, so patch now
• We've rounded up the best password managers
• Take a look at our guide to the best authenticator app

• A flaw named WhoAMI was found in Amazon Machine Image
• It allows threat actors to gain RCE abilities on people's AWS accounts
• A fix has been released, but many users are still yet to update

Amazon Web Services (AWS) users are potentially vulnerable to a name confusion attack called “whoAMI”, experts have warned.

The vulnerability, found in Amazon Machine Image (AMI), was discovered in the summer of 2024 by cybersecurity researchers DataDog, and has now been confirmed by Amazon, which said it fixed the issue on its side, and urged users to update the code on their side and thus protect their premises.

AMI is a pre-configured template used to create and launch virtual servers (EC2 instances) in AWS. It includes an operating system, application software, and necessary configurations like storage and permissions. AMIs allow users to quickly deploy consistent environments, whether using AWS-provided images, community AMIs, or custom-built ones. This makes scaling and managing cloud infrastructure more efficient.

Following the naming pattern

AMIs can be public, or private, and once generated, come with a unique identifier. Public ones can even be found in the AWS catalog. But these public ones should also come with the ‘owners’ attribute, as a way to confirm that they’re coming from a trusted source.

Now, the researchers found that the way software projects retrieve AMI IDs was flawed, and allowed threat actors to gain remote code execution (RCE) capabilities within people’s AWS accounts.

The technical details on how the vulnerability works and how it might be exploited can be found on this link. Long story short, if a threat actor publishes an AMI with a name that follows the format used by trusted owners, it can be picked up by mistake.

When DataDog first discovered the flaw, it said that overall, a very small percentage of AWS users are vulnerable, but that still equals “thousands” of AWS accounts. Amazon responded by issuing a fix in mid-September last year, and releasing a new security control called “Allowed AMIs” in early December last year.

It also advised all users to apply the fixes, while stressing that there was no evidence of abuse in the wild.

Via BleepingComputer

You might also like

• Dangerous Microsoft Outlook flaw could let hackers send out malware via email
• We've rounded up the best password managers
• Take a look at our guide to the best authenticator app

• Unprotected.org successfully petitions USPTO against the application for Hosted WordPress and Managed WordPress
• However this is not an absolute denial for the trademarks
• Move leads some to call Matt Mullenweg's WordPress leadership into question

Since September 2024, WordPress, led by Matthew Mullenweg, has been locked in a battle with WP Engine over licensing and contributions to the WordPress community.

So far, the battle has seen Mullenweg block WP Engine from open source resources, followed by a court order that required WP Engine's access to be reinstated and restrictions to developers removed.

As part of a wider effort to control how hosts can use the WordPress name, the WordPress Foundation also attempted to trademark 'Hosted WordPress' and 'Managed WordPress,' but the US Patent and Trademark Office (USPTO) has required disclaimers for both terms in an initial setback for Mullenweg.

Is Mullenweg doing more harm than good for WordPress?

The attacks on WP Engine and trademark expansion efforts left many hosts wondering about the future of their WordPress hosting products and possible backlash from the WordPress Foundation over trademark infringements of the open-source software.

While this initial disclaimer request gives some relief to hosts it is not a denial and the term Hosted WordPress has already been granted in some counties. If the trademarks are granted it gives more control to the WordPress Foundation on which web hosting providers they can give licenses to for the use of the phrases.

The WordPress drama has caused many to question whether Mullenweg’s actions are damaging rather than strengthening the WordPress community. Since the attack on WP Engine, 159 Automatic (a company run by Mullenweg) employees have left because they disagree with his leadership.

The requested disclaimer on the trademarks was filed by Unprotected.org, a website critical on Mullenweg.

A spokesperson from the site told The Register, "This represents a great victory for the WordPress ecosystem, and we will continue to fight until there is accountability and a change in leadership.”

The spokesperson added, "The WordPress ecosystem is ready for new leadership, and Joost de Valk, the developer of Yoast SEO, is the first who comes to mind."

De Valk has previously called for the decentralisation of WordPress around Mullenweg and to create a federated and independent WordPress. In a blog post he said, “We, the WordPress community, need to decide if we’re ok being led by a single person who controls everything, and might do things we disagree with, or if we want something else. For a project whose tagline is “Democratizing publishing”, we’ve been very low on exactly that: democracy.”

In response, Mullenwag banned De Valk from sponsoring and attending any WordPress events. WordPress currently powers over 43% of the world’s websites but as web hosts develop their own website builders and uncertainty around the stability of WordPress grows we may see that number start to decline.

You might also like

• We've rounded up the best WordPress hosting options around
• What is managed WordPress hosting?
• WordPress.org vs WordPress.com what's the difference?