Feed aggregator

• Check Point Research exposed coerced insider flaw in ChatGPT’s agent architecture
• Containers shared metadata via internal service, enabling cross‑account prompt injection and data theft
• OpenAI closed the path, but CPR warns similar risks may exist in other AI platforms

ChatGPT’s AI agents were allowed to pull sensitive data from one account shared with an entirely different account because, colloquially speaking, all agents used to walk down the same hallways, experts have warned.

A new report from security experts Check Point Research (CPR) dubbed the flaw “coerced insider”, since it revolves around persuading the agent instead of abusing a vulnerability.

Coerced insider

When an AI agent is given a task that needs code execution, it handles that task in an isolated container which also sometimes needs to install software. To enable that, without giving containers direct internet access (which would be too risky), OpenAI routes those packages through an internal JFrog Artifactory instance. As a separate security contingency, containers from different accounts cannot talk between themselves.

However - they can reach the same internal service (our proverbial hallways), which exposes an item management feature that lets the containers attach text or binary properties to a repository item. As a result, any container can read back the properties written by any other container.

“Check Point Research confirmed the isolation gap directly: a property written from one account’s container was fully readable from a different account’s container moments later, with data too large for one property simply split into chunks and reassembled on the other end,” the researchers explained.

“The package delivery metadata effectively became a shared clipboard between containers that were supposed to be walled off from one another.”

From there, the exploit turns into your usual, off-the-shelf prompt injection. The only difference is that the malicious prompt is not delivered directly to the victim, but rather left in the hallways, and the results are not shared with the attackers directly, but rather left in those same proverbial hallways, too.

The attacker then delivers a prompt or a shared conversation that instructs the agent to check the same storage during its next ordinary reply. The agent checks, sees the malicious instructions, executes them, all the while replying to the victim’s question in the usual manner. The victim is oblivious to the fact that data theft is taking place in the background.

What kind of information can be stolen?

But this is just half of the equation. How destructive this attack ends up being still depends on the amount of data being shared with the victim agent.

The bare minimum is the information shared while chatting to the agent. It then grows with every connected app: Gmail, Google Drive, Microsoft Teams, GitHub, and similar. “In Check Point Research’s demonstration, ChatGPT retrieved the victim’s email data through their connected Gmail account and delivered it to the attacker’s session, all within a single ordinary turn,” CPR stressed.

The good news is that you’ll likely never be exposed this way, at least not via ChatGPT. CPR says it disclosed the findings to OpenAI, who then confirmed that the specific internal Artifactory instance identified in the research has been commissioned. In other words, the hallways attack path has been closed.

The bad news is that this doesn’t automatically mean everyone’s safe. This particular path might be closed, but the architectural pattern behind the flaw could be present in other platforms, CPR warns.

“Any AI assistant that operates inside an organization’s trust boundary, holding credentials, running code, and reaching connected services, can become what Check Point Research calls a coerced insider,” the report states. “The model itself does not need to be malicious. It only needs to be persuaded, through text it was never meant to trust, to use access that was granted for entirely legitimate reasons.”

Going forward, businesses are advised to learn which AI tools their employees are using, and what those tools are connected to. Then, they should govern what AI tools and agents are allowed to do, treating all of their actions (not just output) as something that needs to be monitored.

Bangladesh has seen over 50,000 confirmed or suspected cases of measles this year. The death toll among children just surpassed 1,000. How are officials trying to bring the outbreak under control?

(Image credit: Ali Asif Shawon for NPR)

• CD Projekt Red says The Witcher 3 Remastered and its many upgrades helped realize Songs of the Past
• Game director on Songs of the Past, Jakub Rokosz, says the team had a "certain vision" for the expansion, but there were technological challenges
• He says the remaster "helped with applying optimization" to the expansion

According to CD Projekt Red, The Witcher 3: Wild Hunt's upcoming expansion, Songs of the Past, may not exist without The Witcher 3 Remastered and its many upgrades.

If you missed the bombshell news the Polish studio dropped during Opening Night Live, The Witcher 3 Remastered is a massive free update arriving on September 29. It offers many improvements to the base game and the Hearts of Stone and Blood and Wine expansions, such as upgraded visuals, a refined combat system, new Witcher Signs, transmogrification, and more.

The remaster will essentially rework the game in its entirety, giving fans of the 11-year-old role-playing game (RPG) an excuse for another playthrough, ahead of the upcoming expansion Songs of the Past.

Now, in an interview with TechRadar Gaming at Gamescom, Jakub Rokosz, CEO of Fool's Theory and game director on Songs of the Past, and senior quest designer Despoina Anetaki shared that without these technological improvements, their vision for the expansion may not have been fully realized.

"The stars had to align, so to say, because a few things had to happen," said Rokosz, who was also one of the original senior quest designers for The Witcher 3. "The Songs of the Past expansion and Letten itself, the area that it takes place, it had a certain vision. We had a certain vision for it, and it posed us certain specific technological challenges that was impossible to realize without the remaster changes.

"So the conditions had to grow for the remaster to appear on the horizon for us to be able to realize this, so this is actually two things meshing together to make it a reality."

The developer explained that Letten's lush vegetation, small bushes, and dense forests caused issues during development, but were resolved with the remaster's advancements.

"One of the challenges when we started prototyping the area was [that] it was impossible to ride with a horse through it, basically," he continued. "So we had to do some improvements on the horse handling side, right? So it feels more organic or more natural."

Letten also features a big lake in the middle of its map that allows players a view of the horizon, and as Rokosz explained, having this huge amount of open space where you can see fishermen fishing "posed a challenge in terms of how the engine was handling this. So the remaster helped with applying optimization."

Songs of the Past is due to launch in 2027.

A Trump census proposal would stop asking about race and exclude millions of U.S. residents from a count that the 14th Amendment says must include the "whole number of persons in each state."

(Image credit: J. Scott Applewhite)

• The Ninja Crispi DualZone is an air fryer designed to cook two foods at once
• It has two glass containers that you can take from fridge or freezer to table
• It's on sale in the UK today (September 9) for £299.99

Ninja has launched another version of the Crispi glass air fryer — this time with a pair of cooking zones so you can prepare two different foods at once, and time them to finish at the same time so your meal is ready to serve with no effort on your part.

Like the original portable Ninja Crispi and the family-sized Ninja Crispi Pro, the Ninja Crispi DualZone Air Fryer has non-toxic glass dishes with resealable lids, which you can use to store, cook, and serve food without any mess. The two CleanCrisp containers have a capacity of 3.8 litres each, and are thermal shock-resistant so you can start cooking your food direct from the freezer, and immediately place the dishes on cold stone work surfaces once your meal is ready.

Like the Ninja Crispi Pro, the Crispi DualZone has eight cooking modes (Air Fry, Bake, Max Crisp, Roast, Dehydrate, Re-crisp, Keep Warm, and Prove), and the twin cooking zones mean you can use two different settings at once.

SharkNinjaSharkNinja

Once you're done, the glass containers, crisper plates, and lids are all dishwasher-safe, unlike conventional air fryer baskets, which require careful hand-washing to avoid damaging their non-stick coating.

The Ninja Crispi Dual Zone air fryer is available to buy now direct from SharkNinja for £299.99 in the UK, and comes in three colours: Cyberspace (dark blue), Bone (off-white), and Rose Quartz (pastel pink). We'll be putting it to the test soon to see how it compares with traditional dual-zone air fryers, and the rest of the Ninja Crispi range.

• Qualcomm to help build custom silicon and optical networking for AWS data centers
• Amazon will acquire 25 million Qualcomm shares, worth over $4 billion
• AWS data centers are more efficient than the industry average, but behind Google's

Qualcomm and Amazon have announced what they have describing as a "multi-generational product collaboration," in which Qualcomm will help build future AI data center infrastructure for AWS.

Because this isn't a one-off project, instead of announcing a single component, the companies are likely to work together across multiple generations of infrastructure spanning both custom inference silicon and high-speed optical connectivity.

Together, the two companies are expected to combine their expertise in AI infrastructure and data centers to improve performance, power efficiency and cost efficiency for a future where data centers exist in even higher volumes.

AWS and Qualcomm to work together on future data centers

The project spans two key areas – custom AI silicon and networking hardware. On the silicon front, Qualcomm will help produce custom AWS silicon with a major focus on inference, giving the chipmaker a potentially big entry into the data center semiconductor market – one that Nvidia has seen major success in.

As for networking, Qualcomm will be behind high-performance optical connectivity for AWS data centers, covering connections of up to 1.6T initially but plans to scale.

With data centers under intense pressure from all angles to boost efficiency and reduce resources consumption, operators are being forced to make progress in all areas, and reducing losses between components with efficient networking is often overlooked compared with outright chip efficiency.

"As AI demand accelerates, data center infrastructure will require advances in both computing and connectivity to deliver greater performance with more efficiency,” Qualcomm CEO Cristiano Amon summarized.

In Amazon's 2025 sustainability report, it declared a global average PUE of 1.14, which was nearly 9% better than the public cloud industry average of 1.25. Still, Google says it managed a fleet-wide average of 1.09, so Amazon's work with Qualcomm could be key to it overtaking Google.

"By working together on customized silicon and advanced connectivity, we’re delivering more performant, efficient, and cost-effective infrastructure for our customers," AWS VP Prasad Kalyanaraman added.

As part of the deal, Qualcomm has issued warrants for Amazon to acquire around $4.03 billion in stock, per a SEC filing – the equivalent of 25 million shares priced at $161.26. Qualcomm is currently trading at $174.09, up nearly 5% in five days.

• The final season of Call of Duty: Black Ops 7 has been revealed
• It includes a Doom crossover and lots of Halloween content
• Only one new multiplayer map is being added, however

Activision has lifted the curtain on the content coming to Call of Duty: Black Ops 7 in its final season ahead of the Modern Warfare 4 launch.

With a new game on the way, it shouldn't come as much of a surprise that this update is going to be a smaller one. Still, it includes plenty of spooky Halloween content as part of the series' annual The Haunting event.

This includes a bombastic Doom crossover, allowing players to unlock loads of Doom-themed rewards in a new limited-time event pass. As usual, this pass will be divided into two tracks: a free one and a premium option.

The free track contains new operator skins for Wei Lin and Falkner, plus an adorable Mini Slayer weapon charm and a range of other rewards. Completing the entire track will grant you the Roc 20mm special weapon — a massive, almost cannon-like rifle that fires devastating 20mm rounds.

The premium track, on the other hand, unlocks the Doom Slayer himself as a playable operator in the JSOC faction. By default, he's wearing his Doom: The Dark Ages attire, but the premium track's final reward is an alternate classic 1993-style skin.

(Image credit: Activision)

Other weapons introduced in the upcoming season include the TR41 Para Assault Rifle, which will be available at launch, and the Arcblade melee weapon added later on. There are also some new weapon attachments on the way for the VS Recon Sniper Rifle, Dravec 45 SMG, and Sokol 545 LMG.

Unfortunately, the offering is pretty sparse on the map front, with only one new one being introduced. It's called Quantum, and it looks like an appropriately ruined laboratory. There is at least also a new version of the existing Homestead map, called Bloodstead, for you to explore as well.

The cool Hordepoint mode will be returning from previous games in season too, challenging players to face off against a rival team and spawning waves of zombies.

All of this comes alongside a fresh battle pass full of new skins and unlocks, not to mention additional content for the Zombies and Endgame modes, plus the free-to-play Warzone battle royale.

Call of Duty: Black Ops 7 is available now for PC, PlayStation 5, PlayStation 4, Xbox Series X and Series S, and Xbox One.

• Invoke Studios head Dominic Guay says Warlock: Dungeons & Dragons' imp companion Skev provides narrative and gameplay assistance
• Skev will provide lore to the player as they explore Darkon, but can't be controlled in combat
• Guay says the imp is "his own character" with his own personal motivations

Warlock: Dungeons & Dragons' imp companion Skev serves multiple purposes for the player, but was designed to be "his own character."

In an interview with TechRadar Gaming at Gamescom, Invoke Studios head Dominic Guay offered more insight into Skev and his relationship with the warlock protagonist Kaatri, revealing that the imp will assist the player as they navigate the open world of Darkon both narratively and with his utilities.

However, while Skev will act as a companion to Kaatri, Guay stressed that he still has his own motivations, including seeking out his lost master, just like how Kaatri is searching for someone too.

"We iterated a lot on him. So he actually is his own character," Guay said. "That was important to us. He wasn't born to be a sidekick. That was the mentality. He used to be the advisor to the ruler of Darkon, so in a way, he had a really important role in that land before his master vanished, and his quest is to find his lost master, and Kaatri's quest is to find someone too.

"So, he tags along, and Kaatri's kind of like, 'Oh well, he knows this place, and that's useful.' So he gives you a lot of insights as you travel around. You find new places."

The studio head said Skev will offer lore about new locations as players explore, adding that he thinks of the imp as "your unreliable guide to Darkon."

Skev won't be controlled by the player either, Guay confirmed, so he won't die in combat or get in the way, but will sometimes help in certain instances, like by picking locks.

"You can actually decide how you want to upgrade his abilities throughout the adventure," he added, "So [there's] some practical application."

Warlock: Dungeons & Dragons arrives in 2027 for PS5, Xbox Series X, Xbox Series S, and PC.

• iRobot has unveiled its new flagship Roomba Max 875 Combo
• It sprays out heated mist to dislodge dried dirt from hard floors
• There's a seal around the vacuum to maximize pickup power

iRobot has revealed its latest flagship robot vacuum, and I've just seen it in action at IFA 2026. The Roomba Max 875 Combo is a combination robovac-and-mop, and boasts some intriguing features on both of those floorcare prongs.

One that caught my attention in particular is the mopping setup. Robot vacuum mopping has historically been rather underwhelming, but the latest models are showing some significant improvements. iRobot's Roomba Max 875 Combo's USP is its 'ThermaMist' function. I got the low-down from Adam Pope, Chief Engineer & Vice President of Product Planning & Development at iRobot.

"The ThermaMist is a pretty interesting technology," he tells me. "It's an ultrasonic warm mist, meant to pre-treat stains. So it'll detect stains and pre-treat them before it mops. The customer can also activate it in a certain area if they want to just do that whole area.

"The heat gives you a little better power to break up some of the oily or sugary stains that might be sticky. So that's the reason that we have the heat in there, really, try to help break those up."

Unlike on Dreame's new mist-expelling robovac, which heats the water on the robot itself, here the heating happens in the dock. From an engineering perspective, what's interesting is that the on-bot process of creating ultrasonic mist also naturally generates warmth.

"The dock can put hot water in the robot, but the actual ultrasonic ThermaMist part of that process, it does actually heat the water as it's coming out," explains Pope. "When [the water] goes through that process, it comes out warm, even if there's room temperature water in the robot."

The fact there's no active heater on the robot itself is good news for battery life. "With the ThermaMist on, it's going to have a small impact on the battery life, but not a large one, because the process of heating is part of the ultrasonic component," Pope confirms. "It's not like there's a special heater that you're running all the time."

Unfortunately, iRobot wasn't demoing the ThermaMist feature when I visited the booth. However, based on the video (above) I was sent over later, it looks promising — there's a good amount of mist created, and it's being delivered evenly. In theory it sounds like a truly interesting innovation, and one that has plenty of potential to deliver strong mopping results.

(Image credit: Future)

iRobot has also extended the mop, which is larger in diameter and spins faster than its predecessors, for extra scrubbing power. It can kick out to one side to get close to the edges of rooms, and there's a shield that will slide under the roller as extra protection against unwanted sogginess when the bot is moving over carpet.

Sealing the deal

There's interesting innovation on the vacuuming front too. iRobot has added a 'SealForce' system, which is essentially a kind of skirt around the suction inlet on the Max 875 Combo, which lowers to form a loose seal against the carpet fibers.

"The SealForce brings a component of the cleaning head down to the floor, so you get all that airflow going right through the deep fibers of the carpet, trying to pull everything deeply out of the carpet, or even out of crevices, grout lines, that sort of thing," says Pope. "It's really kind of applying the suction where you need it most."

In the demo, the 'SealForce' seal wasn't as dramatic as I had been expecting. You can see the moving frame around the suction inlet below, and in the right-hand photo it's about as extended as it gets. That small shift could be enough to make a big difference to the suction, through — we'll need to test it out to find out.

(Image credit: Future)

iRobot is traditionally shy about publishing its suction specs, but here it has done: the Roomba Max 875 Combo offers 35,000Pa, which is right up there with the highest numbers you'll find anywhere. I'm increasingly of the opinion that it doesn't really matter anyway. The suction is only as effective as the wider design you put it into — and if it works as intended, this seal could make a big difference to the Max 875's pickup power.

Saving grace

The past few years have been a rocky time for once-supreme robot vacuum brand iRobot. Its bots have fallen behind the pack when it comes to features, so it's exciting to see the company starting to find its footing again when it comes to innovation.

The iRobot Roomba Max 875 Combo is on sale now in the UK, priced at £1,099. (equivalent to about $1,490 / AU$2,060). iRobot has a presence in Australia, but I don't yet have launch details for the 875 Combo there.

What about the US? Due to the new FCC ban on foreign-made "advanced" robots, the Max 875's launch is on pause for now. iRobot was founded in the US, but at the start of the year it was taken over by Chinese manufacturing company Picea. However, this takeover isn't to blame for the fact the brand can't now sell its new products in the US; the ban refers to where the product is manufactured, and Roomba bots have long been made outside of the US.

In terms of future plans, Pope has this to say: "We're working through with our advisors on how we continue to sell in the US. So we have multiple plans that we're working through and multiple strategies that will come to fruition, and we'll still be able to bring cool technology to the US as well."

DDR5 might be the memory standard for newer PCs, but it costs far more than DDR4, especially when you're shopping for faster or higher-capacity kits.

If you're upgrading a DDR4 system, I've found a terrific deal on the G.Skill 32GB Trident Z RGB for $215 (was $258) at Newegg when you use code LUYRF222.

For your money, you get two 16GB DDR4-3600 modules with CL18 timings, aluminum heat spreaders, customizable RGB lighting and Intel XMP 2.0 support.

Today's best DDR4 RAM deal

32GB TridentZ RGB Series DDR4 3600 RAM: was $257.49 now $214.99
This memory kit provides 32GB of dual-channel DDR4-3600 RAM across two 16GB modules, with CL18 latency, 18-22-22-42 timings, aluminum heat spreaders, RGB lighting and XMP 2.0 support. For the full discount, use code LUYRF222View Deal

Why we recommend it

32GB is enough memory for demanding multitasking, gaming and creative applications. The DDR4-3600 speed is also a good fit for compatible DDR4 systems, particularly if you're upgrading an older machine that's currently running slower or lower-capacity memory.

The two matched 16GB modules provide a dual-channel configuration, with timings of 18-22-22-42 at 1.35V.

G.Skill's Trident Z design adds aluminum heat spreaders and customizable RGB lighting, making this a good choice if your PC has a windowed case and you want your memory to complement other RGB components.

Price context & historical value

Newegg sells the 32GB Skill Trident Z RGB kit for $257.49, and it's currently available for $249.99. Using coupon code LUYRF222 brings the final price down to $214.99.

While that's a decent price in today's inflated memory market, it's worth remembering just how much cheaper DDR4 used to be. Newegg sold this exact G.Skill kit for just $69.99 in August 2023, less than a third of its current discounted price.

Unfortunately, it's unlikely to drop down to that level again, so the current deal is worth considering if you need to upgrade your DDR4 system now.

Should you buy it?

✅ Buy the 32GB G.Skill Trident Z RGB kit if...

You have a compatible DDR4 desktop and want to increase its memory capacity without replacing your motherboard and processor to move to DDR5. The combination of 32GB capacity and DDR4-3600 speed makes it a solid upgrade for an older system.

❌ Skip the 32GB G.Skill Trident Z RGB kit if...

You're building a completely new PC around a recent processor and motherboard. In that situation, I'd seriously consider DDR5 instead, as you'll be investing in the newer memory standard rather than spending more than $200 on DDR4 that you won't be able to carry over to a future platform.

The Catch: What to know before you buy

You'll need to enable XMP, DOCP or A-XMP in your BIOS to reach the kit's advertised DDR4-3600 speed on compatible hardware. Otherwise, the memory will initially run at its default SPD speed.

Compatibility is also important. G.Skill recommends checking its motherboard QVL, and it specifically warns against mixing separate memory kits, even when their specifications appear identical, as doing so can cause stability problems.

• Shop all RAM at Newegg

The U.S. military says it has destroyed five Iranian oil tankers as the conflict continues to escalate. And, Smithsonian head Lonnie G. Bunch III says he plans to step down by the end of the year.

(Image credit: Tasos Katopodis)

For the past decade, quantum computing has been relegated to the horizon, a distant, theoretical challenge rather than an immediate operational imperative.

Breakthroughs have generated headlines, but without a firm deadline for when quantum computers could threaten today’s cryptographic standards, CISOs have had little incentive to make quantum readiness an immediate priority. IBM’s 2025 Quantum-Safe Readiness Index shows how much work remains, with the average organization scoring just 25 out of 100.

Yet, that window of preparation is contracting rapidly. This spring, Google warned that “quantum frontiers may be closer than they appear,” pointing to advances in hardware, error correction, and algorithms that are accelerating progress toward cryptographically-relevant quantum computers. Google has since set a 2029 target for completing its own migration to post-quantum cryptography (PQC).

Yet, that growing urgency still comes without a firm deadline. Unlike the fixed, binary deadline of Y2K from years ago, the quantum risk profile is nebulous; the exact threshold of peril remains undefined, even as the required remediation timeline spans several years.

Large enterprises need time to identify cryptographic dependencies, assess legacy infrastructure, coordinate vendors, and secure the budgets required for migration. Those decisions reach well beyond cryptography, affecting how the business plans, funds, and prioritizes risk. In fact, the availability of quantum computing capabilities will touch almost every aspect of companies around the globe.

Turn quantum risk into an enterprise-wide priority

Quantum readiness already has a prescribed technical solution. In 2024, NIST finalized its first set of PQC standards and encouraged organizations to begin transitioning as soon as possible. For some systems, migration may just involve a software update or configuration change. For others it may be much more complex. The challenge for a large enterprise is knowing where those changes need to happen.

A practical first step is to build a cryptographic bill of materials (QBOM), cataloging what cryptography is in use and where, which systems can be updated relatively easily, and where legacy infrastructure may require replacement or another mitigation strategy. This is an intensive effort, but there are software platforms that can help.

Next, organizations should prioritize systems based on the value of the data and how long it needs to remain protected. Financial records or customers’ customer personally identifiable information (PII), for example, may warrant greater urgency than an internal chat log with little long-term value.

Data longevity is especially important because adversaries are already collecting encrypted traffic in the expectation that future quantum computers will be able to read it, a practice known as "harvest now, decrypt later." For sensitive data, the attack may already have happened even though the business impact hasn't.

Data exposure is only one part of that assessment. Public key cryptography (PKC) also underpins authentication, including single sign-on and other mechanisms used to establish identity. If that cryptographic foundation can no longer be trusted, organizations face the risk of impersonation, with consequences for access to applications, infrastructure, and data.

Much of this work sits outside the CISO's direct control. Infrastructure, applications, data, and third-party relationships are often jointly owned across the business, and so are many of the budgets needed to address them. CISOs must make a business risk pitch that shows other leaders – the CIO, CTO, procurement, compliance, and executives – where the greatest risks are and what needs attention first.

Make crypto-agility part of the migration

Preparing for a post-quantum future also requires crypto-agility. Cryptographic standards will continue to evolve as new vulnerabilities emerge and algorithms change. Organizations need systems and platforms that can accommodate those changes without another costly, multi-year migration. That means buying and building in ways that can accommodate a future algorithm change through software rather than hardware rip-and-replace.

The UK's National Cyber Security Centre recommends building that flexibility into PQC migration plans and establishing criteria for retiring traditional algorithms, with the goal of removing sole dependence on traditional public-key cryptography. It also means rethinking how organizations budget for the transition.

Large enterprises often operate on annual budgets approved well in advance and adjusted only slightly throughout the year, but a post-quantum transition may require more flexibility as standards and threats evolve.

This principle should also influence procurement. SOC 2 compliance already requires organizations to scrutinize vendors' privacy policies, certifications, and access controls, but cryptographic dependencies can be another blind spot. Asking which standards a vendor supports, how cryptography can be updated, and what its PQC roadmap looks like can reduce the risk of buying infrastructure that becomes difficult or expensive to migrate later.

The quantum timeline starts with today’s budget

While a firm deadline for quantum readiness remains elusive, budget cycles offer a more disciplined, actionable framework for planning. Executing a multi-year migration requires sustained financial commitment over successive cycles; every passing budget period effectively compresses the available window for remediation.

Most CISOs already understand quantum risk. The challenge is moving quantum readiness from something planned for next quarter or next year into funded work. With only a limited number of budget cycles available for a multi-year migration, repeated delays quickly add up.

The priority is to turn that awareness into an inventory, clear business priorities, and a phased migration plan that can be funded over the coming budget cycles. By starting now, organizations have time to focus on the highest-risk systems and address legacy infrastructure before the timeline becomes critical.

We've featured the best web hosting services.

This article was produced as part of TechRadar Pro Perspectives, our channel to 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/pro/perspectives-how-to-submit

This is easily one of the best-value SIM deals I've seen right now - and it's a steal for data-guzzlers, regular travellers, and anyone who hates being locked into a monthly contract.

Right now, this EE Rolling Unlimited SIM is just £15 a month (was £35) at Scancom when you use the code HUKDSAVE.

As a rolling SIM, there's no monthly contract to sign, letting you cancel any time. The plan includes unlimited monthly calls, text, and data. There's also 25GB Europe Roaming included at no extra cost.

Today's best unlimited SIM deal

Unlimited Rolling SIM: was £35 now £15
Unlimited data, texts, calls | 25GB European roaming | No monthly contracts

Running on EE's fast network coverage, this is one of the best SIM deals I've seen. The included 25GB data roaming in Europe makes it ideal for business travel and holidaymakers who want to avoid expensive roaming fees. For the full discount, use code HUKDSAVE. View Deal

Why we recommend it

Unlimited data, calls, and text, and free European roaming data make this one of the best-value SIM deals around right now.

At £15 a month, rather than £35, you get EE's reliable network speeds with absolutely no contract to sign, so you can cancel any time.

The generous 25GB data roaming for Europe is ideal for frequent travellers (whether that's for work or pleasure). And it's more than enough data allowance for streaming, video calls, map navigation, and checking messages and emails when public or hotel Wi-Fi is slow.

I haven't seen a similar cost with similar benefits from EE either on Amazon or direct from the network provider.

Price context & historical value

No two ways about it, £15 a month for unlimited data, calls, and text is an excellent price. For comparison, an Unlimited Rolling SIM plan without the 25GB European data roaming is priced at £25 a month at Amazon. Buying a monthly contract unlimited SIM with no free roaming data direct from EE is currently £23 a month.

Should you buy it?

✅ Buy the EE Unlimited Rolling SIM if...

You don't want to sign up to a monthly contract, use a lot of data at home, or you're often traveling to Europe for work.

❌ Skip the EE Unlimited Rolling SIM if...

You'd rather buy direct from EE to get the network provider's support and sign-up extras, or you need more than 25GB European data roaming.

The Catch: What to know before you buy

As it's part of the EE network, make sure you get the service in your area. You can use EE's coverage checker here.

It's also worth pointing out that Scancom is also technically a B2B site, and you'll use the company for support, rather than EE's in-house team.

That doesn't mean you need to be a business to buy and use the SIM, and we've seen reports online that the company won't check for business credentials anyway. But that's its main purpose is for professionals.

• Shop the full SIM range at Scancom

• The first trailer for Neon's Sam Altman biopic Artificial has been revealed
• The Andrew Garfield-led movie will be released in late 2026/early 2027
• It covers Altman's controversial rise to prominence as OpenAI's co-founder

The first trailer for Artificial, a movie about controversial OpenAI co-founder Sam Altman, has been revealed.

Unveiled yesterday (September 8), the teaser for the Neon-distributed film strikes an unusually sinister tone for what's being billed as a biographical comedy-drama.

Indeed, from the gun-laden room that lead star Andrew Garfield, who'll portray Altman, visits, to Garfield's unsettling voiceover that plays over Artificial's publicly-released footage, it's certainly foreboding in nature. I get the feeling, then, that it'll be a dark-comedy dramatization of Altman's rise of prominence, the launch of artificial intelligence (AI) chatbot ChatGPT, and his brief and somewhat disputable ousting as OpenAI CEO in late 2023.

Artificial will initially debut in US theaters on December 25, 2026, before receiving a wider international rollout in early 2027. UK viewers, for example, will have to wait until January 22, 2027 to catch one of many new movies that'll arrive in the months ahead.

Artificial cast and crew: who's involved?

Garfield (left) will be joined by a whole host of famous faces in Artificial (Image credit: Neon)

Oscar-nominated filmmaker Luca Guadagnino (Challengers, Call Me By Your Name) is on directing duties, while Saturday Night Live writing alumnus Simon Rich has penned its screenplay. Meanwhile, Blur and Gorillaz frontman Damon Albarn has also created original music and songs for Artificial.

Where the movie's cast is concerned, Garfield, who impersonates another Silicon Valley techbro following his portrayal of Eduardo Savarin in The Social Network, aka the 2010 dramatization about Facebook's creation, is arguably its most recognizable star — and based on what little footage we've seen thus far, he appears to have Altman's mannerisms, vernacular, and other unique traits down to a tee.

Here's who else you'll see in Artificial:

• Yura Borisov (Anora) as Ilya Sutskever, OpenAI's chief scientific officer and co-founder
• Monica Barbaro (Crime 101) as Mira Murati, OpenAI's chief technology officer
• Cooper Hoffman (The Long Walk) as Greg Brockman, OpenAI's chairman
• Ike Barinholtz (The Studio) as Elon Musk
• Mark Rylance (Bridge of Spies) as Geoffrey Hinton, the so-called 'Godfather of AI'
• Jason Schwartzman (The Grand Budapest Hotel) as TBC
• Chris O'Dowd (The IT Crowd) as TBC

What is the plot of Artificial?

Artificial will seemingly focus on a specific period of OpenAI's history (Image credit: Neon)

Here's the official logline for Artifical: "[The movie] chronicles the incredibly consequential days leading up to the sudden firing and reinstatement of Sam Altman, as the fate of who gets to control the technology at the centre of an AI arms race hangs in the balance."

It's unclear how much of Altman's backstory, his various relationships with his OpenAI colleagues, and ChatGPT's launch and subsequent sociopolitical impact will be covered as part of Artificial's story. One thing is for sure, though: given that it dramatizes events concerning Altman and OpenAI, the discourse that will surround Artificial's release will be impossible to ignore.

The map covers 75% of the sky, and you can explore all sorts of cool celestial objects once you know where to look.

• Google is investing over $15 billion into Finland over two years
• Nuclear energy, wind power and battery storage are top of mind
• The company is also improving local forests, adding new saunas

Google has announced plans to invest around $15.1 billion into Finland's AI and digital infrastructure over the course of the next two years, making it its largest single investment in Europe to date.

The company said the extra capacity is needed to meet growing local demand for services like Gemini, Maps and YouTube, but instead of just growing its data centers, Google is also allocating money for grid upgrades, clean energy projects, battery storage and more.

The country's cool climate and low-carbon electricity are especially attractive for hyperscalers as they try to improve the efficiency of data centers against a backdrop of heightened public and regulatory scrutiny.

Google is investing heavily in its Finnish data centers

The Hamina campus, which it's been running for around 15 years, is especially notable because it uses both seawater-based cooling to minimize its water consumption impact, as well as an offsite heat-recovery scheme to redirect waste heat into local properties.

Local communities in Hamina, Kajaani, Muhos, and Vaala are set to receive a combined $36 million in funding to support upskilling and education, while as many as 37,000 jobs are expected to become available throughout the 2027-2028 construction phase.

On the energy front, Google has signed a 22-year power-purchase agreement to support the extension of the Loviisa nuclear plant and has drawn up plans to add new onshore wind capacity and a new 94MW battery facility to bridge the gap between renewable energy availability.

Google Global Infrastructure VP Bikash Koley also stressed the company's other community and environment investments, including native forest and wetland regeneration, accessible recreational trails, public saunas and fishing piers.

For Finland in particular, Google claimed to have invested around $4 billion in Hamina and related infrastructure alone since acquiring the site, a former paper mill, in 2009.

God, the search for Joe (Zoe Saldaña) is taking its sweet time, isn't it? At this rate, things still won't be resolved by next week's Lioness season 3 finale.

The CIA has narrowed its rescue search to the Houston area, but fallout from dismantled Russian networks in Washington has led to some tense confrontations and a pretty disastrous decision for one of our leading senators.

But surely this week has to bring better things... right? And when does Lioness season 3 episode 7 arrive on Paramount+?

What time can I watch Lioness season 3 episode 7 on Paramount+?

Lioness season 3 episode 7 will drop on one of the world's best streaming services in the US and Canada on Sunday, September 13 at 12am PT / 3am ET.

Here's when it will be released in other nations globally:

• US – 12am PT / 3am ET
• Canada – 12am PT / 3am ET
• UK – 8am BST
• India – 12:30pm IST
• Singapore – 3pm SGT
• Australia – 5pm AEST
• New Zealand – 7pm NZST

When do new episodes of Lioness season 3 come out?

(Image credit: Paramount+)

Lioness season 3 will have a total of eight episodes, with new entries airing weekly. That gives us the following schedule:

• Episode 1: out now
• Episode 2: out now
• Episode 3: out now
• Episode 4: out now
• Episode 5: out now
• Episode 6: out now
• Episode 7: September 13
• Episode 8: September 20

• Chrome will get new updates every two weeks to minimize real-world attacks
• Smaller, incremental changes should also make it easier to track any issues
• Other major browsers are also shifting to two-weekly release cycles

Google has confirmed that Chrome has officially moved from a four-week Stable release cycle to a two-week one, with effect from September 8 2026 with the launch of Chrome 153.

Besides adding new features, the main objective is for Google to give us performance and security fixes more promptly in an era of increased complexity and heightened vulnerabilities.

And it's exactly those security concerns that pushed Google to want to double Chrome's update frequency in the first lace, with the company now having to deal with a higher volume of patches than ever before.

Chrome will get updates every two weeks

Together with a higher volume of attacks in general, Google has also benefitted from AI-powered vulnerability discovery tools, giving developers more work to do than before. "With automated AI discovery tools and community bug reports generating higher patch volume, shorter release cycles make managing security fixes significantly simpler," the company wrote.

Crucially, the faster cycle reduces the time between a fix landing in Chrome's public codebase and it actually reaching end users, making it less likely for a bug to be exploited in the wild.

Chrome has been adhering to a four-week cycle since 2021, but the world is a very different place in the space of just half a decade.

With the higher pace of releases, Google also anticipates a smaller scope of changes, which could actually be a good thing because it could be easier for developers to identify what went wrong in the event of any issues.

Desktop, Android and iOS versions of Chrome will all be impacted by the change, with Chrome 154 already slated for a September 22 release.

More broadly, Google says Chrome isn't the only browser getting more frequent updates. Microsoft, Mozilla and Brave are also making similar changes, and they're all undergoing those changes right now.

Enterprise infrastructure has always adapted as scale increased. Virtualization tackled server sprawl, cloud computing reduced the need to provision physical resources for every application, and automation made increasingly complex environments manageable. Artificial intelligence presents a different kind of scaling problem.

The discussion around enterprise AI has largely centered on models, GPUs, and inference performance, but those technologies represent only a fraction of what organizations must operate. Every production AI deployment creates a continuous flow of data that must be ingested, protected, moved, analyzed, retained, governed, and eventually archived.

Those activities place demands on infrastructure that are very different from the workloads storage systems were originally designed to support.

This is becoming increasingly apparent as organizations move beyond pilot projects. AI is no longer a single workload running on isolated infrastructure. A single application may include high-speed storage for model training, object storage for inference data, lower-cost capacity for operational datasets, immutable storage for cyber resilience, and long-term archives to satisfy regulatory requirements.

Traditionally, those functions have been handled by separate products with separate management tools, security policies, and operational teams. That architecture worked reasonably well when data moved slowly and applications followed predictable lifecycles. AI changes both assumptions.

Training datasets expand continuously. New models are introduced far more frequently than traditional enterprise applications. Inference workloads fluctuate as demand changes. The same dataset may move repeatedly between active processing, backup, compliance, and archival over its lifetime.

Each transition introduces another operational task, another opportunity for inconsistency, and another point where administrators must intervene. Before long, the effort required to manage the infrastructure begins to rival the effort required to build the AI applications themselves.

Complexity becomes the real infrastructure challenge

For years, the answer to operational complexity was automation. Administrators automated provisioning, scripted maintenance, and orchestrated repetitive tasks. Those capabilities remain valuable, but they were designed to execute predefined actions under predefined conditions.

AI environments are considerably less predictable. Infrastructure must continually adapt to changing workloads, shifting performance requirements, evolving security policies, and rapidly growing data volumes, often without the benefit of stable operating patterns.

That is where autonomous data infrastructure represents something more substantial than another automation framework. Rather than treating storage as a collection of independent systems, it starts with the assumption that the platform itself should continuously optimize how data is managed throughout its lifecycle. Capacity, performance, protection, and cost become policy decisions rather than infrastructure projects.

Data moves between performance tiers automatically according to business requirements instead of being exported, migrated, and re-imported into separate platforms. A single namespace spans workloads that historically required multiple storage systems, allowing infrastructure to evolve without repeatedly forcing administrators to redesign the environment.

That architectural change may ultimately prove more important than the automation itself. Many organizations underestimate how much operational complexity accumulates simply from running multiple storage platforms. Every environment has its own authentication model, monitoring tools, lifecycle policies, upgrade schedules, recovery procedures, and performance characteristics.

As AI expands across the enterprise, those management layers multiply alongside the data. Reducing the number of operational boundaries often creates greater long-term value than introducing another layer of orchestration.

The same principle applies to cyber resilience. AI has increased the value of enterprise data far beyond traditional business records. Training datasets, model checkpoints, vector indexes, and inference pipelines have become strategic assets in their own right. Protecting them requires more than backup software. It requires infrastructure that assumes failures and attacks will occur and is designed to recover without depending on manual intervention.

Governance becomes part of the data lifecycle

The conversation also extends beyond security to control. As AI initiatives become more strategic, organizations are under growing pressure to understand where data resides, who can access it, and which legal and regulatory frameworks govern it.

That is especially true for enterprises operating across multiple countries or in highly regulated industries, where data residency requirements, digital sovereignty initiatives, and industry-specific compliance obligations increasingly influence infrastructure decisions.

Rather than treating these as separate governance exercises, modern infrastructure must make location, retention, and access policies part of the data lifecycle itself, enabling organizations to meet regulatory requirements without introducing additional operational complexity.

One of the more significant design decisions behind autonomous data infrastructure is that immutability exists within the storage engine itself rather than being implemented solely through administrative policy. Instead of modifying existing data in place, new versions are written separately while previous versions remain intact.

Combined with distributed self-healing that rebuilds only affected objects instead of entire disks, this creates a fundamentally different operational model for resilience. Recovery becomes part of normal system behavior instead of an exceptional event requiring administrators to coordinate lengthy repair efforts.

Infrastructure operators become infrastructure architects

Perhaps the most interesting implication has little to do with storage technology itself. Infrastructure teams are already responsible for environments that are growing faster than headcount, and AI is accelerating that imbalance. The objective is not to remove people from operations, but to reduce the amount of time highly skilled engineers spend on repetitive maintenance that adds little strategic value.

As more routine activities become policy-driven and continuously optimized, infrastructure professionals can devote more attention to architecture, governance, capacity planning, and aligning technology decisions with business priorities.

That evolution mirrors what is happening across software engineering, networking, and cybersecurity. AI is steadily shifting human expertise away from repetitive execution and toward system design, governance, and strategic decision-making. Autonomous data infrastructure reflects the same progression.

Rather than asking administrators to manage an ever-growing collection of storage products, it treats the infrastructure as an adaptive system that operates within policies established by the people responsible for it. The most effective approach is not to take humans out of the loop, but to keep them in control of the decisions that shape security, compliance, and business outcomes while allowing the platform to execute routine operational tasks autonomously.

Viewed from that perspective, autonomous data infrastructure is less about storage than it is about preparing enterprise IT for the next decade. AI has exposed the limitations of architectures built around isolated products, manual coordination, and steadily increasing operational overhead.

Organizations will continue investing in faster GPUs and more capable models, but those investments will deliver their greatest value only if the infrastructure beneath them becomes equally capable of managing complexity. The next generation of enterprise infrastructure will not simply store data more efficiently. It will actively participate in operating the environments that modern AI depends upon.

We've featured the best business intelligence platform.

This article was produced as part of TechRadar Pro Perspectives, our channel to 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/pro/perspectives-how-to-submit

• Audio-Technica knows what it's doing with open-back headphones
• However, this is the company's first ever wireless open-back offering
• …and also, they feature active 'noise reduction'

All it takes is a quick glance at our Audio-Technica ATH-ADX700 review to know that Audio-Technica knows precisely how to make a great set of open-backed headphones.

That said, this is the first time the company is cutting the wire, when it comes to delivering such audio architecture. Today, (September 9), the company is expanding its open-back headphone line-up with the new ATH-AL5NC, its first ever wireless open-back headphones.

At their core are specially designed 53mm drivers, promising "detailed, expansive sound", while the open housings should improve airflow and reduce heat around the ears, if your closed-back cans often feel a little toasty.

Tipping the scales at 240g (for reference, the Sony WH-1000XM6 weigh 254g), they’re light too. Longer listening sessions at home should be comfortable, whether listening to your hi-res music, working or, as suggested by A-T, gaming.

And here's a new duo in this space: despite their open-back design, the ATH-AL5NC also feature Sound Leakage Reduction, helping to reduce the amount of sound escaping into those shared spaces (the London Underground carriages spring to mind), alongside "noise reduction" (available in wireless and wired modes) to minimize everyday background distractions.

Audio-Technica ATH-AL5NC: key specs and features

(Image credit: Audio-Technica)

Want the cold, hard specs? Of course you do. Other key features include up to 75 hours of battery life, or 50 hours with noise reduction enabled (so it is an active technology, involving the creation of antiphase to counteract noise), plus up to eight hours of playback from a 15-minute quick charge.

You also get 24-bit/96 kHz high-resolution audio, LDAC support, multipoint connectivity for staying connected to two Bluetooth devices at once, a low-latency mode for gaming and video and an integrated microphone for calls and meetings. And those last four perks are not features you'll regularly find among audiophile-grade open-back headphones, or even the best wired headphones generally.

Customisable EQ and Soundscape modes are also available through the A-T Connect app and (of course) you get optional wired listening, with a bundled 1.2m cable featuring an in-line mic.

Something for our best noise-cancelling headphones buying guide? Not so fast; "noise isolation" is perhaps not quite the same as "noise cancellation", despite the 'NC' used in A-T's naming structure. Although the company hasn't expanded on this, adding true ANC in a set of headphones that are open through the ear cups (with perforated grilles to let sounds to move directly through their driver housings) really would be a feat of engineering.

The Audio-Technica ATH-AL5NC are available from today (September 9), priced at a refreshingly reasonable £219, which is around $299 or AU$410.

How good are they for this money? We're working on getting our hands on a set. As soon as we know, so shall you.