The United States military is among the largest and most technologically sophisticated in the world. Needless to say, it has access to some extraordinary equipment, with the U.S. Air Force operating some of the fastest fighter jets in service today. These are aircraft that most civilians will never ever come close to flying, but by the same token, there is some civilian technology that Air Force personnel are not permitted to use while in uniform.
Smart glasses are becoming increasingly popular, with International Data Corporation noting in July 2025 that the second-generation Meta Ray-Ban glasses captured nearly two-thirds of the market after a successful end to 2024. The broader concept of smart glasses, however, touches on significant fears about privacy and security, meaning that there’s a smart reason not to wear models such as Meta’s AI glasses. It’s also for this reason that the Air Force has prohibited the wearing of such devices. On January 9, 2026, the Secretary of the Air Force Public Affairs issued new guidance targeting “dress and personal appearance.” The policy noted that, with regard to traditional eyewear, specific shades of frames and lenses for sunglasses and eyeglasses must be adhered to, while “It is unauthorized to wear mirrored lenses or smart glasses with photo, video or artificial intelligence capabilities while in uniform.”
Elsewhere in this memo, the iconic duty-identifier patches were brought back, powerful symbols of pride in a shared role and expertise. This same collective spirit, however, means that any Air Force member could endanger their fellow members through the use of smart glasses, which is why they were banned.
Advertisement
The potential issue with smart glasses for the Air Force
tete_escape/Shutterstock
If there’s one term that’s pivotal to U.S. Air Force operations, it’s OPSEC: Operations security. There’s a wealth of information, from troop movements to the specs of particular vehicles or weapons, that could be potentially disastrous if it leaked. When smart glasses are in the picture, such a leak could occur accidentally, or even without the responsible parties knowing it happened.
This is the argument put forward by Dana Thayer, the Information Protection Chief of the 104th Fighter Wing, in a statement shared by dvids in response to this prohibition of the devices. Thayer warns that such devices can and will continue to record even when the wearer hasn’t specifically set them to do so, which Thayer equates to receiving a plethora of pop-up ads about a specific product that you haven’t expressly searched for online but only discussed. Imagine the conversation is about a top-secret new military operation rather than a fast-food craving, and the potential gravity of wearing smart glasses in uniform becomes obvious.
Advertisement
As is the case essentially anywhere AI and similar technologies are in use, there are significant advantages to be gained. Smart glasses may even serve as a viable alternative to computer monitors, for instance, increasing versatility in work environments. It’s vital to be aware of the dangers too. The U.S. Army and Navy have not yet instituted such a blanket measure as an outright ban on in-uniform use of smart glasses, but perhaps they will. It’s an evolving technology and an evolving situation, and the devices may find a place where their utility can be deployed, even in the most sensitive situations.
Mikko Hyppönen is pacing back and forth on the stage, with his trademark dark blonde ponytail resting on an impeccable teal suit. A seasoned speaker, he is trying to make an important point to a room full of fellow hackers and security researchers at one of the industry’s global annual meet-ups.
“I often call this ‘cybersecurity Tetris’,” he tells the audience with a serious face, reeling off the rules of the classic video game. When you complete a whole line of bricks, the row vanishes, leaving the rest of the bricks to fall into a new line.
“So your successes disappear, while your failures pile up,” he tells the audience during his keynote at Black Hat in Las Vegas in 2025. “The challenge we face as cybersecurity people is that our work is invisible… when you do your job perfectly, the end result is that nothing happens.”
Hyppönen’s work, however, has certainly not been invisible. As one of the industry’s longest serving cybersecurity figures, he has spent more than 35 years fighting malware. When he started in the late 1980s, the term “malware” was still far from everyday parlance; the terms instead were computer “virus” or “trojans.” The internet was still something few people had access to, and some viruses relied on infecting computers with floppy disks.
Advertisement
Since then, Hyppönen estimated he has analyzed thousands of different kinds of malware. And thanks to his frequent talks at conferences all over the world, he has become one of the most recognizable faces and respected voices of the cybersecurity community.
While Hyppönen has spent much of his life trying to keep malware from getting into places it is not supposed to, now he is still doing much of the same, albeit a slightly different tack: His new challenge is to protect people against drones.
Hyppönen, who is Finnish, told me during a recent interview that he lives about two hours away from Finland’s border with Russia. An increasingly hostile Russia and its 2022 full-scale invasion of Ukraine, where the majority of deaths have reportedly come from unmanned aerial attacks, have made Hyppönen believe he can have renewed impact by fighting drones.
For Hyppönen, it is also a matter of recognizing that while there are still long-standing problems to solve in the world of cybersecurity — malware is not going anywhere and there are plenty of new problems on the horizon — the industry has made huge strides over the last two decades. An iPhone, Hyppönen brought up as an example, is an extremely secure device. The cybersecurity aspects of drone warfare, on the other hand, remain almost uncharted territory.
Advertisement
Image Credits:courtesy of Mikko Hypponen
From viruses and worms to malware and spyware…
Hyppönen started early in cybersecurity by hacking video games during the 1980s. His love for cybersecurity came from reverse engineering software to figure out a way to remove anti-piracy protections from a Commodore 64 games console. He learned to code by developing adventure games, and sharpened his reverse engineering skills by analyzing malware at his first job at Finnish company Data Fellows, which later became the well-known antivirus maker F-Secure.
Since then, Hyppönen has been on the front lines of the fight against malware, witnessing how it evolved.
In the early years, virus writers developed their malicious code often exclusively out of passion and curiosity to see what was possible with code alone. While some cyberespionage existed, hackers had yet to discover ways to monetize hacking by today’s standards, like ransomware attacks. There was no cryptocurrency to facilitate extortion, nor a criminal marketplace for stolen data.
Form.A, for example, was one of the most common viruses in the early 1990s, which infected computers with a floppy disk. A version of that virus did not destroy anything — sometimes just displaying a message on the person’s screen, and that was it. But the virus travelled around the world, including landing on the research stations at the South Pole, Hyppönen told me.
Hyppönen recounted the infamous ILOVEYOU virus, which he and his colleagues were the first to discover in 2000. ILOVEYOU was wormable, meaning it spread automatically from computer to computer. It arrived via email as a text file, purportedly a love letter. If the target opened it, it would overwrite and corrupt some files on the person’s computer, and then send itself to all their contacts.
Advertisement
The virus infected over 10 million Windows computers worldwide.
Malware has changed dramatically since then. Virtually no one develops malware as a hobby, and creating malicious software that self-replicates is practically a guarantee that it will get caught by cybersecurity defenders capable of neutralizing it quickly, and potentially catching its author.
No one does it for the love of the game anymore, according to Hyppönen. “The age of viruses is firmly behind us,” he said.
Seldom do we now see self-spreading worms — with rare exceptions, such as the destructive WannaCry ransomware attack by North Korea in 2017; and the NotPetya mass-hacking campaign launched by Russia later that year, which crippled much of the Ukrainian internet and power grid. Now, malware is almost exclusively used by cybercriminals, spies, and mercenary spyware makers who develop exploits for government-backed hacking and espionage. Those groups typically stay in the shadows, and want to keep their tools hidden to continue their activities and to avoid cybersecurity defenders or law enforcement.
Advertisement
The other differences today are that the cybersecurity industry is now estimated to be worth $250 billion. The industry has professionalized, in part as a necessity, to fight the increase in malware attacks. Defenders went from giving away their software for free, to turning it into a paid service or product, said Hyppönen.
Computers and newer inventions like smartphones, which began to take off during the early 2000s, have become much harder to hack. If the tools to hack an iPhone or the Chrome browser cost six-figures or even a few million dollars, Hyppönen argued, this effectively makes an exploit so expensive that only the highly resourced, like governments, can use them, rather than financially motivated cybercriminals. That’s a huge win for consumers, and for the cybersecurity industry that’s a job well done.
Image Credits:courtesy of Mikko Hypponen
From fighting spies and criminals… to countering drones
In mid-2025, Hyppönen pivoted from cybersecurity to a different kind of defensive work. He became the chief research officer at Sensofusion, a Helsinki-based company that develops an anti-drone system for law enforcement agencies and the military.
Hyppönen told me that was motivated to get into a developing new industry because of what he saw happening in Ukraine, a war defined by drones. As a Finnish citizen, who serves in the military reserves (“I can’t tell you what I do, but I can tell you that they don’t give me a rifle because I’m much more destructive with a keyboard,” he tells me), and with two grandfathers who fought the Russians, Hyppönen is acutely aware of the presence of an enemy just over his country’s border.
“The situation is very, very important to me,” he tells me. “It’s more meaningful to work fighting against drones, not just the drones that we see today, but also the drones of tomorrow,” he said. “We’re on the side of humans against machines, which sounds a little bit like science fiction, but that’s very concretely what we do.”
Advertisement
The cybersecurity and drone industries may seem leagues apart from one another, but there are clear parallels between fighting malware and fighting drones, according to Hyppönen. To fight malware, cybersecurity companies have come up with mechanisms, known as signatures, to identify what is malware and what is not and then detect and block it. In the case of drones, Hyppönen explained, defenses involve building systems that can locate and jam radio drones, and by recognizing frequencies that are being used to control the autonomous vehicles.
Hyppönen explained that it’s possible to identify and detect drones by recording their radio frequencies, known as their IQ samples.
“We detect the protocol from there and build up signatures for detecting unknown drones,” he said.
He also explained that if you detect the protocol and frequencies used to control the drone, you can also try to conduct cyberattacks against it. You can cause the drone’s system to malfunction, and crash the drone into the ground. “So in many ways, these protocol level attacks are much, much easier in the drone world because the first step is the last step,” Hyppönen said. “If you find a vulnerability, you’re done.”
Advertisement
The strategy in fighting malware and fighting drones is not the only thing that hasn’t changed in his life. The cat-and-mouse game of learning how to stop a threat, and then the enemy learning from that and devising new ways to get around defenses, and on and on, is the same in the world of drones. And then, there’s the identity of the enemy.
“I spent a big part of my career fighting against Russian malware attacks,” he said. “Now I’m fighting Russian drone attacks.”
Starting this week, Microsoft has begun force-upgrading unmanaged devices running Windows 11 24H2 Home and Pro editions to Windows 11 25H2.
According to the company’s Lifecycle Policy site, Windows 11 24H2 will reach end of support in roughly six months, on October 13, 2026.
Also known as the Windows 11 2025 Update, Windows 11 25H2 began rolling out in September to eligible Windows 10 or Windows 11 devices as a minor update installed through enablement packages less than 200 KB in size.
“The machine learning-based intelligent rollout has expanded to all devices running Home and Pro editions of Windows 11, version 24H2 that are not managed by IT departments,” Microsoft said in a Monday update to the Windows release health dashboard.
Advertisement
“Devices running these editions will no longer receive fixes for known issues, time zone updates, technical support, or monthly security and preview updates containing protections from the latest security threats,” it added.
“These devices will automatically receive the update to Windows 11, version 25H2 when they’re ready. No action is required, and you can choose when to restart your device or postpone the update.”
Those who don’t want to wait for the automatic upgrade can manually check whether the update is available in Settings > Windows Update and click the link to download and install Windows 11 25H2.
Advertisement
If you’re not ready to upgrade, you can also pause updates from Settings > Windows Update by selecting the amount of time you’d like to pause them. However, you must install the latest updates after the time limit has passed.
Microsoft also provides a support document and a step-by-step guide to help users resolve problems encountered during the Windows 11 25H2 upgrade process.
Since the March 2026 Patch Tuesday updates were released, Microsoft has issued several emergency updates, including one that addresses a known issue breaking sign-ins with Microsoft accounts across multiple Microsoft apps, such as Teams and OneDrive.
Automated pentesting proves the path exists. BAS proves whether your controls stop it. Most teams run one without the other.
This whitepaper maps six validation surfaces, shows where coverage ends, and provides practitioners with three diagnostic questions for any tool evaluation.
[Washington, DC – April 2, 2026] – IREX, a global pioneer in ethical AI and intelligent video analytics deployed across 10+ countries and over 300,000 cameras, announced a major update to its FireTrack smoke and fire detection module. The update doesn’t require any additional hardware and broadens FireTrack’s applicability to critical infrastructure such as energy facilities and transportation hubs, public institutions including schools and hospitals, residential and commercial buildings, and parks, national parks, and forests.
Built on IREX’s ethical AI platform, the new module processes visual data in just 75–105 milliseconds –or about 0.1 second-, identifying danger almost instantly. This advancement – combined with improved model accuracy and resilience in poor lighting or weather – empowers early intervention by first responders, reducing the risk of catastrophic loss.
The updated model analyzes how fire and smoke evolve over time, distinguishing genuine hazards from harmless visuals like fog, headlights, or glare. This dramatically cuts down false alarms, allowing safety teams to focus on incidents that truly require attention.
To boost accuracy, IREX changed how the system “sees” fire and smoke. Instead of traditional bounding boxes around objects, the updated module uses segmentation, applying a color mask over the exact areas where fire or smoke appears: green for fire and red for smoke, thus better reflecting their irregular shapes. This approach improves the system’s ability to localize hazards precisely within the scene.
Advertisement
Credit: Irex
The updated FireTrack delivers early warning that is significantly faster than traditional optical or heat-based detectors by analyzing live video feeds for the visual signatures of smoke and fire in real time.
“Because the IREX AI platform seamlessly operates on existing camera networks, cities and organizations can strengthen fire safety without installing specialized sensor hardware – simply by connecting their CCTV systems to IREX,” said Serge Smirnoff, Head of PR at IREX. “Each detection event comes with a video snapshot for instant visual verification, enabling operators and first responders to quickly assess the situation and respond effectively.”
By leveraging the surveillance infrastructure already in place, the new FireTrack model offers a cost-effective path to comprehensive fire safety across both built environments and natural landscapes.
“The pride I feel for the IREX team today is immense. This FireTrack launch is a monumental achievement that reflects our core mission, to deploy ethical, intelligent AI to solve the world’s most critical problems,” said Calvin Yadav, CEO of IREX. “We are strengthening the resilience of entire communities globally, proving that every hour of hard work put into responsibly designed artificial intelligence is actively saving lives long before a single alarm sounds.”
The SDIC 8-bit MCU. (Credit: electronupdate, YouTube)
In this wonderful world of MEMS technology, sensor technology has been downsized and reduced in cost to the point where you can buy a car tire pressure sensor for less than $3 USD on a site like AliExpress. Recently [electronupdate] got his mittens on one of these items to take a look inside, and compare it against his trusty old mechanical tire pressure gauge.
Perhaps unsurprisingly, there isn’t a whole lot inside these devices once you pop them open to reveal the PCB. The MEMS device is a tiny device at the top, which has the pressurized air from the tire guided to it. The small hole inside the metal can leads to the internals that consist of a thin diaphragm with four piezoresistors that enable measurements on said diaphragm from which pressure can be determined.
Handling these measurements and displaying results on the small zebra connector-connected LCD is an 8-bit MCU manufactured by Chinese company SDIC. Although the part number on the die doesn’t lead to any specific part on the SDIC site, similar SDIC parts have about 256 bytes of SRAM and a few kB of one-time programmable ROM.
This MCU also integrates the clock oscillator, thus requiring virtually no external parts to work. Finally, its sigma-delta ADC interacts with the MEMS device, rounding out a very simple device that’s nevertheless more than accurate enough for a spot check as well as quite portable.
For decades, modern navigation has relied heavily on GPS, but another, less visible system plays an equally critical role in helping aircraft, ships, smartphones, and military platforms determine their position.
Earth’s magnetic field, constantly shifting and evolving, underpins the World Magnetic Model (WMM), a global reference that supports navigation systems used by billions of people every day.
Maintaining the accuracy of that model depends on reliable measurements of the magnetic field, yet much of the satellite infrastructure used to gather this data is aging, while the field itself is changing at an accelerating rate.
Advertisement
Article continues below
Quantum diamond magnetometers
These pressures have driven a search for new technologies capable of monitoring the magnetic field with greater precision and frequency.
Advertisement
In response, the US National Geospatial-Intelligence Agency (NGA) launched the MagQuest Challenge in 2019, a seven-year, multi-million-dollar competition designed to identify next-generation sensing technologies.
The goal is to develop compact, highly accurate systems that can provide continuous magnetic data, reducing reliance on periodic measurements and helping ensure the long-term reliability of global navigation systems.
One of the companies emerging from this effort is SBQuantum, a Canadian firm specializing in quantum sensing technology. Its approach centers on quantum diamond magnetometers, compact devices that use the principles of quantum physics to measure magnetic fields with exceptional sensitivity.
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
Advertisement
Recently, the company reached a major milestone when its sensor was launched into orbit as part of the final phase of the MagQuest program. The deployment represents a step toward continuous, space-based monitoring of Earth’s magnetic field and highlights the growing role of quantum technologies in navigation, defense, and public safety.
To better understand the development of this technology, the challenges involved in bringing it to space, and the potential applications beyond navigation, I spoke with David Roy-Guay, Founder of SBQuantum.
Before we start, can you give us a brief overview of what the WMM is and why it is so important for us.
The World Magnetic Model (WMM) is what powers every electronic compass, including the one in your watch and cellphone. It is essential to keep up to date as the Magnetic North Pole is moving. It was in the Canadian north and is now shifting toward Siberia. This has a real impact on the precision of every analog and digital compass.
Everyday, we use the WMM, just think of the blue arrow in your favorite navigation application telling you to head left or right as you exit a subway station or a hotel. This directional information is complementary to GPS, which provides location information, but doesn’t tell you which way you are facing.
Advertisement
You mentioned that the satellites feeding it with data are reaching their end of life. What happens next?
Typically the WMM is updated every 5 years when a new official version is released. However recently a new update was released after only 4 years because the movement of the field had accelerated.
Once the mission of the current ESA SWARM constellation of satellites comes to an end, the existing magnetic field maps will be of little value 2-3 years after that. This means the navigation systems on board aircraft and drones will be off significantly, especially in the northernmost areas, possibly up to dozens of degrees. I can think of one example in Alaska when recently a landing strip had to have its numbers changed since it was no longer facing the same direction according to the WMM.
In comparison, our platform ‘Diamond Polaris – 1’ will allow the continuous production of magnetic data for the WMM. This approach is far more cost-effective, gathers and assembles faster, and offers data well suited for accurate positioning.
How does the data from the WMM project convert into something that can be an alternative to the ubiquitous GPS?
Data collected over a year of orbit is processed and curated by the US NOAA and the US NGA, to inform future versions of the WMM. Although the data is coarse it is applicable to compass applications. Higher resolution versions can be produced by deploying multiple satellites and drones to gather data at different altitudes.
These high-resolution maps will act as a calibration reference to navigation systems (INS systems) and could provide positioning data without GPS to up to 100m precision.
Our spring 2026 space-launch came after years of testing and retesting with NASA and other organizations. SBQuantum’s sensor was deemed to be fit for use in space. This first space deployment is the next step on the road to making magnetic navigation widely available as an alternative to GPS which cannot be jammed or distorted.
Advertisement
Your company built something called a diamond quantum magnetometer. Why diamond and why quantum?
Being solid state, diamonds are exceptionally stable and provide the right environment to preserve quantum coherence for an extended period, even at room temperature. This enables highly sensitive and very accurate magnetic field measurements for extended satellite missions at a global scale.
Furthermore, the atomic structure of diamonds is well suited to provide measurement of magnetic fields along three axes. For the purposes of navigation it is essential to gather all of that in order to provide directional information.
You mentioned the size of the device (roughly a quart of milk — about 1L in metric or a cube with 10cm size). Does your roadmap contain products that are smaller? What would something “better” differ in terms of features?
We are still in the early stages of this diamond technology. One of its advantages is that it can eventually be shrunk further, to about the size of a matchbox, without degrading its performance.
This is not the case for classical directional magnetometer technologies. We expect to reach that point in about 3 years, once we scale the production to industry standard wafers, which are of course widely used in the semiconductor industry.
How does the data captured by a quantum sensor allow for “advanced interpretation algorithms” that conventional sensors simply cannot support? What other applications could these sensors have?
By building an array of directional diamond magnetometers, we can enable real-time magnetic signals interpretation in a way which was otherwise not possible. For instance, we can locate metallic objects underwater, in real-time.
This is also true for metallic objects on the other side of a wall or underground. We are therefore also looking to employ the technology to support security and defense applications.
For instance this could be used for tracking submarines from a drone, or enhancing security at sporting events, or even security at schools and corporate events.
We miss the old Heathkit. You could build equipment that rivaled or even surpassed commercial devices. The cost was usually reasonable and, even if you could get by with less, the satisfaction of using gear you built yourself was worth a lot. Not to mention the knowledge you’d gain and your confidence in troubleshooting should the need arise. So we were jealous of [RCD66] when he found a Heathkit AJ-43C stereo tuner in the recycle bin.
As you can see in the video below, it needed a lot of love to get back to its former self. The device dates from around 1965, when the kit cost $130. In 1965, that was a lot of money. Back then, that would have bought you about four ounces of gold and would have been a great down payment on a $1,500 VW bug.
Things were a bit of a mess, so he removed all the parts and replaced most of them. Unsurprisingly, the electrolytic capacitors all tested bad. The transistors were all germanium, but if they tested good, his plan was to reuse them. There were several PCBs inside, and he made some changes, such as replacing the zener diode power supply with something more modern.
Advertisement
How did it sound? Watch the video and see for yourself. We usually like troubleshooting specific problems on gear like this, but in this case, it was probably smart to just do a total rework.
Season 1 hasn’t even aired yet, and Star Wars: Maul – Shadow Lord is already coming back for more. Chief Creative Officer Dave Filoni has announced that Season 2 is officially in the works at Lucasfilm Animation.
Star Wars: Maul – Shadow Lord Season 1 kicks off on Disney+ with a two-episode premiere on April 6, dropping two episodes weekly after that. No release date for Season 2 has been shared yet, but the early renewal signals serious confidence in the show.
This 10-episode animated series picks up after The Clone Wars, with Maul trying to rebuild his criminal syndicate on a planet the Empire hasn’t touched. Along the way, he encounters a disillusioned young Jedi Padawan, who might become the apprentice he needs.
With Season 2 locked in before Season 1 even premieres, Maul’s story is clearly just getting started.
Advertisement
The stellar cast includes Golden Globe winner and Oscar nominee Wagner Moura as Brander Lawson, Richard Ayoade as Two-Boots, Dennis Haysbert as Master Eeko-Dio Daki, Gideon Adlon as Devon Izara, and several others.
When are the new episodes of Star Wars: Maul – Shadow Lord season 1 coming?
Star Wars
Star Wars: Maul – Shadow Lord follows a two-episode-per-week format, rolling out every Sunday this month. Here’s the full breakdown:
April 6 – Episodes 1 and 2: “The Dark Revenge” and “Sinister Schemes”
April 13 – Episodes 3 and 4: “Whispers in the Unknown” and “Pride and Vengeance”
April 20 – Episodes 5 and 6: “Inquisition” and “Night of the Hunted”
April 27 – Episodes 7 and 8: “Call to the Oblivion” and “The Creeping Fear”
May 4 – Episodes 9 and 10: “Strange Allies” and the as-yet-untitled Season 1 finale
Windows PCs are about to get a little more touchy. Microsoft is now testing a new kind of interaction in Windows 11 that doesn’t just show you what’s happening on screen, but it lets you feel it too.
Microsoft
Rolling out in the latest Insider build, the update introduces haptic feedback for a bunch of everyday actions. It’s subtle, it’s optional, and if done right, it could make Windows feel a lot more responsive.
What’s changing in Windows 11 with haptics?
With Windows 11 Insider Preview Build 26300.8155, Microsoft is adding haptic feedback effects to compatible devices like advanced trackpads and possibly some mice. The idea is simple: certain actions across the OS will now trigger a small physical response, almost like a tap or vibration.
Microsoft
These aren’t random buzzes either. The system is designed to respond to specific interactions, things like snapping windows into place, resizing them, aligning objects in apps like PowerPoint, or even hovering over the close button. The feature lives under input settings, where users can toggle it on or off and tweak how it behaves. And importantly, it’s limited to hardware that actually supports haptics, meaning this won’t magically show up on every old laptop.
Microsoft
Alongside the headline haptics feature, this build also brings a few smaller but useful refinements. The Xbox full-screen experience is now rebranded as Xbox mode, with a smoother first-run setup to make things feel more seamless for gamers. There are also under-the-hood improvements, including faster startup app launches, fixes for recent sign-in issues in certain apps, and a patch for a printing-related crash that had been affecting some Insider users.
Why Windows suddenly wants you to “feel” your actions
Haptics have long been a natural part of smartphones, adding subtle vibrations to confirm taps and gestures, while Windows has mostly relied on visuals and sounds. Now, Microsoft is bringing that same tactile layer to PCs, especially as more devices adopt haptic trackpads and stylus-friendly designs. The idea is simple: reduce the need to constantly look for on-screen confirmation by letting users feel their actions.
It also signals a broader shift in how Windows is evolving, moving toward a more immersive experience that blends sight, sound, and touch. If done right, it could make everyday interactions feel more intuitive and responsive—but it’s a delicate balance. Too much feedback could get annoying, but if Microsoft nails it, this might end up being one of those features that quietly becomes hard to live without.
Perplexity, one of the fastest-growing AI search tools right now, is suddenly facing some serious heat. And this time, it’s not about accuracy or hallucinations.
Perplexity
A fresh lawsuit is raising uncomfortable questions about what actually happens to user data behind the scenes, especially when people assume their chats are private. And if the allegations hold any weight, this could be one of those moments that prompts many users to rethink how casually they share information with AI tools.
Is Perplexity’s “incognito mode” actually private?
According to a newly filed class-action lawsuit by an anonymous Perplexity user, John Doe, not quite. The complaint alleges that Perplexity’s so-called incognito mode is essentially a “sham” that fails to protect user data as most people would expect.
Tushar Mehta / Digital Trends
The lawsuit claims that user conversations, including potentially sensitive topics like financial advice, health concerns, or legal queries, were shared with third parties like Google and Meta. And as reported by Ars Technica, this happened even when users explicitly chose incognito mode, which is supposed to limit tracking and data collection.
Joe Maring / Digital Trends
What’s more concerning is the kind of data allegedly involved. Reports suggest that information such as IP addresses, email IDs, geolocation data, and even full chat transcripts may have been passed along for ad targeting purposes. The lawsuit also accuses Perplexity of embedding tracking tools similar to those used in online advertising, without clearly informing users. In some cases, it even claims that entire conversations could be accessed via publicly reachable links.
Why this lawsuit could change how we trust AI
This goes beyond one app as AI tools feel personal, which makes oversharing easy. The lawsuit also claims years of chats were shared with ad giants, and that Perplexity doesn’t clearly surface its privacy policy like rivals do.
If true, it could force stricter transparency across AI platforms. For now, they’re just allegations, but enough to make that next AI prompt feel a little less casual.
Tencent Holdings has launched ClawPro, an enterprise AI agent management platform built on OpenClaw, the open-source framework that has become the fastest-growing project in GitHub’s history and the unlikely centrepiece of a national technology craze in China. The tool, released in public beta by Tencent’s cloud division on Thursday, allows businesses to deploy OpenClaw-based AI agents in as little as 10 minutes, with controls for template selection, model switching, token-consumption tracking, and security compliance. During its internal beta, ClawPro was adopted by more than 200 organisations across finance, government, and manufacturing, sectors that require the kind of strict data governance that the open-source version of OpenClaw was never designed to provide.
ClawPro is the latest and most commercially significant addition to Tencent’s growing suite of OpenClaw products, which now spans individual users, developers, and enterprises. In March, the company released QClaw, a mini-programme that embeds OpenClaw inside WeChat, giving the framework access to the app’s 1.3 billion users. It simultaneously launched WorkBuddy, a workplace AI agent tested by more than 2,000 non-technical employees across human resources, administration, and operations, and ClawBot, a WeChat plugin supporting multi-modal interactions. The speed of the rollout reflects Tencent’s determination to position WeChat not just as a messaging platform but as the primary interface forthe agentic AI wave that is reshaping how software gets used.
The object of all this enterprise engineering is a tool created by Peter Steinberger, an Austrian developer who published the first version under the name Clawdbot in November 2025. The software, built to let large language models operate computers, call tools, and execute tasks autonomously, was renamed twice in three days in late January 2026, first to Moltbot after Anthropic raised trademark concerns over the phonetic similarity to “Claude,” then to OpenClaw because Steinberger found Moltbot “never quite rolled off the tongue.” In February, he announced he would be joining OpenAI and transferring the project to an open-source foundation. By that point, the project had already passed React to become the most-starred software repository on GitHub, a record it reached in 60 days that took React more than a decade. As of late March, OpenClaw had 335,000 GitHub stars, 27 million monthly visitors, 2 million active users, and more than 13,700 community-built skills on its ClawHub marketplace.
In China, the adoption curve has been extraordinary. The country now has more OpenClaw users than any other, roughly double the activity of the United States according to analysis by SecurityScorecard. The phenomenon has been given a name: “raise a lobster,” after OpenClaw’s crustacean logo and mascot, which Steinberger chose because a lobster sheds its shell to grow. Tencent organised public installation sessions in Shenzhen that drew retirees and students. Baidu held similar events in Beijing. A cottage industry of technicians began charging 500 yuan, around $72, for on-site installations. Nvidia’s Jensen Huang told CNBC that OpenClaw was “definitely the next ChatGPT.” The Chinese state media apparatus amplified the enthusiasm. “Claw-powered” one-person companies became a talking point at the National People’s Congress, and local governments began offering grants to startups building applications on the framework.
The enthusiasm collided with reality almost immediately. In March, China’s National Computer Emergency Response Team warned that OpenClaw had “extremely weak default security configuration” and that attackers could exploit the tool by embedding malicious instructions in web pages or distributing poisoned plugins. The Ministry of Industry and Information Technology’s National Vulnerability Database published formal security guidelines urging users to run only the latest version, minimise internet exposure, and grant the agent the minimum permissions necessary. State-owned enterprises and government agencies, including the country’s largest banks, received notices warning them against installing OpenClaw on office devices. Several were instructed to report existing installations for security review and possible removal. Bloomberg reported that China moved to curb OpenClaw use at banks and state agencies, a striking reversal for a tool the government had been celebrating weeks earlier.
Advertisement
Tencent’s own relationship with OpenClaw has not been without friction. On 11 March, Tencent Cloud launched SkillHub, a Chinese-localised mirror of OpenClaw’s ClawHub marketplace, by scraping more than 13,000 skills from the original registry. The bulk scraping pushed Steinberger’s server costs into five digits and caused slowdowns on official servers. He complained publicly on X. Five days later, Tencent Cloud and Tencent AI appeared on OpenClaw’s official sponsor list, providing lightweight application servers for one-click deployment. The episode encapsulated a dynamic familiar in Chinese tech:a European project supplies the foundational innovation, Chinese companies scale it faster than anyone else, and the relationship between creator and commercialiser oscillates between parasitism and partnership.
The competitive context is fierce. Alibaba, which holds a 35.8 per cent share of China’s AI cloud market compared with Tencent’s smaller position, integrated its Qwen AI assistant across Taobao, Tmall, Alipay, and other consumer platforms, reaching 300 million monthly active users by early 2026 and delivering roughly 140 million first-time AI shopping experiences during a Chinese New Year promotional campaign. ByteDance is pursuing platform independence through Douyin and a state-media partnership. Baidu’s AI-powered business now accounts for 43 per cent of its core revenue, up from 26 per cent a year ago. Tencent’s strategy depends on WeChat’s unmatched distribution, its 1.3 billion users, and the bet that AI agents will become features of existing super-apps rather than standalone products. The company spent 18 billion yuan on AI products in 2025 and plans to double that in 2026.
ClawPro is the piece of that strategy designed to generate cloud revenue. Enterprise AI agent deployments require infrastructure, compute, model hosting, security layers, and compliance tooling, all of which Tencent can bill for even when the underlying agent framework is free. The 200 organisations that trialled ClawPro during its internal beta represent the beginning of a conversion funnel: take the enthusiasm for a consumer phenomenon, channel it through enterprise-grade tooling, and extract recurring cloud revenue from the result. It is the same playbook thatEuropean cloud companies have used to monetise open-source software, applied at a scale and speed that only the Chinese tech ecosystem can achieve.
The security concerns are not trivial. OpenClaw, by design, grants AI agents broad access to local files and the ability to communicate with external services. In an enterprise context, a misconfigured agent could exfiltrate sensitive documents, execute unauthorised transactions, or expose internal systems to prompt-injection attacks. The tension between the open-source community’s permissive defaults and the compliance requirements of banks, government agencies, and manufacturers is precisely the gap that ClawPro is designed to fill. Whether Tencent’s security layer is robust enough to satisfy Chinese regulators, who have already demonstrated their willingness to restrict the tool entirely, will determine whetherthe year of governed AIproduces governed AI agents or merely governed press releases about them.
Advertisement
The broader significance of the OpenClaw phenomenon is what it reveals about the geography of AI adoption. The tool was built by a single developer in Austria, renamed after a trademark dispute with an American AI company, transferred to an open-source foundation after its creator joined OpenAI, and then adopted at a velocity in China that dwarfs anything that has happened in the West. The country that produced DeepSeek, the AI model thatrattled Silicon Valley’s assumption that scale required American infrastructure, is now demonstrating that it can also adopt, adapt, and commercialise foreign AI tools faster than the markets that created them. Tencent’s ClawPro is, in that sense, less a product launch than a proof of concept for a pattern that will repeat: the open-source AI stack is global, but the speed of enterprise adoption is determined by the ecosystems that can distribute it. In China, that ecosystem runs through WeChat, and WeChat runs through Tencent.
You must be logged in to post a comment Login