Technology

• Security researchers found a PDF app for Android sporting a banking trojan
• The trojan was introduced with a patch, six weeks after release
• It had more than 50,000 downloads, so users should beware

A dangerous Android banking trojan has found a way to the Google Play Store once again, potentially affecting tens of thousands of North American users, experts have warned.

Security researchers from Threat Fabric found an app on the Play Store, called ‘Document Viewer – File Reader’, published by a company called ‘Hybrid Cars Simulator, Drift & Racing’ roughly two months ago and having amassed a significant following - some 50,000 people.

Until only recently, the app was clean, working as intended. Then, between June 24 and 30, it received an update that turned it into a banking trojan called Anatsa.

How to stay safe

This is a known piece of malware that's been smuggled into the Play Store on multiple occasions in the past.

BleepingComputer claims in November 2021 researchers found a trojanized app with 300,000 downloads, and in June 2023 a separate one with 30,000 downloads. In February 2024 there was another app with Anatsa, counting 150,000 downloads, and in May the same year, two apps with 70,000 downloads between them.

Every time, Google removes the apps, but the attackers seem to find a way back.

Anatsa is a banking trojan that first scans the victim’s mobile device, looking for North American banking apps.

If it finds any, it serves them an overlay that grabs credentials and other login data, granting the attackers the ability to log into accounts and make transactions. At the same time, the victims are presented with a message that the app is undergoing scheduled maintenance.

The app has now been removed from the Play Store, and if you have it installed, it would be wise to remove it and then run a full system scan using Play Protect. Resetting banking account credentials would also be advised.

“All of these identified malicious apps have been removed from Google Play,” a Google spokesperson told BleepingComputer. “Users are automatically protected by Google Play Protect, which can warn users or block apps known to exhibit malicious behavior on Android devices with Google Play Services."

Via BleepingComputer

You might also like

• This dangerous new malware is hitting iOS and Android phones alike - and it's even stealing photos and crypto
• Take a look at our guide to the best authenticator app
• We've rounded up the best password managers

Recently, a new sentiment has emerged in AI security circles: "RAG is dead." I've observed firsthand how organizations are increasingly abandoning Retrieval-Augmented Generation (RAG) architectures in favor of agent-based approaches.

This shift represents not a temporary trend but a fundamental recognition of RAG's inherent security and performance limitations.

The inherent flaws in RAG architectures

When enterprises first implemented AI applications, RAG emerged as the default architecture -- extracting data from internal systems, centralizing it in vector databases and using it to augment AI model outputs. While conceptually straightforward, this approach revealed critical vulnerabilities at scale:

RAG architectures create significant security risks by centralizing data from disparate systems into repositories that frequently bypass the original access controls. These centralized stores become potential data exfiltration points, often circumventing authorization checks that existed in source systems. Additionally, data quality degrades quickly as information in these repositories becomes stale, requiring constant synchronization with source systems.

The technical challenges compound as organizations scale. Each new data source added to a RAG system requires custom extraction logic, formatting rules and ongoing maintenance. When multiplied across dozens or hundreds of internal systems, this creates an unsustainable maintenance burden. Furthermore, performance bottlenecks emerge as vector databases grow, resulting in slower response times and degraded user experiences.

The security challenges become prevalent in regulated industries like education or healthcare. Consider how an AI system using RAG might handle sensitive student records or patient information -- once extracted from secure systems with proper access controls, this data enters a parallel repository with potentially weaker protections, creating compliance risks and security vulnerabilities. For financial institutions, the risks extend to potential regulatory violations and financial penalties if customer data becomes exposed through these secondary repositories.

The agent-based alternative

In response to these challenges, forward-thinking enterprises are pivoting to agent-based architectures. Rather than extracting and centralizing data, these systems employ software agents that query source systems directly at runtime, respecting existing access controls and authorization mechanisms.

This architectural shift offers several critical advantages:

1. Elimination of duplicate data repositories -- information remains in its original systems with their established security controls
2. Preservation of authorization models -- access controls from source systems remain in effect
3. Improved data freshness -- queries always access the most current information
4. Reduced attack surface -- fewer data stores means fewer potential breach points
5. Enhanced user experience -- responses reflect the most current organizational knowledge
6. Simplified compliance -- data governance policies remain consistent across all systems
7. Reduced maintenance overhead -- no need to continuously update and synchronize extracted data

Many large enterprises that initially implemented RAG as their first AI initiative have subsequently moved to agent-based approaches after encountering these limitations in production environments.

Implementation reality vs. media perception

Despite media excitement about fully autonomous agents, the reality in enterprise environments is more measured. The productive implementations I've observed involve specific, well-defined agent workflows with clear security boundaries rather than completely autonomous systems.

Most organizations are currently implementing agent systems that:

• Operate within defined parameters and workflows
• Have explicit permission models
• Maintain comprehensive audit trails
• Include guardrails that prevent unauthorized actions
• Employ human-in-the-loop verification for critical operations
• Implement circuit breakers that automatically terminate suspicious activities

The distinction between theoretical capabilities and practical implementations is important. While academic research may showcase fully autonomous agents, enterprise deployments prioritize security, reliability, and predictability over complete autonomy.

Security implementation for agent-based systems

For organizations transitioning to agent-based architectures, several essential security controls should be implemented:

1. Authentication and Authorization

Agent systems require robust user authentication tied directly to authorization, with granular controls at document and data chunk levels. Role-based, relationship-based and attribute-based access control models provide the necessary flexibility for enterprise environments. Implementing just-in-time access provisioning further reduces the risk profile by limiting access duration to only what's necessary for task completion.

2. Visibility and Monitoring

Security teams need complete visibility into agent operations, including model versions, authentication events, prompts, behaviors, data citations and all interactions with external systems. Real-time alerting for anomalous patterns and comprehensive logging for forensic analysis are essential components of a robust monitoring system.

3. Content Protection

Real-time content filtering capabilities must be implemented to prevent sensitive data exposure, detect malicious content and protect organizational information assets. Sophisticated DLP (Data Loss Prevention) mechanisms should be deployed to recognize and redact sensitive information before it leaves controlled environments.

Case study in secure AI implementation

Grand Canyon Education, a publicly-traded education services company, developed an AI chatbot platform for thousands of students and staff across 22 university partners. Rather than building their own redaction solution, which would have significantly delayed their project, they implemented API-driven security guardrails that could programmatically redact sensitive data from user prompts and uploaded files before they reached backend AI models.

This approach allowed their security team to make redaction policy changes without requiring developer sprint cycles. The result was a secure, managed AI platform with sensitive data automatically redacted in real-time and no perceptible latency for users, reducing the risk of that data ending up in AI model training sets.

The path forward

The shift from RAG to agent-based architectures represents a natural evolution in enterprise AI implementation. As organizations gain practical experience, they're adapting their approaches to better address security, performance and user experience challenges.

While some security teams may consider developing in-house solutions, the organizations succeeding most so far with agent-based AI, are those leveraging specialized security tools that integrate seamlessly into their AI workflows. These purpose-built solutions provide the right balance of control and flexibility while minimizing development and maintenance costs.

This transition mirrors similar evolutions in other technology areas, where initial approaches give way to more sophisticated, secure designs as implementation experience grows. By embracing agent-based approaches with appropriate security controls, enterprises can deliver more powerful, secure AI capabilities while avoiding the pitfalls of first-generation RAG implementations.

We've featured the best online cybersecurity course.

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

Samsung Galaxy Unpacked's many new products and features have not left out AI examples. Plenty involved Google and its Gemini family of AI models, with a host of new features coming to Android devices with the new Android 16 and Wear OS 6 systems. Here are some of the ones to be the most excited for.

Gemini Live gets way more useful on foldables

(Image credit: Samsung)

Gemini Live is a way for Google's AI companion to be present on a continuous basis. Rather than just asking a question and moving on, you can have it on hand to help as you follow a cooking tutorial, fix your bike, or do yoga. Starting with the Galaxy Z Flip7, Gemini Live will now be accessible right from the external screen, meaning you won't have to even unfold the device to interact with the AI.

You'll also be able to link Gemini Live via Flex Mode with full camera integration. So, you might flip your phone halfway open, tap the camera button, and have Gemini be a hands-free AI assistant that can actually see what you're doing or what's happening around you and offer advice on your half-completed DIY project or your latest outfit. You can show Gemini what you're looking at and get on-the-spot feedback without fully unfolding your phone.

Circle to Search gets a big Gemini upgrade

(Image credit: Google)

Circle to Search is Google Gemini's party trick of looking up things you draw a circle around on your screen, like a photo or a phrase written in a text you're looking at. It's a way to get Google Search results without switching apps. The trick is becoming a lot more impressive with an AI Mode upgrade.

Now, when you circle a word, image, or phrase, Gemini doesn’t just look it up on Search; it starts up the AI Mode version of online search with a conversation that allows you to ask follow-up questions and look into related ideas within the same setup. You can discuss complex topics right from your screen with Gemini without switching among multiple tabs and apps.

Gaming the Circle

That's not the only major upgrade to Circle to Search. The feature will now try to entice people playing video games to look things up just like someone circling an unfamiliar plant. The new Gemini-powered feature offers mobile gamers help in the context of the game. You can just circle something on the screen as you're playing, like an item, enemy, or puzzle, and Gemini will identify it and offer timestamped advice based on your progress about what to do next if you're stumped. It's sort of an interactive walkthrough and strategy guide that you don't need to stop playing to look through.

Gemini Live starts talking to your apps

(Image credit: Future | Alex Walker-Todd)

Until now, Gemini has mostly been incorporated into Google apps and services. But, Gemini Live will now start integrating with native device apps, starting with Samsung’s Calendar, Notes, and Reminders apps. So, you could ask Gemini to summarize your day, add reminders about your next meeting, and organize your notes about what to buy for a vacation you took last week without opening those apps. Other apps will start offering their information to you through Gemini Live soon, with the end goal apparently a more proactive AI manager of your life.

Gemini finally shows up on your wrist

(Image credit: Samsung)

Google Assistant has often seemed to struggle on smartwatches, but Gemini will apparently feel right at home on your wrist, starting with the Galaxy Watch8 series and Wear OS 6, with other smartwatches to follow. Gemini will provide better notifications, real-time voice support, and contextual responses to what you ask the smartwatch. The interface is supposed to be more natural as well, with the AI not feeling like an afterthought.

You might also like

• I tried Google’s new Search Live feature and ended up debating an AI about books
• I tried Google's new AI mode powered by Gemini, and it might be the end of Search as we know it
• Google’s AI Mode can explain what you’re seeing even if you can’t

• TBK DVRs and Four-Faith routers are carrying known, old security flaws
• The flaws were used to build Mirai botnet in the past, and are now used to build RondoDox, too
• Users are advised to patch, firewall, or replace vulnerable endpoints

A new malicious botnet, called RondoDox, is being built at this very moment, potentially targeting thousands of products around the world e xperts have warned.

Cybersecurity researchers from Fortinet FortiGuard Labs, who said they saw multiple vulnerabilities in different digital video recorders (DVR) and routers being exploited to create the botnet.

The vulnerabilities in question are tracked as CVE-2024-3721, and CVE-2024-12856. These two were found in TBK digital video recorders, models DVR-4104 and DVR-4216, as well as in Four-Faith routers, models F3x24 and F3x36.

Defending your endpoints

According to BleepingComputer, the flaws were exploited before by threat actors building the infamous Mirai botnet. They are popular among cybercriminals because these devices are often used in retail stores, warehouses, small offices, and similar places, where they “often go unmonitored for years”.

As such, they’re prime targets - easy to compromise, and active for years without patches or updates.

Cybercriminals love building botnets. A network of compromised devices, from routers, to smart home devices, can be used for all sorts of nefarious activities, from distributed denial of service (DDoS) attacks, to residential proxy services that can be rented out.

In fact, RondoDox seems to be used for stealth proxies, hiding command-and-control (C2) traffic for even more malicious activities. It is also used to run layered scams, or to amplify DDoS-for-hire campaigns.

It is also rather good at staying hidden, the researchers claim, trying to spoof gaming traffic.

"To evade detection, it disguises malicious traffic by emulating popular games and platforms such as Valve, Minecraft, Dark and Darker, Roblox, DayZ, Fortnite, GTA, as well as tools like Discord, OpenVPN, WireGuard, and RakNet," Fortinet explained.

"Beyond gaming and chat protocols, RondoDox can also mimic custom traffic from tunneling and real-time communication services, including WireGuard, OpenVPN variants (e.g., openvpnauth, openvpncrypt, openvpntcp), STUN, DTLS, and RTC."

As usual, to defend against these threats, users should make sure their routers and DVRs have updated firmware, and strong, custom passwords. If they are no longer supported by their vendors, they should be replaced by newer models. Furthermore, if possible, the devices should be disconnected from the public internet, or placed behind a firewall.

Via The Hacker News

You might also like

• TVT DVRs become prime target for Mirai botnet
• Take a look at our guide to the best authenticator app
• We've rounded up the best password managers

Apple TV+ have confirmed that their Gary Oldman-starring darkly comic spy thriller, Slow Horses will return for season 7.

It’s a bold move for the streamer considering that Slow Horses season 5 is yet to air, and the renewal will see the series continue until at least 2027. Although, considering the last outing of one of the best Apple TV+ shows became Rotten Tomatoes’ number one series of 2024, it’s hardly surprising. Add to that the fact that Slow Horses has picked up numerous Golden Globe, Emmy and BAFTA nominations across its run so far, and you can see why one of the best streaming services think this horse is still very much in the race.

As mentioned, the next adventure for the Slough House gang doesn’t arrive until later in the year, but to get a taste of what’s to come, you can check out the trailer below.

Based on the series of books by Mick Herron, Slow Horses follows the “losers, misfits and boozers” of Slough House, a dumping ground for MI5 operatives who’ve messed up, but not enough to be worth the red tape of sacking. Thrust into Slough House’s mildew ridden administrative hell, the expectation is that the misery, drudgery and abuse from irascible boss Jackson Lamb (Gary Oldman) will soon send the unlucky few on their way.

However, despite making thoughtless mistakes, suffering unsociable addictions and being victims of career sabotage, those that make up the residents of Slough House, or the 'slow horses’, are in fact very good at what they do, and seem to find themselves routinely, and quite accidentally, tasked with saving the country from various threats.

There’s definite Killing Eve vibes to the damp, unglamorous depiction of a British spy game still secretly fighting the Cold War, however once the plot of each season kicks into gear, the show feels closer to the relentless pace and shocking plotting of 24, just with a heavy dose of sardonic British humour.

The cast too, are excellent, with Oldman’s slobbish Lamb as unpleasant to watch as he is a joy to behold. Jack Lowden is the closest the show has to a hero, playing River Cartwright, a young agent with family ties to the old game, while Christopher Chung plays antisocial tech expert Roddy. Other notable ‘horses’ include reliable and level-headed Louisa (Rosalind Eleazar) and Aimee-Ffion Edwards as pocket rocket Shirley.

What should we expect from Slow Horses season 7?

(Image credit: Apple TV Plus)

We won’t go in to too much detail about what to expect from season 7 of the show –after all, there’s still two seasons to come before that – but the fact that it’s set to be based on the eighth book, Bad Actors is intriguing, as thus far, it’s been a one to one book to season match. We do however have this tease from Apple TV+: “Lamb and his Slow Horses are on the hunt to find and neutralize a mole at the heart of British Government before they can bring down the state.”

Season 6 of the show looks to continue the usual tradition, being based on the equivalent book, Joe Country, however, the novel Slough House and two novellas take place between that and Bad Actors, so it’ll be interesting to see what fits where when the episodes hit screens.

At time of writing, there’s only one Slough House novel after Bad Actors, however, with a handful of same universe novels and novellas telling stories in and around MI5 yet to have much of a presence in the show, it’s possible we’ve got plenty more adventures with the ‘slow horses’ yet to come.

But we know for sure that we’ve definitely got at least three more seasons, the first of which – season 5, based on the novel London Rules – premieres on Apple TV+ on September 24.

You might also like

• 3 new Apple TV+ movies and shows you should stream this month
• 10 new Apple TV+ shows and movies are coming in mid-2025 – here are the 3 I'll be streaming on my Apple TV 4K
• 'I’m such a huge sci-fi nerd': Murderbot stars reveal what attracted them to the unusual Apple TV+ show

Marvel Rivals ranked mode lets you slowly progress against harder and harder teams, with rewards associated with reaching certain thresholds in Season 2.5. Getting into the upper ranks is no easy feat, as the rules change slightly as you climb, including one that lets you and your team choose a character to ban from play.

In terms of Competitive Mode, Marvel Rivals already offers a whole host of rewards for reaching the higher ranks, including an exclusive new skin for Ultron, one of the new Heroes set to be ranked in our Marvel Rivals tier list. Challenging yourself as you move up higher and higher in ranks is part of Marvel Rivals' appeal, making it a new contender for our best crossplay games list.

Here's all of the Marvel Rivals ranks in order, so that you can know just how close you are to the top. Also, we'll explain how things change as you reach Gold Rank, and walk you through some of the rewards you'll get once the season ends. With Season 3 just days away, we'll also cover what's changing.

Marvel Rivals Season 3 Competitive Mode changes

(Image credit: NetEase Games)

Marvel Rivals Season 3 will make the following Competitive Mode changes when it launches on July 11, 2025:

• New Domination Map - Klyntar: Celestial Husk
• Rank Reset: Based on your final Season 2.5 rank, ranks will be demoted by 7 divisions. For example, if you ended the season at Diamond I, your Season 3 journey begins at Gold II
• New Rank Settlement Rewards: Reach Gold for the incredible Phoenix - Emerald Flames costume, and earn fresh Crests of Honor for Grandmaster, Celestial, Eternity, and the illustrious Top 500
• Nameplate Frame rewards for Diamond and Platinum ranks.

Marvel Rivals ranked: here's what's changed for Season 2.5

(Image credit: NetEase)

There's been a few key changes to Competitive Mode thanks to the arrival of Marvel Rivals Season 2. You'll find the highlights below:

• Map rotation will be in effect from Season 2 onwards (ranked), Hellfire: Gala (new map)
• New limited-time modes will be added over time
• Bans added to Gold III
• Minimum level requirement now 15
• One Above All and Eternity ranks only allow duo queuing
• Increasing the rate of individual performance vs match outcome.

These are just the main updates that arrived in Marvel Rivals alongside the new season. It's worth noting that Marvel Rivals makes updates to ranked play fairly frequently. stay up to date with our latest Marvel Rivals Patch Notes coverage.

Marvel Rivals ranks in order

(Image credit: NetEase)

Below, you'll find a list of the Marvel Rivals ranks in order:

• Bronze: Tier 3-1
• Silver: Tier 3-1
• Gold: Tier 3-1
• Platinum: Tier 3-1
• Diamond: Tier 3-1
• Grandmaster: Tier 3-1
• Celestial: Tier 3-1
• Eternity: Rack Up Points
• One Above All: Rack Up Points (limited to the top 500 players)

As you can see, all but the top two ranks have three tiers that you must move through. Once you get to Eternity, you'll simply need to rack up points to rank up. Be aware, however, that at Eternity and One Above All, you will gradually lose points if you're not active enough.

How to unlock Ranked in Marvel Rivals

(Image credit: NetEase)

To unlock Ranked Play in Marvel Rivals you will need to reach level 15 (as of Season 2). This is done by simply playing matches. We recommend jumping into Quick Play to get a feel for the characters, and the two main game modes. To make sure you're winning as many games as possible, and featuring as the MVP regularly, you may want to consider equipping custom Marvel Rivals Crosshairs.

If you're looking to team up with some friends, be aware that players in Gold and below can squad up without any restrictions. For those in Gold Tier 1 to Celestial tiers, teaming is limited to a maximum range of three divisions. Players in Eternity or One Above All can team with Celestial Tier 2 players within 200 points. However, if they exceed 200 points, the point difference must be less than 200. One Above All and Eternity ranks only allow duo queuing

Marvel Rivals Ranked rewards for Season 2.5

(Image credit: NetEase)

There are a few different rewards that you can pick up by playing Season 2.5 of Marvel Rivals' Competitive Mode. They are as follows:

• Golden Ultron skin: Reach at least Gold Tier 3 by the end of Season 2.5
• Silver Crest of Honor: Reach at least Grandmaster Tier 3 by the end of Season 2.5
• "One Above All" Crest of Honor: Reach One Above All by the end of Season 2.5

Note that rewards are given out corresponding to the highest rank achieved during a season in Marvel Rivals. This means that as long as you reached, for example, Gold Tier 3, you'll get the Ultron skin regardless of whether you go up or down in rank by the end of Season 2.5 (end date TBC).

Marvel Rivals Ranked FAQs

(Image credit: NetEase)How does the Marvel Rivals Chrono Shield work?

The Chrono Shield is a unique mechanic in Marvel Rivals that can save you from losing points after losing a ranked game. This Chrono Shield will charge up as you lose games, and once it's fully charged, it'll shatter in place of a rank demotion, giving you another chance to claw things back.

Can you ban Heroes in Marvel Rivals ranked mode?

Once you get into a match where all players are at least Gold Tier 3 rank, you'll get the option to ban 2 Heroes before the game starts. This can counter some pesky meta-plays, like those using Hela to revive Loki and Thor.

What are the rules in Marvel Rivals ranked mode?

Game modes work a little differently in Marvel Rivals ranked when compared to Quick Play. Each game can have multiple rounds, and the team that wins the most rounds wins the game. You'll gain a point for each objective captured while attacking and then must make sure the enemy team scores fewer points while you're on the defensive side. If there's a tie, the match is decided on distance pushed in terms of Convoy, or with another set of rounds in terms of Domination.

(Image credit: NetEase Games)How to get your Marvel Rivals ranked rewards

Once you've hit the requisite rank to receive a reward, you'll just need to wait until the end of Season 2.5 to receive it. This is set for July 11, 2025. Unfortunately, this is the earliest you'll be able to get new reward skins, so it's best to just focus on climbing even higher before the season ends.

How do I rank up in Marvel Rivals?

To rank up in Marvel Rivals, you need to be playing Competitive Mode. For each win, you'll be awarded points, and for a loss, your point value will decrease. Once you hit 100 points total, you'll go up to the next rank.

You Might Also Like...

• Here's when to expect the next season of Marvel Rivals
• Best free games
• Marvel Rivals Mutant list: here's all of the playable Mutant Heroes

The recent announcement that Google is embedding a new AI mode into its search engine is yet another example of our daily exposure to AI. In theory, almost everything could be automated if we really wanted it to be, but getting the most out of AI tools within a workplace setting will always come down to thinking strategically.

What is the value it will create and how will it impact people? That’s not just employees or users, but partners and ultimately customers. If the value and purpose aren’t defined – and the roadmap to achieve success is unclear – then AI adoption is likely to fail.

From appetite to adoption

In a recent survey we carried out with over 4,000 UK workers, we found that the appetite to use AI within the workplace is there, with more than half of respondents (56%) feeling optimistic about the benefits. But when it comes to getting the most out of it, workers were less positive. 61% admitted to being overwhelmed by the rapid pace of AI advancements and that the sheer number of tools available make it difficult to know where to start (24%).

So how can the gap be bridged to ensure AI enhances workers’ lives and is used strategically? Although the pace of AI advancement is rapid, it doesn’t mean you need to throw best practice out of the window and adopt it without due diligence or thorough assessment. Knowing what to automate, and equally what not to automate, is about assessing the risk, setting objectives and putting measurable performance indicators in place. Only by having a clear goal for AI can businesses bring stakeholders along on the journey and achieve real value.

Reducing risk

Adopting AI brings different levels of risk – from operational to reputational. Without clear guidance for usage or contingency for failure, day-to-day business operations can be negatively impacted. Reputational risk and preservation of value must also be considered when looking to automate certain processes. For example, what experience do you want people to have when interacting with your business? What value do you want to create? What risks are you going to take, if something goes wrong, and how resilient is your business against these risks?

And what would you lose as a result of automation? It’s harder to define what shouldn’t be automated but it’s just as important as knowing what should. It requires reflection on what a business does and what it values. AI shouldn’t be about humans being replaced by machines but about how AI technology can be used effectively, to augment processes and support people to do their best work.

Tied in with assessing risk is ensuring responsible usage of AI. The OECD AI principles promote the use of AI that is innovative and trustworthy and respects human rights and democratic values, with an emphasis on human-centered values. By following these principles, organizations can not only adopt AI responsibly but have a basis on which to create clear guidelines for employee usage and expectations.

Transparency and trust

Setting a clear strategy and defining the right role for AI is not enough. The right culture needs to be created and any plans communicated effectively to those it will impact, to allay any anxiety around usage. AI comes with many connotations and preconceptions, so it’s important for organizations to remove fear of the unknown.

This can be done by creating a safe environment for people to explore what AI can do, with the support of clear guidelines on what tools are available to them. This will give users the confidence to use AI in their roles and remove any misalignment between perception and reality.

Dedicated AI training and guidance are crucial for the success of any initiatives. This is backed up by our survey in which a quarter (24%) of UK workers cited that a lack of training makes it hard to use AI effectively at work. Almost a half (49%) said their employer has no guidelines at all for using AI. When asked how inclined they would be to use or use AI more in the workplace if provided with training, the majority (60%) said they would be more inclined.

This suggests that many workplaces still lack positive facilitating conditions for AI usage, which could be leading workers to feel overwhelmed and under-supported when it comes to using AI tools within their role. Creating a culture of openness around AI usage will help build transparency and trust for users and ensure they can get value from it. Providing an environment for sharing experiences – positive and negative – will further support adoption. As not every organization has an ‘AI expert’, sharing these experiences will support best-practice creation and adoption across the organization.

Leading from the top down

Like with any new technology adoption, there will be short and long term value. ROI will not always be immediate. As such it’s important to effectively manage change and expectations. Any type of business transformation will create different ways of working and a period of adjustment, which can take time for users to adapt to.

It is therefore important that AI adoption initiatives are driven from the top, with senior managers providing clarity for users. Leaders need to make sure people understand the strategy behind the introduction of AI, the ultimate goal, and how AI supports it. This will help reduce uncertainties and fear of the unknown, and ensure both the business and users achieve value. Only through creating a culture of transparency, trust and continuous learning can AI augment workers’ lives rather than overwhelm.

We've featured the best productivity tool.

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

Samsung now sells two Galaxy Z Flip phones. One that costs $1,100 and another that's $900. There's a lot to unpack.

Samsung's newest book-style phone is impressively slim, at just 4.2mm thick when open and 215 grams. But the $2,000 price tag is heavy on the wallet.

From first-of-their-kind health tools to the return of the rotating bezel, here's what's new and noteworthy on Samsung's latest smartwatches.

Google's AI assistant lands on Wear OS 6, starting with Samsung's Galaxy Watch 8. Here are our first impressions, and what this update is bringing to your wrist.

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 Wednesday's puzzle instead then click here: Quordle hints and answers for Wednesday, July 9 (game #1262).

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 #1263) - hint #1 - VowelsHow 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 #1263) - hint #2 - repeated lettersDo any of today's Quordle answers contain repeated letters?

• The number of Quordle answers containing a repeated letter today is 4.

Quordle today (game #1263) - hint #3 - uncommon lettersDo 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 #1263) - 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 #1263) - hint #5 - starting letters (2)What letters do today's Quordle answers start with?

• C

• W

• U

• P

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

Quordle today (game #1263) - the answers

(Image credit: Merriam-Webster)

The answers to today's Quordle, game #1263, are…

• CROSS
• WHEEL
• UNDID
• PENCE

With E-N-C-E in position and just two possibilities for the first letter I picked the wrong one. In this situation I am sure most people would have chosen fence ahead of PENCE – but then again, that's probably only the case in the UK. 

These moments are what makes Quordle such fun – just when you think you’ve got all the answers it’ll kick you in the teeth.

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

Daily Sequence today (game #1263) - the answers

(Image credit: Merriam-Webster)

The answers to today's Quordle Daily Sequence, game #1263, are…

• SNIPE
• ALPHA
• BATHE
• GROUP

Quordle answers: The past 20

• Quordle #1262, Wednesday, 9 July: SHIRE, SIXTH, SINGE, IMAGE
• Quordle #1261, Tuesday, 8 July: PLIED, PRANK, GAWKY, OXIDE
• Quordle #1260, Monday, 7 July: DROLL, TRUCE, TWIRL, SINCE
• Quordle #1259, Sunday, 6 July: AMPLY, SPAWN, EXTOL, RIDGE
• Quordle #1258, Saturday, 5 July: HAVEN, SNAKE, DREAM, TORUS
• Quordle #1257, Friday, 4 July: FAVOR, SKUNK, GAWKY, FLUFF
• Quordle #1256, Thursday, 3 July: DANCE, EYING, GLAZE, EGRET
• Quordle #1255, Wednesday, 2 July: INANE, SCOUR, ELITE, ULCER
• Quordle #1254, Tuesday, 1 July: REBAR, YEARN, FORTH, CROWD
• Quordle #1253, Monday, 30 June: INGOT, INFER, TAPIR, CLUED
• Quordle #1252, Sunday, 29 June: MIRTH, APTLY, SCRUB, SMACK
• Quordle #1251, Saturday, 28 June: AWOKE, SMOKY, DEVIL, SWING
• Quordle #1250, Friday, 27 June: SPEAK, EAGLE, AVERT, SUING
• Quordle #1249, Thursday, 26 June: SLUMP, REBUS, GUAVA, MONEY
• Quordle #1248, Wednesday, 25 June: SOGGY, CLASH, MODEM, SQUAD
• Quordle #1247, Tuesday, 24 June: QUALM, SQUIB, AXIAL, FLACK
• Quordle #1246, Monday, 23 June: PIXEL, FJORD, STEAL, EXIST
• Quordle #1245, Sunday, 22 June: WREST, CRAZE, PAPAL, SCION
• Quordle #1244, Saturday, 21 June: BUDDY, GROUT, BEGIN, MADAM

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 Wednesday's puzzle instead then click here: NYT Connections hints and answers for Wednesday, July 9 (game #759).

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 #760) - today's words

(Image credit: New York Times)

Today's NYT Connections words are…

• DRAB
• BLOW
• MISS
• SPOKE
• TIRE
• DUCT
• WASTE
• FATIGUE
• EXHAUST
• RIM
• KHAKI
• HOSE
• LOSE
• PIPE
• HUB
• CAMO

NYT Connections today (game #760) - hint #1 - group hints

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

• YELLOW: Circular car component
• GREEN: Army garments
• BLUE: Let it flow
• PURPLE: A chance that slipped away

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 #760) - hint #2 - group answers

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

• YELLOW: PARTS OF A WHEEL
• GREEN: MILITARY CLOTHING DESCRIPTORS
• BLUE: USED IN VENTILATION
• PURPLE: FAIL TO TAKE ADVANTAGE OF, AS AN OPPORTUNITY

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

NYT Connections today (game #760) - the answers

(Image credit: New York Times)

The answers to today's Connections, game #760, are…

• YELLOW: PARTS OF A WHEEL HUB, RIM, SPOKE, TIRE
• GREEN: MILITARY CLOTHING DESCRIPTORS CAMO, DRAB, FATIGUE, KHAKI
• BLUE: USED IN VENTILATION DUCT, EXHAUST, HOSE, PIPE
• PURPLE: FAIL TO TAKE ADVANTAGE OF, AS AN OPPORTUNITY BLOW, LOSE, MISS, WASTE

• My rating: Easy
• My score: Perfect

After a terrible streak it felt good to complete a round of Connections without making a mistake.

I can only conclude that today’s edition was an unusually easy one, even the purple would be yellow-level some weeks.

My only pause for thought was over MILITARY CLOTHING DESCRIPTORS – CAMO, FATIQUE, and KHAKI all seemed common enough, while DRAB was a guess based on how dull army clothing is.

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

Yesterday's NYT Connections answers (Wednesday, July 9, game #759)

• YELLOW: PARTS OF A COLLEGE CAMPUS CAFETERIA, DORM, LIBRARY, QUAD
• GREEN: EXAGGERATED, AS A PERFORMANCE CAMP, DRAMATIC, HAMMY, OVERDONE
• BLUE: CLOSE IN ON CORNER, SURROUND, TRAP, TREE
• PURPLE: STARTS OF GREEK LETTERS BET, DELT, LAMB, THE

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.

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 Wednesday's puzzle instead then click here: NYT Strands hints and answers for Wednesday, July 9 (game #493).

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 #494) - hint #1 - today's themeWhat is the theme of today's NYT Strands?

• Today's NYT Strands theme is… Nap time!

NYT Strands today (game #494) - hint #2 - clue words

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

• DECK
• TIES
• SWOT
• WORK
• SPRAY
• GUIDE

NYT Strands today (game #494) - hint #3 - spangram lettersHow many letters are in today's spangram?

• Spangram has 10 letters

NYT Strands today (game #494) - hint #4 - spangram positionWhat are two sides of the board that today's spangram touches?

First side: left, 4th row

Last side: right, 5th row

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

NYT Strands today (game #494) - the answers

(Image credit: New York Times)

The answers to today's Strands, game #494, are…

• DROWSY
• SPENT
• WEARY
• FATIGUED
• TIRED
• KNACKERED
• SPANGRAM: SLEEPYHEAD

• My rating: Easy
• My score: Perfect

Yesterday it was party time! And today it’s nap time! Strands was obviously anticipating quite the knees up.

To be honest I am much more familiar with napping than partying, which is possibly why I made such quick work of today’s word search.

I have been feeling nearly every word in today’s search over the past few days. It’s not due to a lack of sleep or exhaustion due to over exercise, it’s because I am attempting to give up caffeine and am experiencing some quite radical side effects including feeling constantly DROWSY, SPENT, WEARY, FATIGUED, TIRED and zzzzzzzz....

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

Yesterday's NYT Strands answers (Wednesday, July 9, game #493)

• BALLOON
• BUNTING
• PENNANT
• CONFETTI
• STREAMER
• SPANGRAM: DECORATIONS

What is NYT Strands?

Strands is the NYT's not-so-new-any-more word game, following Wordle and Connections. It's now a fully fledged member of the NYT's games stable that has been running for a year and which 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.

• Rockerbox kept an open database online for an unknown period
• The database contained ID card numbers and other vital information
• Following its discovery, it has now been locked down

A tax credit consulting agency inadvertently exposed sensitive data on thousands of its customers by allegedly keeping a database filled with personally identifiable information (PII) open on the public internet.

It was discovered by Jeremiah Fowler, a cybersecurity researcher and analyst known for hunting for unencrypted and non-password-protected databases, and in a new vpnMentor report, Fowler said he found an archive with a total size of 286.9 GB, containing 245,949 records.

“In a limited sampling of the exposed documents, I saw files that detailed PII such as names, physical addresses, email addresses, DOB, and SSN in plain text,” Fowler explained. “There were also driver’s licenses, identification cards, SSN cards, work opportunity tax credit documents that included employment and salary information, and determination letters with acceptance or denials of eligibility.”

Rockerbox leaks

Furthermore, he observed DD214 forms - Certificates of Release or Discharge from Active Duty, issued by the US Department of Defense to veterans and similar military personnel. There were also password-protected PDF files labeled as “forms”, with file names containing PII such as employer names, and applicant first and last names.

Fowler attributed the database to a Texas-based company called Rockerbox, a tax credit consulting organization helping businesses increase their cash flow by identifying and managing employer-focused tax incentives through programs like the Work Opportunity Tax Credit (WOTC), Employee Retention Tax Credit (ERTC), R&D credits, and Empowerment Zone credits.

After reaching out to Rockerbox, the company closed down the archive in a matter of days, but allegedly never replied back to the researcher.

Therefore, we don’t know if the company manages this database, or if that work was handled by a third party - or if any threat actors obtained it in the past, but at press time, there was no evidence about in-the-wild abuse.

You might also like

• More than 3 million records, 12TB of data exposed in major app builder breach
• Take a look at our guide to the best authenticator app
• We've rounded up the best password managers

• Netflix has released a new trailer for Wednesday season 2
• The teaser arrives one month ahead of the hugely popular show's return
• Wednesday's sophomore season will be released in two parts

Netflix has released the official trailer for the highly-anticipated second season of Wednesday that sees the titular character in a race against the clock to save the life of her school bestie.

As the eponymous character and fans alike prepare for their sophomore year at Nevermore Academy, the streaming titan has debuted plenty of new footage via a new teaser. Check it out below:

Until now, we only knew that Jenna Ortega's protagonist would not only return to Nevermore, but also need to solve another hair-raising supernatural mystery. As the above teaser confirms, it'll center around Emma Myers' Enid Sinclair, whose life is in grave danger unless Wednesday can use her psychic abilities and detective skills to stop Enid from being murdered by something or someone.

After her exploits in season 1, Wednesday has also acquired an unwanted hero status and small group of fanatical followers. If she thought some of her fellow students were annoying last term, Wednesday's seen nothing yet.

Wednesday season 2's latest trailer arrives exactly one month before it debuts on the world's best streaming service.

Like many of Netflix's most-watched TV Originals, it'll launch in two parts. Volume 1, which comprises four episodes, releases on August 6, while the final four chapters, aka season 2 part 2, will air on September 3. Funnily enough, both dates fall on a Wednesday. It's almost as if Netflix planned this...

Enid's going to survive... right? (Image credit: Netflix)

This isn't the only footage Netflix has treated us to recently. In April, it unveiled the first trailer for Wednesday season 2. Then, at Tudum 2025 in late May, Netflix released the first six minutes of the series' second season, which showed the eponymous hero using her prescient powers to turn the tables on a serial killer in typically deadpan fashion.

We've also learned plenty more about one of the best Netflix shows' next installment in non-video form.

In mid-May, Netflix lifted the lid on some of the new and returning faces we'll see in Wednesday 2, the former of whom includes Joanna Lumley's turn as Grandmama and Steve Buscemi's new Nevermore principal. Those reveals came over a year after Netflix announced a Westworld alumnus would appear in Wednesday season 2 and the addition of Doctor Who, Star Wars, and Addams Family movie actors to season 2's cast.

A couple of weeks later, a sneak peek video also confirmed we'll be getting tons more Addams family lore next season. Considering we'll be seeing more of Wednesday's immediate family, such as Catherine Zeta-Jones' Morticia and Isaac Ordonez's Pugsley this time around, that's to be expected.

For even more details on the show ahead of its return, read our ultimate guide to Wednesday season 2.

You might also like

• Netflix might have another Suits-like hit on its hands as unlikely crime thriller makes its way onto the popular streamer's most-watched shows list
• Netflix just added my favorite sci-fi thriller series in July – here’s why you should make it your next binge watch
• I watch Netflix for a living – here are the 7 new anime shows you need to to be ready for

Foundation season 3 is almost here. The critically-acclaimed Apple TV Original's latest installment will arrive in less than 48 hours, so you'll want to know exactly when you can stream it.

Below, I'll tell you when its first episode will be available to watch on Apple TV+. You'll also find a full release schedule for Foundation's third season at the end of this article, so you can make block out some time in your busy schedule to view them as soon as they air.

What time can I stream Foundation season 3's first episode on Apple TV+?

Gaal Dornick will return in Foundation's third season (Image credit: Apple TV+)

Foundation season 3 will premiere on one of the best streaming services on Friday, July 11. It should do so globally, too. I say "should" because, like other entries in our best Apple TV+ shows guide, it could arrive earlier in some nations, such as the US.

Occasionally, new episodes of Apple TV Originals actually launch late on Thursdays in certain time zones. Viewers in some North and South America regions, then, may get to see them early.

So, what time will season 3 episode 1 come out? Apple doesn't usually release information about episodic launch times. However, based on when other TV Originals, including Severance, have landed on Apple TV+, here's when we expect Foundation season 3's first episode to be released (NB: if your country isn't listed below, use 12am ET to work out when you can stream it):

• US – Thursday, July 10 at 9pm PT / Friday, July 11 at 12am ET
• Canada – Thursday, July 10 at 9pm PT / Friday, July 11 at 12am ET
• UK – Friday, July 11 at 5am BST
• India – Friday, July 11 at 9:30am IST
• Singapore – Friday, July 11 at 12pm SGT
• Australia – Friday, July 11 at 2pm AEST
• New Zealand – Friday, July 11 at 4pm NZST

Foundation season 3 full episode release schedule

Jared Harris is also reprising his roles as Hari Seldon and Doctor Seldon in Foundation season 3 (Image credit: Apple TV+)

New episodes of Foundation season 3 will land on Apple TV+ every Friday. Read on for a full breakdown on when to expect new chapters:

• Foundation season 3 episode 1 – Friday, July 11
• Foundation season 3 episode 2 – Friday, July 18
• Foundation season 3 episode 3 – Friday, July 25
• Foundation season 3 episode 4 – Friday, August 1
• Foundation season 3 episode 5 – Friday, August 8
• Foundation season 3 episode 6 – Friday, August 15
• Foundation season 3 episode 7 – Friday, August 22
• Foundation season 3 episode 8 – Friday, August 29
• Foundation season 3 episode 9 – Friday, September 5
• Foundation season 3 episode 10 – Friday, September 12

For more Foundation coverage, read my Foundation season 3 review or check out the below section for more from my exclusive chats with its cast ahead of this season's debut.

You might also like

• 'It cracks me up': Foundation actor Lee Pace teases Brother Day's The Big Lebowski-like transformation in season 3
• 'He needs to be a big threat': Foundation season 3 star Pilou Asbaek reveals how The Mule differs from his book series counterpart
• 'We're the last to find out': Foundation stars don't know if it'll return for a fourth season

• Germany is calling on Apple and Google to block the DeepSeek app from the country's official app stores
• The Berlin Data Protection Commissioner has deemed the Chinese AI chatbot "unlawful" over alleged privacy violations
• Italy banned DeepSeek from official app stores in January over similar concerns

DeepSeek could soon disappear from Apple and Google's official app stores in Germany as data protection officials accuse the Chinese chatbot of alleged privacy violations.

"DeepSeek's transfer of user data to China is unlawful," said Berlin Data Protection Commissioner Meike Kamp, in an official announcement dated June 27, 2025. Kamp has called on the Big Tech giant to consider blocking the app in the country.

Another EU member, Italy, already banned Deepseek from the Apple App Store and Google Play Store in January 2025 over similar grounds. The block was enforced about a week after the release of the ChatGPT rival.

Does DeepSeek violate GDPR rules?

(Image credit: Pixabay)

According to German authorities, the company behind DeepSeek AI (Hangzhou DeepSeek Artificial Intelligence Co., Ltd) violates Art. 46 (1) of the GDPR, which rules the need for "appropriate safeguards" when transferring EU citizens' personal data to a third country.

According to Kamp, DeepSeek failed to convince German officials that users' data is protected when these details are transferred to China, as expected by EU laws.

"Chinese authorities have far-reaching rights of access to personal data within the sphere of influence of Chinese companies. Furthermore, DeepSeek users in China do not have the enforceable rights and effective legal remedies guaranteed in the European Union," Kamp added.

These concerns echoed similar GDPR complaints issued after examining the platform's privacy policy and finding "multiple violations" of European rules.

Did you know?

(Image credit: Adobe Stock)

The US is currently pushing for a new bill, known as the 'No Adversarial AI Act,' to ban all AI models built in China, Iran, Russia, or the Democratic People’s Republic of Korea from being used in government agencies – DeepSeek included.

It's worth noting that DeepSeek updated its privacy policy on July 4, 2025, and now states that the company does not engage in 'profiling,' which was among the issues raised by EU privacy watchdogs. That said, other concerns previously raised, such as non-transparent information on data retention periods, may remain.

On May 6, 2025, German data protection officials asked the Chinese firm to "independently remove its apps from the German app stores, cease the illegal data transfer to China, or fulfill the legal requirements for lawful third-country transfers," the announcement reads. A request that, said German officials, Deepseek failed to comply with.

Kamp has then decided to exercise Germany's right under Art. 16 of the Digital Services Act (DSA) to report what he believes to be "illegal content" to app stores' operators.

She said: "I have therefore informed Google and Apple, as operators of the largest app platforms, of the violations and expect a timely consideration of a blocking."

While Italy was the first to launch an investigation into DeepSeek's data privacy and security practices on January 28, 2025, France, the Netherlands, Luxembourg, and Portugal are also investigating DeepSeek's policies over potential GDPR violations.

Can you use a VPN if DeepSeek gets blocked?

It's still early to say whether the popular Chinese AI chatbot will leave the German market for good.

What we do know, though, is that DeepSeek is increasingly under fire worldwide, meaning that some of its nearly 97 million active users may need to find an alternative way to download, update, and use the app.

Using the best VPN services is generally an easy way to bypass online geo-restrictions, thanks to their IP-spoofing capabilities that let you look like you're browsing from a completely different country in no time.

That said, like with the short-lived US TikTok ban, a VPN isn't a one-click solution if DeepSeek gets banned on official app stores.

That's mainly because using a VPN doesn't spoof your location linked to your app stores. This means that you'll need to "find another way of downloading the app other than the Apple App or Google Play stores," Eamonn Maguire, Head of Account Security at Proton – the provider behind Proton VPN – told me back in January.

However, experts suggest doing this with caution as potential privacy concerns may remain.

You might also like

• New bill could see DeepSeek and Chinese AI models banned in government departments
• Worried about DeepSeek? Well, Google Gemini collects even more of your personal data
• German consumer protection group calls on Meta to halt its AI training in the EU

• 33-year-old man was accused of spying for the Chinese government
• FBI alleges he is part of the Silk Typhoon hacking collective
• If convicted, he might be looking at decades in jail

Italian law enforcement has arrested a 33-year-old Chinese national for allegedly spying on the United States.

Landing at Milan’s Malpensa airport on a flight from China, Zewei Xu was apprehended by local police, and according to Italian news agency ANSA, is wanted by the FBI for allegedly participating in cyber-espionage operations for China, targeting data surrounding anti-COVID vaccines that were being produced at the University of Texas back in 2020.

Citing “interior ministry documents”, ANSA said Xu is accused of being part of Hafnium, an infamous state-sponsored threat actor also known as Silk Typhoon. With this group, Xu allegedly “targeted thousands of computers around the world” to get information on “various US government policies.”

Get 55% off Incogni's Data Removal service with code TECHRADAR

Wipe your personal data off the internet with the Incogni data removal service. Stop identity thieves
and protect your privacy from unwanted spam and scam calls.View Deal

Typhoons against critical infrastructure

There will be a hearing early next week, at Milan’s Court of Appeals, to determine whether or not Xu will be extradited to the United States.

While his family claims he is an IT manager at Shanghai GTA Semi Conductor, where he develops IT systems and network infrastructure, American authorities accuse Xu of wire fraud and aggravated identity theft.

If convicted, Xu is looking at a maximum punishment of 20 years in prison, as well as an additional five years for unauthorized access to protected computers.

Silk Typhoon is one of many “typhoon” groups (Flax Typhoon, Silk Typhoon, and others), all of which are apparently state-sponsored and engaged in various forms of cybercriminal activity.

Critical infrastructure firms, government organizations, telecommunications companies, and similar, are the typhoons’ most common targets.

In mid-April 2025, amid a serious escalation of hostilities between the US and China, senior Chinese officials apparently acknowledged behind closed doors that Beijing was involved in a series of cyberattacks on US critical infrastructure, conducted by Volt Typhoon.

This group was infiltrating US critical infrastructure systems for years, including compromising energy, communications, transportation, and water industries.

Via BleepingComputer

You might also like

• CISA tells agencies to patch BeyondTrust bug now
• Take a look at our guide to the best authenticator app
• We've rounded up the best password managers

• A new leak suggests the Sony WF-1000XM6 are arriving soon
• There might be a slight tweak in terms of the dimensions
• It's been two years since the Sony WF-1000XM5 earbuds launched

We've recently seen the flagship Sony WH-1000XM6 over-ear headphones launched, but what of the wireless earbud equivalents? Well, a new leak suggests the Sony WF-1000XM6 aren't far away from being made official either.

As per The Walkman Blog (via Notebookcheck), import documentation has surfaced that mentions the Sony WF-1000XM6 earbuds by name – describing them, somewhat strangely, as "headphones without a frame".

We don't get too much in the way of detail here, but there is a reference to the charging case in the documentation, and we get the dimensions of the earbuds: it looks as though the new pair will be a little smaller overall than the Sony WF-1000XM5.

The dimensions listed in the leak are 84mm x 70mm x 55mm for the Sony WF-1000XM6, which compare to 99mm x 66.4mm x 53mm for the Sony WF-1000XM5, but we don't get any hints of a redesign here.

Will the price be right?

Sony just launched the WH-1000XM6 headphones (Image credit: Future)

Our list of what we want to see from the Sony WF-1000XM6 earbuds mentions better noise cancellation, improved connectivity, better battery life, an upgraded IP rating, more color options, and (of course) a cheaper price.

Head to our Sony WF-1000XM5 review and you'll see those earbuds launched at a starting price of $299.99 / £259 / AU$499. That was back in July 2023 though, and the tech industry and the world in general has changed quite substantially since then.

It doesn't look as though another July launch is likely for 2025, but based on the manufacturing stages mentioned in these import filings, the earbuds should break cover in the coming months – so most probably before the end of the year.

Whenever the Sony WF-1000XM6 make an appearance, they're going to face stiff competition in what is one of the most crowded categories in tech. We're still waiting for the AirPods Pro 3, for example, but they may not show up until next year.

You might also like

• I compared the Sony WH-1000XM6, XM5 and XM4 headphones
• These are the best wireless earbuds you can buy
• Audio-Technica's new earbuds have a mind-blowing upgrade