TechRadar News

Discussions about passkeys replacing passwords as the go-to sign-in method have rapidly accelerated. The National Cyber Security Centre (NCSC) has acted as the catalyst for this transition, describing the shift as "overhauling decades of security practice". Finally, the industry has reached a collective point of recognition: traditional, static passwords are a weak defense against modern, automated threats.

AI has turned cybercrime into a highly efficient, industrial-scale enterprise. With researchers uncovering agentic ransomware attacks, we are fast approaching a reality that outpaces traditional human-centric defenses. Cybercriminals routinely deploy industrial-scale AI tooling to orchestrate and deploy vast credential-stuffing attacks and hyper-realistic phishing campaigns at machine speed and scale.

For organizations, the adoption of passkeys and other passwordless verification methods is dependent on the maturity of existing identity and access management (IAM) systems. Of the businesses with less mature, or no IAM strategy in place, only 30% are embracing password alternatives. Without passwordless systems, threat actors will continue to undermine digital trust.

User friction and weak protection

For years, identity security was treated as an isolated concern for the operations department. However, as customer experience dictates corporate revenue, delivering seamless IAM has become a commercial imperative.

Passwords force a broken trade-off: making users create unique, complex strings of characters results in severe password fatigue, while cutting corners by reusing credentials and using easily guessable phrases leaves gateways vulnerable to threat actors. When security kills usability, the business model breaks.

Another barrier passwords create is that, although most of the British public is genuinely worried about identity theft and fraud, only 17% fully trust organizations managing their identity data. Passkeys directly resolve this trust deficit through cryptographic authentication.

Because private keys stay securely on the user's local device and are never shared with a central server or application, there is no central database for hackers to breach. Passkeys are also phishing-resistant, removing the risk of users falling prey to social engineering attempts. Unlike passwords, they are also cryptographically bound to a specific, verified domain name, adding another layer of technological defense against attackers.

Securing autonomous identities

The shift towards agentic AI – fully autonomous systems operating with minimal human intervention – shatters traditional authentication models. Passwords were fundamentally built for human brains to remember.

An autonomous AI agent, such as autonomous digital proxies, personal shoppers and synthetic workflows, cannot "remember" a password without creating severe vulnerabilities, such as hardcoded credentials or shared session tokens exposing permanent risks.

If an agent inherits over-privileged human credentials, a compromised or rogue AI can exfiltrate sensitive data at machine speed, vastly outpacing the response time of traditional security methods.

To secure the growing autonomous workforce, AI agents must be given their own distinct, first-class identities with task-specific, time-bound permissions. Passkeys and decentralized, cryptographic credentials secure these machine identities and allow enterprises to securely verify and audit agentic workflows and interactions, without exposing underlying human credentials.

From passwords to passkeys

The technology industry is driving the momentum behind passkey adoption. Tech giants like Google and Amazon are seeing the need to improve customer convenience and account security, positioning passkeys as their default authentication standard. Businesses that delay moving towards passwordless frameworks are left defending a perimeter built for an entirely different era.

Although most organizations now see the importance of adopting passwordless options, transitioning away from traditional passwords is hindered by real-world complexities. Many core IT systems and legacy infrastructure are reliant on passwords, which remain deeply embedded within business processes.

IT leaders therefore fear that updating these systems will demand massive code rewrites and extensive specialist development resources, meaning replacing them requires careful orchestration.

To address this challenge, modern, low-code or no-code identity platforms allow organizations to visually map, test and deploy sophisticated authentication journeys seamlessly. This enables businesses to introduce modern passwordless options like passkeys or biometric prompts progressively, without disrupting the active user workflow or having to replace entire underlying legacy systems.

By allowing teams to modernize security processes without affecting user experience or uptime, advanced identity orchestration becomes a critical business enabler.

The future of trust starts with identity security

The threat landscape continues to be redefined by automated threats, deepfakes and autonomous software. Against these sophisticated attacks, the concept of static, point-in- time logins is obsolete. The NCSC’s call to action has been the final push many organizations need to drop passwords and rethink how to safeguard digital trust.

Adopting an identity-first approach centered on passkeys and robust identity orchestration allows enterprises to reduce user friction, strengthen the security of both human and machine identities and establish the continuous trust needed to grow confidently in an AI-driven environment.

We've featured the best antivirus software.

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

• Control Resonant composer Petri Alanko tells TechRadar Gaming how he approached the game's musical aesthetic
• He says "human vocals" and the "significant" relationship between Dylan and Hedron helped inform his writing
• Alanko says Dylan being an "involuntary hero" provided "a lot of writing material"

Control Resonant composer Petri Alanko has shared how the game's new setting and the relationship between protagonist Dylan and antagonist Hedron informed the soundtrack.

"Usually, it’s the scenery that hits best, and how the main character is related to the scenery, be that a connection or a disconnection," Alanko told TechRadar Gaming in an interview when asked about his design process. "It’s interesting, as there also are no rules whatsoever; again it’s a classic case of 'Hey, I think THAT would make you think this…'

"It’s basically psychological to an extent, really. Trying to play with your brain first, and later on with the player’s brain."

Alanko said "human vocals" was a term he and Remedy identified as part of the Resonant's core music aesthetic, "offering a whole dimension, practically." In addition, the "significant" relationship between Dylan and Hedron, a powerful Resonant entity, was another angle that "drove this first phase onwards."

"The path of the main character [Dylan] and the main antagonist’s [Hedron] relationship with him, it’s indeed a very conflicted pairing," he said. "Dysfunctional, one would say, but that provided a very human layer to the original setting.

"Dylan, being one might call 'an involuntary hero', or 'anti-hero' even, provided a lot of writing material; his path from a blank, soulless and partially mindless being into something willing to sacrifice himself just to find his sister is worth a dozen stories alone, not to mention the fact that their relationship hasn’t been exactly peaceful."

The composer went on to say that terms like "rough," "larger than life," "gut-wrenching," "human, yet inhuman," and "organic" were all on his Post-It notes at the beginning of development, concerning how he approached Dylan.

On the other hand, for Hedron, he had "strangely familiar yet frightening," "clangorous," "structured and sequenced," "mathematical and procedural" as his terms.

"And when I mention 'procedural', I mean the appearance," Alanko added. "It’s almost as if Hedron communicated with circular shapes, hence the term."

Alanko also shared his approach to the whole Remedy-verse score and musical landscape with TechRadar Gaming, saying, "There’s always the absolute evil lurking in there," at the games' core.

As global AI investment continues to climb — it surpassed $300 billion in 2025, according to IDC — enterprise IT remains primarily focused on supporting the development of training models, with architecture such as large GPU clusters and high-density compute taking center stage.

But now AI inference is becoming the dominant workload, and the architecture that supports models in production, at scale, and to real users and systems has quickly become the future of AI.

The demand for AI inference is there — according to Deloitte, inference already accounts for roughly two-thirds of compute, while McKinsey projects that it will represent between 30% and 40% of total data center demand by 2030 — however, the dedicated infrastructure is not.

Inference is more than a scaled-down training environment

What training and inference require from an IT point of view are very nearly polar opposites. Enterprises have facilitated AI training via data centers with massive, centralized data GPU clusters for extreme power density and specialized, high-speed interconnects meant to maximize communication within a single room.

Inference, on the other hand, is defined by a continuous flow of information and unpredictable user demand as opposed to a pre-defined workload. Once the prompts start coming in, you need geographically distributed compute that responds to and scales with real-time demands.

Because inference runs 24/7 and requires millisecond-fast responses for end users, it relies on commodity ethernet, geographically distributed edge sites, and low-latency network connections rather than isolated throughput within a single data center. The very design choices that make an enterprise data center optimal for training actively work against running inference at scale.

The cost structure is entirely different, too. Training is a one-time capital event — once a model is built, that cost is already sunk. Organizations that continue to optimize their spending for training are now being hit with the reality that they have the wrong foundation.

Inference, meanwhile, is a recurring operational cost that scales with every user, query, agent interaction, and more. It’s more like a living, breathing organism which needs to grow alongside the kind of workload that is going to determine the future of AI economics.

Agents are creating new strain on architecture

Autonomous AI agents are also compounding the infrastructure deficit. Agents don’t generate a single response and then stop; they plan, execute across multiple steps, call external tools, coordinate other agents, and pass outputs between systems. The ripple effect of an agentic workload multiplies the data traffic flowing through a network. As demand for agentic solutions grows, so will the demand for supporting infrastructure.

That growth is already consequential. Gartner predicts that by the end of 2026, 40% of enterprise applications will be integrated with task-specific agents, while IDC projects that agent usage will surge tenfold by 2027, with token and API call volumes multiplying by a factor of thousands.

Inference is also acting as a north-south traffic multiplier in ways that traditional WAN and data center network architectures aren’t designed to handle. McKinsey found that most enterprise WANs are currently unable to support AI data traffic patterns, and must evolve as AI now ranks as the second-highest infrastructure investment priority.

Burst traffic is another issue impacting architecture strain. Agentic workloads aren’t a steady flow, they’re a torrential downpour that comes and goes. A single orchestration event can trigger dozens of downstream API calls simultaneously.

This causes networks that are sized for average loads to hit capacity ceilings when agents coordinate at scale, creating the same latency and reliability problems that lead enterprise AI pilots to crash before reaching production. To thrive in the AI era, enterprises must build the proper infrastructure from the ground-up.

Inference is now the critical CPU use case

Building inference-forward architecture starts at the most basic layer: hardware. CPUs are critical for planning, scheduling, tool-call coordination, and the multi-step reasoning loops that define agentic workloads.

However, the industry’s heavy focus on the GPU market during the AI training boom has directed attention away from the necessity for CPUs in AI inference workloads. As inference workloads increase in scale, CPUs move from a supporting role to a critical one, acting as the main filter to the orchestration layer.

When viewed through a purely agentic lens, high-core-count server CPUs can handle the workload more economically than GPU accelerators. But it’s important to understand that one cannot function without the other. Inference still demands GPU function; it’s simply that CPU capacity is the bottleneck that most organizations aren’t accounting for.

Aligning compute, network, and storage is key

The reality is that AI use cases are generating abysmal ROI in most cases. Gartner’s most recent survey of infrastructure and operations leaders found that only 28% of AI use cases fully succeed and meet ROI projections, while 20% fail outright. The leading cause of failure was expecting too much, too fast from infrastructure that was incapable of supporting the workload.

How can enterprises avoid that trap? Take the time to set up their underlying infrastructure before diving headfirst into building systems they can’t support. Putting real effort into building dedicated compute, network, and storage that function as individually managed resources is the key.

Inference at production scale requires all three to work in tandem, with compute handling model execution, storage determining how quickly context, embeddings, and retrieval data are available, and network defining whether compute and storage can communicate externally with the appropriate latency

The organizations that can make this distinction early will be better prepared for the next phase of AI, and will spend less time managing infrastructure failures during production, and more time actually focusing on generating value through AI inference.

We've featured the best AI tool.

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

• Half of UK workers admit to taking personal breaks daily, four in five take 30 to 120 minutes
• Household chores and personal errands are the most common activities
• European workers take fewer breaks and spend less time on them

New research has claimed three in four (76%) UK employees handle personal tasks during working hours at least a few times a week, with half (48%) doing so every single working day, but it might not be such a bad thing.

The vast majority (86%) of respondents in a new LiveCareer study reported that taking short personal breaks during working hours actually helps to improve overall productivity, with a further 11% seeing no meaningful impact. That leaves just 3% who believe there could be harmful effects.

In fact, only 5% of workers choose never to handle personal tasks during working hours, challenging traditional definitions of productivity at work.

Taking personal breaks at work helps to improve productivity

Spread across the working day, nearly three in five (57%) spend between 30 and 60 minutes on personal tasks, with another one in five (19%) spending between one and two hours.

And by far the leading distraction are household chores (54%) and personal errands (51%), which are likely most synonymous with hybrid and remote working.

But despite illustrating just how widespread personal breaks, LiveCareer's data reveals some interesting geographical differences, positioning the UK in a totally different ballpark to other European countries.

While around half (48%) of British workers take personal breaks daily, only around a quarter (27%) of European employees do the same. The figures also point to European workers spending less time on breaks when they do take them.

What the report fails to note is just how productive workers are from each country, to detail whether taking personal breaks has a positive or negative impact on work, however British workers are clearly redefining what productivity looks like in a hybrid working landscape thanks to the added flexibility.

"For employers, the findings suggest that occasional personal breaks are no longer viewed by employees as time lost but as part of a more sustainable approach to managing work and life throughout the day," the report concludes.

• Bose recently launched $99 / £99 'Noise Cancelling Wired Earbuds'
• Also, Bose is 'happy to take credit for making open-ear mass market'
• Bose's President of Premium Audio spoke to me about both developments

Last month was huge for Bose's earbuds output, thanks to the release of not one but two successors to the excellent Bose Ultra Open, a delightfully emo collab on the company's flagship cans and a set of wired, inexpensive earbuds that boast ANC, controlled by the inline remote.

So hang on, on the one hand we've got an evolution on a groundbreaking, clip-style, wire-free product Bose pioneered just over two years ago, (something President Raza Haider says he's "happy to take credit for — making open-ear mass market"). And on the other hand we have… wired earbuds like it's 2003?

Yes. And Bose President of Premium Audio, Raza Haider, is happy to talk about the nostalgia-driven technology trend many of us are seeing. After all, who could forget the various 'digital detox' resolves at the start of the year, compounded by the advent of the great privacy revolt concerning glasses, TVs and even wireless earbuds that want to know a little too much about us?

I start off asking Haider simply, "Who knew wired earbuds would be so popular in 2026?"

"Right? I know", Haider laughs. "It was a funny story actually. Over the last three, four years, the team's done a really good job of being able to capture insights and turn them into products much faster than we historically were able to.

"So the idea [for the wired earbuds] actually came a year and a half ago, when I was dropping my daughter off at university. We saw them everywhere. And my daughter was like, 'You know, these are the coolest things ever.'

"And she's like, 'A year ago, we would have made fun of kids who had wired products, and now we all wear them.' So we came back and as a group back in the office, we're like, 'You know, we should do something about it.'

"We saw the trend, but we saw that all the products in the market didn't have great quality audio — none of them had noise cancellation. So, typical Bose, we wanted to redesign something that was a big part of our history and DNA, and bring back something a little bit more retro, a little bit more fun.

"It's a little bit of an homage. I don't know if you remember the old Bose [wired earbuds], it had the little block in the middle, so we kept the block, except it's more like retro color blocks versus the large, big thing.

"These launch in Europe on the 15th October, but we launched them for presale in America on Monday. It is now Wednesday, and they're gone. They're sold out."

'Something is definitely happening'

(Image credit: Bose)

I'm always keen for presidents of audio companies to talk to me about any trends they're seeing. Haider doesn't shy away from telling me how much he thinks this particular trend might stem from buyers avoiding apps that require data collection, as well as trying to bypass the more intrusive aspects AI.

Is that why people are heading back to wires? "There's two dimensions to it. One, there is something larger happening, whether it sticks or doesn't," he said. "You used the term 'digital detox' — now whether it's trust or whatever it is, there's something happening.

Companies, I think, do realize that they need to do a better job of earning consumers' trust, not letting them feel like there's an invasion of privacy

Raza Haider, Bose President of Premium Audio

"Most tech companies will figure out a way around it. There's no way we're going to back away from technology — the technology is here to stay. But companies, I think, do realize that they need to do a better job of earning consumers' trust, not letting them feel like there's an invasion of privacy. So I think one part is definitely that, there's a return to less technologically driven products. That being said, there's a lot of tech in these that makes these great, right?"

He shows me the earbuds and I have to agree, they're pretty cool. But Haider isn't done. "The second part is this: there is a certain convenience and almost comfort in this," he continued, "and I think this is where technology actually should help.

"So, you can leave these in your bag for weeks on end and never lose a charge, never worry about a connection, never have to worry about a sensitive area. You just pull them out and you plug them into your phone physically. There's no app, no settings, and you just go — that's it!

"There's something oddly satisfying about taking this lovely silicone wire, playing around with your hands, hearing the physical click into the interface; you've got actual buttons for volume, and so it's just oddly fun."

'I think the average consumer doesn't know what a Bluetooth codec is, and they couldn't care less'

(Image credit: Bose)

I ask Haider how much he thinks the knowledge that Bluetooth codecs (however improved) still can't quite compete with wired audio for resolution plays a factor in the wired trend? Or is it simply that physical media is having a moment and if you've got a CD Walkman, you want wired buds?

"It's both", he answers. "I think the average consumer doesn't even know what a Bluetooth codec is, and they couldn't care less. All they know is that when they have a wireless headphone, it sometimes drops out, and this doesn't.

"That goes back to my earlier point; it's reinforcing to technology companies that technology needs to work in a way that's invisible to consumers. The days of people using technology just for technology's sake? We've moved past that for certain categories.

"It also goes back to my point on re-plumbing our company's operating model and product development to be able to move faster to capitalize on trends, because there always will be trends.

"The trend of wired now? There'll be a trend for something else. There are trends for colors too. We've done a great job over the last couple of years of launching I think 60 colors to take advantage of that. Companies have to be a lot more nimbler in listening and adapting to consumer trends."

Let's be completely open

(Image credit: Bose)

As well as literally writing the book on active noise cancellation, Bose pioneered what I like to call the 'ear cuff style' open earbuds, with the Bose Ultra Open. To be clear, this was in February 2024, a time when nobody else was doing that.

Did Haider anticipate how popular that particular design would become, and why does he think production of such designs has skyrocketed since? "Yeah! I think as a category overall, it's like the fastest growing category in consumer wearables, right?" he said. "We saw there was a huge need for people to have an alternative to in-ear products.

"For many reasons, starting from the obvious simple logic, in-ear products don't fit in everyone's ear. Some people don't like jamming stuff in their ear. Then, there are many instances where you want to actually be more aware of your surroundings, where it's actually safer to be aware of your surroundings. You're walking home late night, you're going for a run, you're out walking the dog, you want to hear the cars, the roads, the fire trucks.

"There are also many places where there's a social cue — where having one earbud in is disrespectful, people don't engage with you, you lose the human connection. But having a cue or clip which looks and feels like jewelry? It feels more inviting, you don't feel disconnected.

"And so as a result of that, we took a bunch of technology we had, and we spent a lot of time actually designing that cuff, much more than the 10 months it took for the wired earbuds.

The best thing that can happen is, everyone copies us. And that's what happened

Raza Haider, Bose President of Premium Audio

"We were very deliberate about form, fit, feel, design, design intent and look. And when we launched it, people actually asked us whether this would take off, whether people would understand what it is, whether it'd be alienating, would men want to wear something that looks like jewelry? Our answer was, 'We don't know. We've heard our consumers, they said they wanted it, we're giving them an option for something new.'

"And the best thing that can happen is, everyone copies us. And that's what happened. In the last two years, this thing has blown up."

La vie en Bose?

(Image credit: Future)

Our time together is coming to an end, but I'm keen to ask where Haider, as head of a market leader and pioneer in the portable audio space, thinks consumer audio will be in 10 years?

"Given how fast things are changing now, it seems almost impossible to make a prediction," he said. "But let me give you a few just general trends, right? There's always a nostalgia-driven customer, and companies have to be positioned to be able to successfully capitalize on trends.

"Companies well-positioned to capitalize on trends will be at an advantage. But from a tech standpoint — especially with what's happening right now with AI and the pace at which things are getting smarter — I think you'll see edge or wearable devices becoming a lot more intelligent and contextually aware of what a user wants to do.

"As a result, I think you're going to see a lot more fragmentation in the combination of things a user puts on their body, to optimize something that they're passionate about.

"People will soon realize that most of us have a passion point around something. For some people it might be music, for some it might be health and fitness, for some might be wellness, some might be an active lifestyle, some might be content creation, some might be creating and producing.

"And for all of those people, there's a different set of both awareness and sensory capture and sensory orchestration, let's say, that will allow you to do what you're passionate about better. And companies that realize that, and deliver something that's more consumer and passion point-centric, win.

"So, what my daughter will want to do with the combination of her earbuds, her phone, and her watch would be very different than what you and I might want to do, which will be different from someone who wants to go on a run, which will be different from someone who might be interviewing someone at a concert."

It's an insightful and hopeful look at a more tailored wearable future. Will we still be embracing wires, or will wireless audio finally be hi-res enough (and carefree enough) to be the better standard? Time will tell. But as a daughter myself, I for one really hope the people in charge will always listen to theirs.

The conversation around AI and the jobs market is dominated by anxiety. People worry that roles will be lost to agents – that the job market will dry up from the impact of AI. As a result, there will be a general constriction in the volume of career opportunities out in the world. But is this concern well-founded, what does the data indicate – is it a bleak outlook?

Drilling down into the latest trends and labor market statistics, we can see that the facts paint a different picture. What’s going on? It’s quite a positive shift with the data showing that the job market is in fact rebounding and settling around AI.

Job postings mentioning AI have surged. Since last year, specialist AI job postings have increased by 61% in the UK. This is according to PwC’s AI Jobs Barometer, recording 68,000 new roles and bringing AI hiring volumes back to around 2022 levels. Why are we seeing this uptick? The rise in hiring reflects a broader pressure on companies to deploy and manage AI effectively.

The new AI jobs drive

From a safety standpoint, we’re seeing concerns mounting around open-weight models. We’re also seeing mounting pressure from AI legislators and industry watchers, calling for watertight regulation and governance processes to be put in place.

That pressure has direct implications for workforce strategy – are you hiring the right talent for AI management, do you have the skills in place within your organization to best handle AI operations? Better AI starts with people, and this is where we’re seeing the job market respond over time.

AI roles are vital for the success of the UK AI economy. According to Coursera research, 88% of leaders agree that planned AI investments will not succeed without increasing investments in skills development. The future of work will be defined by those who navigate these changes effectively – knowing how to work seamlessly alongside AI is critical.

The conversation around AI has been dominated by fears of automation replacing workers. But our UK research tells a very different story. While 85% of organizations are already using AI agents in workflows, businesses are increasingly investing in the people needed to govern, secure and scale them.

Among UK IT leaders: 65% expect overall headcount to increase over the next three years, compared with just 14% expecting it to fall. On top of this, only 8% say AI agents are primarily eliminating jobs today. Meanwhile, 48% are hiring AI agent operators, 31% are hiring security, risk and compliance specialists, and 26% are hiring AI ethics and governance specialists.

AI governance becomes a growth function

This is a positive hiring strategy shift, generating scope for AI upskilling, and talent pool growth within the UK economy. What we’re seeing here is that AI governance has become a growth function. Historically, in a negative sense, governance has been framed as a constraint designed to slow down innovation for safety. In the agentic era, that attitude is being reversed.

Strong governance has become the ultimate accelerator. With the UK’s rigorous focus on expanding AI investment and infrastructure, governance needs to be front-of-mind for those innovation-hungry organizations. Additionally, according to our research, nearly 94% of UK organizations say stronger governance would help them move faster with AI, while 93% believe it's essential to scaling AI responsibly.

This is about a concerted focus on ‘re-plumbing’ fundamental structures. UK business leaders understand that teams cannot move fast without confidence in their safety structures. Without clear guardrails, organizations trying to scale AI hit a wall. Employees can leak corporate data, or autonomous systems can surface restricted content. Governance, permissions, and access controls make AI trustworthy, transforming compliance into a growth engine.

How AI is creating new careers, not just changing existing ones

LLM Engineer, Prompt Engineer, AI Governance & Compliance Manager, Chief AI Officer – these are unique new roles we are seeing. AI job risk has been widely discussed in the media, but in fact a bump in opportunities is happening – new opportunities are being created for new, focused AI roles.

As emerging technology becomes embedded into day-to-day operations, demand is growing for specialists who can oversee AI systems, manage risk, ensure compliance and build trust.

For candidates looking to enter this field, knowing how to work alongside digital agents will be the baseline competency of the 21st century – those entering the workforce will increasingly need to build AI literacy. This is where the job market is heading, and skills strategy must align with that too. And it’s not just for deeper technical knowledge – it’s the skills to interrogate and control AI tools from a critical thinking perspective.

The ‘human in the loop’ discussion is not going away – we can see from machine-vs-machine cyber threats that human control is more important than ever. The question is, how do I become the person for that job? For those in the job market seeking these kinds of opportunities, the goal must be a hybrid skill set blending technical literacy with human judgment

The data disconnect with AI engines

Governance is only part of the opportunity. Data and knowledge management are becoming equally important. What we’re seeing now, in that wider ‘re-plumbing’ discussion, is that data libraries are also failing.

While 96% of organizations say access to company knowledge is critical for AI success, only 32% have connected AI agents to trusted internal content across multiple use cases. This disconnect is hurting the growth potential of AI, and limiting its scope to single-use, isolated areas.

Without this connection, agents are left blind and prone to errors. This content disconnect is the primary reason why enterprise pilots stall. Bridging this operational gap requires organizing and securing corporate content pools, and thus hiring the right talent for the job – in the process, creating a massive, multi-year runway of work for data stewards and AI operators.

These kinds of needs or requirements can’t be ignored – especially by larger enterprises that are betting big on AI ROI. They need to iron this out now, and will look to hire the right people for that process.

On the horizon

It’s true there has been significant disruption to the jobs market from AI. Any major technology that has such a wide-reaching impact will create huge ripple effects. But as we study that impact, we can see new opportunities emerging for eagle-eyed candidates and employees. The bottom line is that good governance will be the core foundation for AI.

By framing governance as the foundation of a modern, managed digital workforce, we can unlock unprecedented productivity. And in this transition, we create new sets of roles, reordering the skills priority list. It’s those individuals who are receptive to those changes, and stand out with the right, competent skill sets, who will be the ones to prosper and benefit. This is an exciting time for the technology sector – one that is creating high-value careers for the next generation.

We've featured the best AI website builder.

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

• UK AI minister warns cyber defenses will 'likely' be breached in the future
• More regulation is needed to prevent 'rogue' AI agent incidents
• The UK's AI testing body is working to help secure frontier models before release

UK AI minister Kanishka Narayan has warned that it is “likely, if not inevitable” that the nation's cyber defenses will be breached by AI agents.

Narayan’s warning comes after numerous frontier models have escaped testing environments and hacked third party companies, including one case where an OpenAI agent hacked an Australian government website in June 2026.

Speaking at a press conference, Narayan also said that AI companies should, “Listen when your scientists warn of serious dangers. Investigate unexpected behaviour. And if you cannot adequately control the risks, do not release the system.”

UK government to use AI to defend

The AI minister also revealed that the UK government is planning to use advanced AI to counter the threats posed by AI, which will build upon two pilot programs that have helped spot and patch over 1,000 vulnerabilities across government systems.

In addition, frontier AI systems are undergoing “rigorous evaluations” by the UK’s AI Security Institute (AISI), in order to “identify risks early” and “test robust safeguards to help keep people safe.”

However, there have been some roadblocks in AISI’s ability to test some of the latest models. In September 2026, US President Donald Trump asked OpenAI and Anthropic to not share their latest models with UK testers until US labs had conducted their own testing.

Narayan said that the UK government has since worked with the White House, “to retain access for Britain's AISI to the most powerful AI models before they are released" (via BBC).

AI protections are also likely to be introduced as part of the Cyber Security and Resilience bill, Narayan said, with a particular focus on regulations surrounding the reporting and disclosure of AI incidents.

“It's promising to see the Government taking frontier AI cyber defences seriously,” said Richard Wainwright, Field CTO, EMEA at Okta.

“As AI agents become more autonomous, safety must be built into the infrastructure UK businesses and government rely on. Everyone wants a 'big red button' to turn off rogue AI agents. The industry is increasingly talking about AI 'kill switches' to help protect critical infrastructure, and I hope they are considered as part of the Cyber Security and Resilience Bill regime.”

Why does AI need more regulation?

In 2026 so far, there have been multiple incidents of frontier AI agents escaping testing environments and breaking into the networks of other companies.

OpenAI has notified more than 100 organizations about unauthorized activity tied to the company’s AI agents, including the Hugging Face attack. Anthropic has disclosed two unauthorized AI agents events, and both Meta and Google have disclosed a single testing environment escape each.In multiple cases, the AI agents were being evaluated by Israeli AI testing firm Irregular.

The companies involved are facing significant scrutiny over the safety of their AI models, but are yet to face any punishment for failing to contain their agents.

There is also significant debate around labelling the AI agents as ‘rogue’, with some experts such as Tenable Field CTO Bernard Montel saying it is “a dangerous way of deflecting blame.”

But while regulation may force AI companies to be more rigorous in the testing and deployment of their models, it may also slow down AI development.

For the Trump administration, that is a line that cannot be crossed. Slowing down AI development would mean handing the AI development lead to China and potentially compromising US national security.

The US midterm elections are already fraught with debate over AI development, especially when it comes to the topic of the thousands of AI data centers that are being proposed across the US.

Watch HOUDINI: Back from the dead online

Even Harry Houdini can't escape the wave of AI. In this FOX Nation original, Houdini's likeness has been simulated as an AI avatar. As we hear his story straight from the horse's mouth, here's how to watch FOR FREE from anywhere with a VPN.

Release date: Wednesday, October 14

Stream: FOX Nation

Use NordVPN to watch any stream

HOUDINI: Back From The Dead tells the story of history's most famous magician, told by the man himself.

Released just in time to commemorate 100 years since his death, the show sees Harry Houdini brought to life through the power of an AI that used over 400 pictures from his lifetime.

The series is available online via the FOX Nation app, or the FOX Nation is available as an add-on to Youtube TV or Amazon Prime.

Watch HOUDINI: Back From The Dead FOR FREE

HOUDINI: Back From The Dead is available to watch on FOX Nation in the United States. Fox Nation is currently offering a Free Trial for new members, with full access to the full catalog and free one-day delivery during that time.View Deal

How to watch HOUDINI: Back From The Dead online from outside your country

If you’re traveling abroad and want to watch HOUDINI: Back From The Dead but FOX Nation isn't available where you are, there’s an easy solution.

Downloading a VPN will allow you to stream online, no matter where you are. It's a simple bit of software that changes your IP address, meaning that you can access on-demand content or live TV just as if you were at home.

Use one of the best VPNs to watch HOUDINI: Back From The Dead from anywhere.

Editors Choice

NordVPN – get the world's best VPN
We regularly review all the biggest and best VPN providers and NordVPN is our #1 choice. It unblocked every streaming service in testing and it's very straightforward to use. Speed, security and 24/7 support available if you need – it's got it all.

The best value plan is the two-year deal which sets the price at $3.39 per month. There's also an all-important 30-day no-quibble refund if you decide it's not for you.

- Try NordVPN 100% risk-free for 30 daysVIEW DEAL ON

Can I watch the new Houdini documentary in the UK?

Unfortunately, HOUDINI: Back From The Dead is only available on FOX Nation, which is currently geo-blocked in the UK.

Visiting from across the pond? You can use a VPN to access all of the streaming services you usually do when you're stateside.

HOUDINI: Back From The Dead episode guide

HOUDINI: Back From The Dead is a three part documentary, with all three episodes available immediately

HOUDINI: Back From The Dead trailer

VPN services are evaluated and tested by us in view of legal recreational use. For example:a) Access to services from other countries, (subject to the terms and conditions of that service).b) Safeguarding your online security and making your online privacy more robust when abroad.Future plc does not support nor condone the illegal or malicious use of VPN services. We do not endorse nor approve of consuming pirated content that is paid-for.

As AI becomes embedded into critical public services, sovereignty is rapidly becoming one of the biggest strategic questions facing government.

Much of the debate still treats sovereignty as a question of geography: where is the data stored, which country hosts the infrastructure, and whether information remains within national borders?

These questions matter, but they are only part of the picture. True sovereignty is about maintaining control across the entire AI lifecycle from where data resides and how it is accessed, to how models are governed and tested.

For public sector organizations, that distinction is becoming increasingly important. AI is moving beyond experimentation and into services where the consequences of losing control can be far greater than a poor chatbot response. Healthcare, education and central and local government all rely on data and systems that need to remain secure, resilient and accountable.

Sovereignty needs to go beyond the data center

Keeping information in the UK does not automatically make an AI deployment sovereign. An organization could host its data domestically while relying on external models, platforms or services over which it has limited visibility or control. It may not fully understand where information is being processed, how models are accessing it, or what happens when those systems change.

That creates a much broader question of dependency. As AI tools become increasingly embedded in public services, teams need to understand not just where their data sits, but what sits between that data and the decisions being made from it. Sovereignty means having sufficient control over the infrastructure, connectivity, security, data and governance that underpin the AI environment.

This is particularly important as AI becomes more autonomous. The shift from simple assistants towards agents capable of taking actions on an organization's behalf means that the stakes are rising. The more agency an AI system has, the more important it becomes to know what information it can access, what decisions it can influence and where accountability ultimately sits.

Bring AI to the data, not the other way around

One of the most important principles for public sector AI should be 'bring AI to the data' rather than automatically sending sensitive data to wherever the AI happens to live.

That requires a different approach to AI adoption. Instead of starting with a model and asking what data can be connected to it, organizations should start with the service or decision they are trying to improve and work backwards.

Organizations don't always struggle with insight. They struggle to make decisions. Therefore, rather than putting every problem into the same generative AI model, the technology should be chosen based on the problem.

For government, this could mean keeping sensitive information within controlled environments while connecting appropriate AI capabilities to it through secure, governed infrastructure. It also means accepting that there is no single AI architecture that will work for every use case.

Some applications may benefit from generative AI. Others may be better suited to traditional automation, machine learning or optimization algorithms. The objective should not be to use the newest technology, but to create the right outcome safely and effectively.

Governance cannot be bolted on afterwards

The other misconception is that sovereignty is primarily a cybersecurity issue. Security is obviously fundamental, but governance is equally important. Public sector organizations need to be able to explain what their AI is doing, test whether it is performing as intended and maintain appropriate human oversight.

This becomes especially important because AI systems are based on probability. They do not always produce exactly the same result from the same input, and they can generate answers that appear convincing while being incorrect. That creates a very different challenge from traditional software, where a defined process is expected to produce a consistent output.

AI needs to be treated as something that must be continuously tested and governed, rather than deployed and forgotten. That means bringing security, compliance and risk teams into AI projects from the outset, rather than asking them to approve systems once they have already been built.

It means establishing clear rules around data access and usage, ensuring there is a transparent audit trail for how AI systems operate. Most importantly, designing human oversight into the system from day one.

Infrastructure is becoming part of the AI strategy

None of this is possible without the right foundations - a sovereign AI strategy requires secure cloud environments, resilient connectivity, strong cybersecurity, appropriate data architecture and infrastructure that can support changing AI workloads.

That is particularly important because AI is evolving so quickly. The public sector is already having to balance what it builds in-house with rapidly changing platforms and services from technology providers. Long-term flexibility is becoming just as important as today's technical capability.

For the public sector, this should be viewed as a strategic opportunity rather than simply another IT challenge. Building flexible, secure infrastructure now creates the foundation for adopting new AI capabilities later, without having to surrender control every time the technology changes.

The goal should not be to isolate public sector organizations from the wider AI ecosystem. That would neither be realistic nor desirable. It should be to make sure they can participate in that ecosystem on their own terms.

Sovereignty is about control

The UK's AI ambitions will depend on its ability to adopt new technologies at scale while protecting the information, services and citizens that depend on them.

We need to move the conversation beyond where data is stored. Sovereign AI is ultimately about control over data, infrastructure, access, governance and the decisions AI systems are helping to make.

Public sector organizations that build those principles into their AI strategies now will be better positioned to take advantage of increasingly capable systems without becoming dependent on technologies they cannot fully govern.

The future of AI in the public sector should not be about choosing between innovation and sovereignty. With the right foundations, it should be possible to achieve both

We've featured the best AI chatbot for business.

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

• CMF by Nothing has the compact Speaker S and larger Speaker M
• Both tote Bluetooth 6.0, dual device connection, and multiple audio modes
• You can detach the speakers' panels and 3D print your own replacements

There are two new audio devices in town, and they come from Nothing’s sub-brand CMF — but they’ve attracted substantial attention in such little time for one particular reason.

Hot on the heels of the recent Nothing Headphone (1) Pro, the new Speaker S and Speaker M mark CMF’s portable speaker debut, mainly because of their quirky milk carton-inspired design which, as you can see, gives it quite an edge over the best Bluetooth speakers on the visual front. The smaller Speaker S is going for $59 / £49 and the larger Speaker M is $109 / $99, with both being available in white, dark grey, orange, and turquoise.

(Image credit: Nothing / CMF)

While both are affordable entry-level options, they’re built for different purposes. The Speaker S is more compact, designed for easy transportability and listening on the go, packing a ‘custom-tuned’ 45mm driver and two passive radiators. If filling a room with your favorite tunes for a party or gathering is the bigger priority, that’s where the Speaker M comes in handy, which packs a 60Hz–20kHz frequency range.

As far as their respective features go, that’s pretty much it, and despite the size difference there’s a lot of common ground between the two models sitting beneath its carton-like surface. For starters, both speakers give you 18 hours of playback at 50% volume with Ultra Bass on, an audio feature that adds a little impact to music that needs ‘extra weight’. Fast-charging for 10 minutes will also give you an hour’s worth of playback.

Both support Bluetooth 6.0, allowing a connection range of over 100 metres between the speaker and your connected device. If you find yourself flicking between two different devices to control your audio playback, both models support dual-connection, meaning you can move between devices without having to keep pairing them to your speaker each time. I can see this coming in handy if there are two passionate maestros at the same gathering who will stop at nothing to assume total control over the party’s soundtrack — sharing is caring.

(Image credit: Nothing / CMF)

Speaking of sharing, CMF’s new speakers offer an extensive range of playback options. Firstly, Stereo Mode turns two Speaker S or Speaker M devices into individual left and right output channels for a stereo pair. Party Mode is even more flexible however, giving you the freedom to pair both models together to beef up the sound, including, apparently, compatible models from other brands.

CMF is also taking its first speakers to another level by adding a unique customization element to the aesthetic. The front panels of both the Speaker S and Speaker M are removable, and can be swapped with another one of a different shade. CMF is also encouraging 3D printer owners to go ham on the creativity front, providing a guide for you to print your own custom panels and volume knobs so you can build your speaker around your home’s interior. It’s so extra, but I love the creative freedom.

Slaying vampires always feels good, and 2D action-adventure and exploration game Castlevania: Belmont’s Curse is no exception. The first entry in the iconic horror platformer in more than a decade, Evil Empire and Konami have crafted a love letter to past titles that, while not changing much of the formula, delivers an experience worthy of the Belmont name.

Review info

Platform reviewed: PS5
Available on: PS5, Xbox Series X|S, Nintendo Switch 2, PC
Release date: October 15, 2026

Despite a fantastic animated adaptation (and great spinoff series with Nocturne), Castlevania has been six feet under as Konami barely revisited the series. Seldom for tie-ins with Dead Cells and re-releases, so fans were ready to sink their fangs into a new title.

Fortunately, it would come from the Dead Cells team, Evil Empire. Its mastery of the craft with the Metroidvania genre made it a safe but effective pick to hand the Vampire Killer whip to. And, lo and behold, they ran with it and made something fans will love.

You must whip it, whip it good

(Image credit: Konami)

Surprisingly, the game is a sequel of sorts to both Netflix’s Castlevania and Castlevania 3: Dracula’s Curse. As Rose Belmont, the daughter of Trevor Belmont and Sypha Belnades, she is tasked with stopping the newfound dangers flooding Paris.

With her father’s guidance and all the Belmont tactics, she’ll discover the root of it all, as well as how she and her mysteriously cursed arm fit into things.

Castlevania: Belmont’s Curse does go a good distance with its narrative, but it doesn’t reach too high. It’s got some twists, turns, and great moments, though it can feel sparse. Of course, most of the fun will come from your experience dashing and slashing, but it can feel uneven with its cinematic counterparts. Still, it’s serviceable enough for you to want to keep playing and see more.

Rose Belmont, as the daughter of Trevor and Sypha, is a pretty likeable protagonist. She’s not one-note by any means and struggles with her emotions, especially with the spotlight now on her. However, it’s the supporting cast that provide her with what she needs to be a worthy Belmont.

Without spoilers, her Confidantes will be her moral compass, and it’s up to her to choose which direction she wants to go in. Props go to many of the fantastic performers in the game, as everyone works perfectly with their respective characters.

A missable little pile of secrets

(Image credit: Konami)

Once you set foot off the docks of the Seine, you’re instantly thrust into classic Castlevania gameplay: slay monsters and maneuver through obstacles. There are small tutorials, but for the most part, Paris is an open book, and you’re free to flip through the pages almost immediately.

Action-wise, it also harkens to classic Castlevania, with Rose using a variety of weapons to kill the many creatures of the night. There are many types of weapons, from the iconic whip to flashy gauntlets. Finding new versions of them throughout the world is always a treat, as it never disappoints when you equip them and test them out. Personally, the whip was a go-to, as it rewarded distance with better damage output, though the spear was a close second.

Also, with Rose’s abilities, which she’d earn after certain bosses, it unlocks various attacks, like throwable crosses or pillars of fire. It leaves customization very open, letting you play how you’d like in terms of combat.

Best bit

(Image credit: Konami)

The level design in Castlevania: Belmont’s Curse is simply phenomenal. From trap-filled chasms to eerily beautiful vistas, the team has done a wonderful job making levels look seemingly impossible but enticing enough to make you want to persevere and keep going.

Without going into spoilers, Rose’s journey goes beyond the confines of Paris into various regions. Each of them has some theme and gimmick that makes them memorable in both a playing sense and a backtracking sense. As this is a Castlevania game, you will find yourself going back to certain locations to find secrets and clues for other locations. Luckily, with new abilities like an air dash and, of course, a grappling hook whip, you’ll be able to do so easily and rapidly.

That said, the level design in Castlevania: Belmont’s Curse is phenomenal. Traps and platforming take advantage of Rose’s abilities, including ones she earns throughout her adventure. It takes precision and good reflexes to maneuver through them, and it never gets old. It can get frustrating, though, especially if you’re going back and forth a ton. Still, finding what you need after a seemingly impossible challenge feels so good.

Slaves to our family’s wishes

(Image credit: Konami)

As the first new 2D Castlevania in decades, there have been noticeable increases in graphical fidelity since then.

Belmont’s Curse blends 2D with 3D to bring a Dead Cells-like style to the gameplay and look. It does keep the drawn silhouettes of characters for dialogue, but in-game cutscenes are all flat 3D. While the pixelated depth in past titles like Symphony of the Night and Order of Ecclesia is sorely missed, Evil Empire’s redesign works well with the series and fits right in.

Frame-rate-wise, there were rarely any hiccups during the playthrough. There was a stutter or two during certain moments with tons of monsters, but it was pretty seldom. However, there were a handful of glitches with the audio, as it would cut in and out and, often, restart with each level transition. While rare, some glitches with the gameplay, like phased-out enemies or awkward platforming, would happen.

It’s also important to mention that it can be quite a difficult game. The series is known for being brutal, and Belmont’s Curse is no exception. A death can bring you way back and reset tons of progress. It’s a worthwhile challenge, but it is one that takes time and patience. Bosses will be the bane of many people’s existence, as some of them, especially late-game ones, are super difficult. It can feel a bit overkill, but it’s nonetheless satisfying to whittle them down and then finally get the last laugh.

In short, Castlevania: Belmont’s Curse is classic Castlevania brought to the modern age. Evil Empire and Konami have crafted a title that, while it’s not the huge blockbuster many would hope for, is a well-done action platformer. It doesn’t go too far outside the box, but it makes the most of what it does have.

With fantastic level design and combat, the franchise is in a position to return to its glory days and its bloody reign over the action platformer genre.

Should you play Castlevania: Belmont’s Curse?Play it if...

You love a great balance of platforming and combat
You get the best of both worlds with Castlevania: Belmont’s Curse, with the platforming and combat. Each aspect works extremely well and, most of the time, it’s a delicate dance between the two, even intersecting each other at times. It’s frenetic but exhilarating at the same time.

You enjoy a challenge in all areas
You will die a ton of times in Belmont’s Curse, and it’ll be in many different ways. However, if that is a big motivator for you, then you’ll have a wonderful time here. Precise platforming, expert hand-eye coordination, and a tinge of hesitation are required to make it, so if that’s up your alley, it’s the perfect challenge.

You love having the world be a puzzle in itself
Exploration in Castlevania: Belmont’s Curse isn’t just required, but rewarded. There are dozens of secrets hidden throughout the many stages, and backtracking only leads to better opportunities for you. Plus, hints for many connecting points, especially bosses, can be found.

Don't play it if...

You’re looking for something easygoing
As mentioned before, Belmont’s Curse is a difficult title. It doesn’t hold your hand much, and you will die a ton. While it may be possible to play this without throwing your controller in frustration, it’s not the most positive or happy-go-lucky title, given its dark world and countless pools of blood.

Accessibility features

While the game touts its difficulty, there are available accessibility features to lessen it. There are also various visual adaptations for thicker outlines or colorblind mode. In addition, subtitles are available in different languages, as well as voice-overs. However, despite some accessibility, some things, like one-hit-kill traps, exist and can be frustrating.

How I reviewed Castlevania: Belmont's Curse

I played Castlevania: Belmont’s Curse for around 16 hours, seeing much of the game’s map, bosses, and secrets. I also got many of the available weapons and spells, as well as heart and mana containers. In addition, I did play the demo, which was released prior to the game’s release, and transferred the save file from that to the full title.

I played the game on a base PS5 with the DualSense Wireless Controller and HyperX Cloud 3 wired gaming headset, all on a Roku 4K TV.

First reviewed October 2026

Google has given its Nano Banana AI image generator another upgrade, and the latest version, Nano Banana 2.1, is arriving inside Google Gemini in an increasingly crowded field of tools that can turn a few sentences into remarkably convincing pictures. The promise is familiar, with sharper details, better instruction-following, more accurate text, and greater consistency across images. The real challenge is finding something that makes it meaningfully better than the competition.

Nano Banana's main rival, ChatGPT's image generator which is called Images 2.5, is hardly sitting around waiting to be overtaken. OpenAI has made considerable progress with realistic lighting, complicated compositions and text rendering, turning image generation into one of ChatGPT's most entertaining and genuinely useful features. With both companies pushing the technology forward, I wanted to see whether Google's latest upgrade had managed to pull ahead.

If you've got a Gemini account you'll find that it should have been upgraded to use Nano Banana 2.1 by now. To test the two image generators against each other I gave Gemini and ChatGPT identical prompts designed to test different abilities. I compared the resulting images side by side, with ChatGPT on the left and Gemini on the right. The results were surprisingly close, although the two generators often approached the same instructions with different artistic instincts.

Storms within storms

A tiny Victorian greenhouse inside a glass snow globe during a thunderstorm. ChatGPT on the left, Gemini on the right. (Image credit: ChatGPT/Gemini)

My first test was designed to challenge the models' understanding of light, glass and reflections, three things that can quickly expose the weaknesses of an AI-generated image. I wanted a scene with multiple transparent surfaces, complicated lighting and enough tiny details to make the results worth examining closely. The prompt was, "Create a photorealistic image of a tiny Victorian greenhouse inside a glass snow globe during a thunderstorm, with realistic reflections, refractions, and lighting."

Both generators produced beautiful images that would look perfectly at home on the cover of a fantasy novel. ChatGPT's version featured an elaborate, golden-lit Victorian greenhouse inside a large glass sphere, sitting on a polished wooden base beside a rain-covered window. Lightning illuminated the storm outside, while the globe contained a surprisingly detailed miniature world, complete with plants, pathways and delicate architectural flourishes.

Nano Banana 2.1 went for something slightly more restrained. Its greenhouse looked more like an actual miniature structure, with dark metal framing, lush greenery and warm interior lighting that contrasted beautifully with the blue-gray storm outside. The globe's reflections were particularly convincing, and the lighting felt a little more natural, although the wet tabletop was an odd choice given that the scene appeared to be indoors.

Titan calling

ChatGPT on the left, Gemini on the right. (Image credit: ChatGPT/Gemini)

Text has historically been one of the most embarrassing weaknesses of AI image generators. Letters would melt into one another, words would become nonsense and asking for an exact phrase could produce something that looked like an alien language. For the second test, I wanted to see whether either model could handle a fairly demanding design brief without mangling the typography.

The prompt was, "Design a retro-futuristic travel poster for Saturn's moon Titan with the exact text 'TITAN IS CALLING,' 'Seven nights beneath the rings,' and 'Starting at $2,499.'"

ChatGPT created a luxurious travel advertisement that looked like something a very wealthy space tourist might encounter in an interplanetary airport. Saturn dominated the sky above a beautiful orange landscape, while futuristic resorts and a couple enjoying drinks suggested that Titan had become the solar system's most exclusive vacation destination. One thing I didn't like was the gap it left between the $ and the 2 in its text rendering.

Google took a more cartoonish approach with a colorful vintage science-fiction poster featuring rockets, astronauts, icy mountains, and a futuristic base. It was actually more realistic about how the moon is not at all hospitable but still had the optimistic quality of a 1950s space-travel advertisement, complete with bold lettering and a slightly weathered finish. Crucially, it also reproduced the requested text correctly, including the price.

Fox library

ChatGPT on the left, Gemini on the right. (Image credit: ChatGPT/Gemini)

Character consistency is another tricky problem for AI image generators, particularly when the same character needs to appear in several scenes. A fox might suddenly change its appearance, lose an accessory, or acquire an extra limb between panels. I decided to test this with a simple illustrated story that required both visual continuity and a coherent sequence of events.

The prompt was, "Create a four-panel comic showing an orange fox wearing a green scarf discovering a magical key, unlocking a door, and entering a library. Keep the fox identical in every panel."

ChatGPT produced four almost 3D illustrations. The lighting seemed more computer-generated, though the fox remained recognizable throughout as requested. The style leaned toward children's book illustrations rather than a conventional comic.

Nano Banana 2.1 delivered something much closer to a traditional comic strip, with bold outlines, clearly defined panels, and a consistent cartoon fox. Its version began with the fox finding a key in the forest, then examining it, unlocking a door in a tree, and finally entering a magical library. The storytelling was clearer, and the character's appearance stayed impressively consistent, right down to its green scarf and white-tipped tail.

Google edged ahead here because its four panels felt like parts of the same illustrated story. ChatGPT's individual images were more elaborate, but the fox's proportions and presentation varied slightly, and the sequence felt more like four separate illustrations.

Dragon on a desk

A desk with a dragon sketched in a notebook. ChatGPT on the left, Gemini on the right. (Image credit: ChatGPT/Gemini)

For my final test, I wanted to see how accurately the models could follow a complicated collection of specific instructions. AI image generators have a habit of overlooking small details when a prompt contains several independent requirements. Getting everything into the picture is one challenge, while getting the exact numbers and arrangements right is another.

The prompt was, "Create a realistic overhead photograph of a wooden desk with a notebook showing a dragon, a compass pointing north, a pocket watch showing 3:15, exactly three red wax seals, and a blue gemstone."

ChatGPT produced an attractive overhead photograph of an antique wooden desk, with an open notebook containing an intricate dragon illustration. The compass, pocket watch, three red wax seals, and bright blue gemstone were all present, and the composition looked carefully arranged.

Nano Banana 2.1 also included the requested objects, but the notebook featured a smaller, cleaner dragon illustration, while the wax seals, gemstone, compass and watch were arranged around it. Google's result looked more like a casually assembled collection of historical objects, but it did make a common AI error and got the time wrong, setting it for 3:00, not 3:15.

Approaching parity

You can access Nano Banana 2.1 in Gemini if you're a Google AI Plus, Pro or Ultra subscriber as it is the new default model. Free users will still be getting the older Nano Banana 2 model.

Nano Banana 2.1 is an excellent image generator, but ChatGPT is already so capable that Google's improvements don't necessarily translate into an obvious advantage. Both tools produced attractive, detailed images from relatively short prompts, and neither suffered the glaring visual mistakes that used to make AI-generated pictures instantly recognizable.

That said, different prompts, styles, and repeated generations could easily produce different results. What these tests do suggest is that the leading AI image generators have reached a remarkably high level of capability and are approaching parity in many everyday creative tasks.

That makes the competition more interesting, even if it makes choosing between them harder. The differences increasingly come down to taste and interpretation. Nano Banana 2.1 may be Google's latest impressive upgrade, but ChatGPT has made sure it won't be claiming the image-generation crown without a fight.

• Poor workforce health could cost the UK economy £241 billion a year
• This is equivalent to £7,000 per worker, and represents around 8% of UK GDP
• Employees who attend while ill increase the burden, as they are not working to their full ability

New research into the impact of sickness in the British workforce has highlighted the financial loss per worker, and makes the case for staying off work while ill rather than “braving it.”

These findings highlight that a poor workforce has an impact on businesses and the wider economy, with an estimated £241 billion a year lost due to sickness.

But the loss is made worse by an increased burden caused by employees attending work while unwell, and unable to work to their full capabilities.

Psychological distress

Conducted for Vitality's Global Health Index 2026, the results paint a picture of a workforce lacking focus – something that appears to be impacting the economy by around 8% of GDP, equivalent to £7000 per worker.

The research also points to challenge posed by employees with a high mental health wellbeing risk impacting productivity due to being less focused at work. While the research doesn't specify concerns over AI, it would not be a surprise to learn that the competition they pose to productivity-based roles could have an impact.

"Britain's search for growth could be aided by paying more attention to one of our greatest productivity issues," noted Neville Koopowitz, Global CEO of Vitality.

"This is not people just being off sick, it is people at work, struggling with their health and wellbeing, most likely within organisations that do not measure the health, wellbeing and productivity challenges affecting their workforce."

In the Gen Z cohort, the findings suggest what it calls serious psychological distress among 80% of those surveyed.

Worryingly, around 33% of young women symptoms of depression, with mental ill health across the board losing an estimated 30 days of productivity time per person per year.

Wellbeing and productivity challenges

It seems that attention to these issues is only given in certain quarters. Vitality found that less than a third of UK employers offers counseling or mental health support.

“It might be an investment,” says Koopowitz, “yet it is those employers who are taking the time to understand their employees’ health and associated challenges and then addressing it that are seeing real returns. The gap between organisations that do this and do it well, will become increasingly visible, with the cost of inaction today already at £241 billion a year.”

With around 1 million young people in the UK classified as not in education, employment or training (NEET), workplace-appropriate mental health support and counseling could help employers make the early-career workplace a more accommodating and positive environment. Given the challenges that AI adds into the mix, inaction is not an option.

If you’re shopping for a new pair of headphones, you’re not exactly short of options. As an audio reviewer, I naturally have to consider sound quality when recommending headphones, but I’m also partial to pairs that look distinctive. Who wouldn’t want to stand out in the crowd for all the right reasons? And if you can get a set for basement-bargain prices without too many compromises, who am I to say no?

That’s where Nothing comes in. The fledgling brand has made a serious mark on the headphone industry with its distinguished style, which you can see clearly across both its over-ear and in-ear headphone models. What’s more, we’ve found they perform exceptionally well under review. A double win.

Right now, some of our favourite pairs have received generous discounts on Amazon, including a 44% saving on the highly rated Headphone (a), and the same 44% discount on the Ear (a) if you prefer in-ear buds. And Black Friday 2026 sales in Australia haven’t even started yet!

The only caveat with these deals is that they’re exclusive to Prime members and ship from Amazon Japan, so you’ll need to wait a bit longer than usual for them to arrive at your doorstep. If you’ve not been a Prime member before, you can sign up for a free 30-day trial, make your purchase, then cancel any time.

The Ear (3) are Nothing’s current flagship buds and admittedly, at full price, they are a little tricky to recommend over the Ear (a) below. This mega discount takes care of that, however, making it money well spent. You get excellent ANC, LDAC support for higher-res streams and some hearing tests that help tailor the sound to your preferences. We said the sound quality had “excitement, energy, bass clout and oomph” in our Nothing Ear (3) review, and we love the level of customisation on offer. Note that the best price is on the white model only.View Deal

We love the Nothing Ear (a) earbuds. They sound fantastic, offer excellent noise cancellation and they boast an impressive 42.5 hours of battery life. Plus, if you pair them with a Nothing phone, you get the added benefit of ChatGPT integration. They were around AU$10 cheaper during Black Friday last year, making this a great price outside of a major sale. All three colours are discounted to below AU$100.View Deal

They’ve now been superseded by the Nothing Headphone (1) Pro for flagship status, but the Headphone (1) are well worth considering for their looks alone. We thought the audio was a little flat in our review, at least compared to price-comparable rivals, and this discount helps to fix that gripe. Plus, they’re brimming with useful features like tactile controls, and they deliver strong ANC.View Deal

The jewel in Nothing’s headphone crown if you ask me, the Headphone (a) represents a big improvement over the Headphone (1). The sound quality is objectively better, while the customisable EQ, spatial audio, and adaptive noise cancellation make them more useful for day-to-day listening. Add the obscene 135-hour battery life and IP52 dust and water-resistance rating into consideration, and it's basically golden. Only the yellow version is available at this price. Black, white and pink are all available for a smidge under AU$210 via Amazon US. View Deal

• Human error remains one of the biggest challenge to online security, according to industry leaders
• 68% of security professionals cite human error, compared with just 55% for email-based scams
• A worrying 80% of contributing factors in recent security incidents can be traced back to poor user practices or lack of training

New research into the challenges facing cybersecurity professionals has revealed the naked truth: that humans are considered the main challenge to security.

A new Kaseya survey of over 1,100 MSPs and IT professionals around the world uncovered a startling trend: that the risks posed by humans to their cybersecurity are considerable. Some 68% of respondents cited human error (including social engineering and distraction) as the most concerning attack vector over the next 12 months.

With the overwhelming majority of cybersecurity incidents attributable to human error, security suites and the teams that use them are increasingly focusing on ways to mitigate the challenges posed by people.

Human error

We know that humans need security as much as the data they’re working with. But the results of this survey illustrate a sad truth: that cybersecurity professionals are more concerned about user error and the vulnerabilities of “soft” targets than the threats from automated attacks against network hardware and servers.

Hackers know that humans make mistakes. Organizations need to be on the same page.

From the results of the report, we can see that after the prospect of risks caused by human error, social engineering, and distraction, 55% of those surveyed are concerned about attacks via email. Gullibility is also considered a risk (41%) as is a lack of cybersecurity training for users (40%).

Auditing isn’t always the answer, either. Of those surveyed, 65% said they would pass a compliance audit today, but 49% noted that incomplete documentation and procedures could create problems.

True cyber resilience

Worryingly, only 20% of IT departments have a budget that is growing in line with the risks they face, with 77% describing themselves as under-resourced.

“People aren’t perfect, and everyone knows it,” noted David Baggett, SVP and GM of Security Suite at Kaseya. “Yet much of cybersecurity still depends on people making the right decision every time.”

Unfortunately, it seems as though human error is here to stay. So what is the solution? Bagget says its about factoring in the poor practices and errors.

“The path to true cyber resilience isn’t about eliminating mistakes. It’s about building an organisation that can withstand them. Expect mistakes, account for constraints and engineer around them.”

As far as Kaseya is concerned, that means building security systems and practices that withstand those mistakes.

• Ukrainian drone strikes knock key Yandex data centers offline in Central Russia
• Bot traffic on Yandex networks has dropped significantly as a result, with many services knocked offline and Yandex's Cloud Zones operating with 'limited functionality'
• Data centers are quickly becoming key military targets as intelligence gathering and communications rely on cloud services

Ukraine has launched a series of drone strikes over the past week that have knocked multiple Russian data centers offline, resulting in internet bot traffic from the country dropping to almost zero.

The data centers were operated by Yandex - Russia’s domestic answer to Google - and traffic for the search engine has dropped to just 0.4% of its two week high.

Yandex traffic first dropped on October 8, 2026 after Ukraine struck a data center in Sasovo, Russia, and a second strike on a data center 220 miles to the west of Sasovo on October 10.

Russian bot traffic collapses following strikes

It is important to define what this bot traffic actually is. There is a common tendency to automatically associate 'bot traffic' - especially Russian bot traffic - with the fake social media accounts that spread Kremlin-aligned propaganda. The bot traffic taken offline in this context is more likely to be automated programs that crawl pages, index search results, and gather data for AI platforms.

The drone strikes have been particularly devastating for Yandex’s cloud services. The Yandex Status Board shows that Yandex’s Central Russian Cloud Zones are currently unavailable, and the platform is running in a “limited functionality” mode with most services offline.

Ukraine’s strikes on data centers appear to be retaliatory. Russia began striking Ukrainian data centers in late September under the reasoning that they were being used by Ukrainian military intelligence. A strike on a building within the business center of Kyiv killed four people and knocked internet connectivity offline for 100,000 households in the Kyiv region.

These strikes have been the latest escalation of the war which has been raging since the Russian invasion in 2022, although some observers say the war began with Russia’s annexation of Crimea following the collapse of Ukraine’s Pro-Russian government in 2014.

Since the outbreak of the war, civilian infrastructure has been a key target of strikes by Russia. Energy production and infrastructure has been repeatedly hit in strikes specifically orchestrated to demoralize the Ukrainian population, especially during the winter.

Data centers become a key target

As military intelligence and warfighting capabilities become increasingly digital, data centers have become a key part of war.

At the outbreak of hostilities between Iran and the US and Israel, numerous AWS data centers in Bahrain and the United Arab Emirates were struck by missiles knocking numerous services offline in the region.

Iranian media later reported that the strikes were deliberate, and were intended to “to identify the role of these centers in supporting the enemy's military and intelligence activities.”

To adapt to the role that data centers now play in military conflict, Palantir have begun deploying shipping containers packed with Nvidia's B300 AI accelerators to be used in warzones. They are designed to be quickly set up and capable of processing information within any connection to public cloud infrastructure and can be fully isolated from external networks when required.

Iran’s strikes in the Middle East show just how vulnerable data centers and the infrastructure they rely on are to airborne munitions. Some experts have argued that the US should begin building redundancy and decentralization into the infrastructure required to coordinate logistics, intelligence gathering, and communications.

• Cybercrime group BYOD has published records on 3,615 Trump Mobile customers, including names, home addresses, phone numbers, and order details
• BYOD says it got in via malware on a Liberty Mobile employee's device, and that Trump Mobile told it there was "no team" to respond
• The published data appears to cover the wireless service rather than T1 phone orders, and includes abandoned sign-ups and the Trump Organization's own CIO

A newly formed cybercrime group calling itself BYOD says it has stolen and published personal data belonging to 3,615 Trump Mobile customers, the latest setback for the Trump-branded wireless venture and its gold-colored T1 smartphone.

Straight Arrow News first reported the leak after reviewing the dataset, which contains names, email addresses, phone numbers, home addresses, and order details, and contacting people listed in it.

Trump Mobile has not responded to Straight Arrow, and neither the Trump Organization nor Liberty Mobile, the Florida-based operator behind the service, answered questions from The Register.

A ransom note that skipped the ransom component

BYOD posted the data to its dark web leak site in late September 2026; a ransomware tracker cited by IBTimes UK dates the entry to September 29. The Register describes BYOD as a new ransomware-as-a-service operation, with Trump Mobile as only the third organization named on its site, though IBTimes UK notes that nothing published so far shows encrypted systems or a ransom demand.

The group's account of how it got in is consistent in outline, even if the details differ slightly: BYOD told International Cyber Digest it compromised a Liberty Mobile employee with an infostealer, while Cybernews and Android Authority describe a remote access trojan.

Regardless of how they got onto the device, BYOD's messaging indicates it pivoted from that foothold to exposed Trump Mobile subdomains, still has live access to a customer dashboard, and claims neither company used multi-factor authentication.

As proof, it sent Straight Arrow a dashboard screenshot showing one customer's personal details, but said access has not been independently confirmed yet. The more damning part of this is that Trump Mobile's answer to this has been "We have no team to handle this" and branding anyone who hacked it a terrorist, an exchange that BYOD alleges took place.

The data, however, has held up; Straight Arrow reached people who confirmed their details were accurate, Cybernews said its samples looked genuine and did not match earlier breaches, and PCMag reached three people in the file, one of whom was surprised to find an accurate record of canceling a 30-day unlimited plan.

BYOD may not even be the first to cash in: a separate group, EndZone, claimed on September 23 to have taken data on 4,000 Trump Mobile users, IBTimes UK reports, and security researcher Dominic Alvieri told The Register the EndZone dump "appears to be the same original breach." BYOD denies a formal tie, with a representative telling PCMag that one member "just knows a few people behind EndZone," according to IBTimes UK.

At launch in June 2025, Eric Trump promised that "Trump Mobile is going to revolutionize cell phones" - but what followed has been less revolutionary.

The "Made in America" promise disappeared from the website within two weeks. Shipping slipped from August 2025 to December, then March, then April, and in April the company rewrote its preorder terms to state that "a deposit is not a purchase." In January, lawmakers urged the FTC to examine the brand's marketing.

For consumers, there is a silver lining: No payment card data has been reported in the dump, so the immediate risk is phishing rather than direct fraud. This doesn't completely solve the problem, however: names, phone numbers, home addresses, and order histories are exactly what scammers need to pose convincingly as a carrier or a delivery service.

Anyone who signed up, or merely started signing up, with Trump Mobile should treat unexpected messages about orders, refunds, or account problems with suspicion at a time when Trump Mobile continues to drag its feet on the matter, much like its May breach, when the company said it was still deciding whether to notify people. This time is a little different, however: it has said nothing at all.

When you read ‘video game fashion’, you likely picture faintly embarrassing Pokémon-inspired onesies, awkward PlayStation-branded rucksacks, or cringeworthy Sonic the Hedgehog T-shirts. But the world of video game and lifestyle collaborations has changed markedly over the last few years. With gaming becoming more of a major cultural force, higher-end brands have started to get involved.

You can probably already think of some examples. It wasn't long ago that cult streetwear brand ARK/8 launched a capsule inspired by the world of Elden Ring at the peak of that game’s popularity, including a coat that made headlines thanks to its eye-watering $1,700 price tag.

On the more approachable end of the cost spectrum, premium watchmaker Hamilton has been collaborating with some of the biggest names in gaming for years and has big plans for the Call of Duty: Modern Warfare 4 launch in just a few weeks.

From Hollywood to high scores

(Image credit: Hamilton)

“What’s fantastic about Hamilton is that the brand keeps working to its DNA and its roots,” explains Hamilton International CEO Francesca Ginocchio as we chat about the company’s gaming efforts. “I would say the American spirit of the brand, now with Swiss movement, is very strong, as is the link with cinema and entertainment, which has been so consistent starting back in 1957 with Elvis Presley.”

The King of Rock and Roll himself famously owned and frequently wore a Hamilton Ventura, the watch even appearing on his wrist in the 1961 musical film Blue Hawaii. The asymmetrical design became so synonymous with the star that the model is still dubbed the ‘Elvis watch’ today.

Hamilton continues to work closely with the glamorous world of Hollywood today, collaborating with directors including Christopher Nolan and Denis Villeneuve to create custom pieces for titles like Tenet and Dune: Part Two.

“We have been in more than 300 movies so far, but we keep on trying to update ourselves,” Ginocchio continues. “Gaming is really a new part of the DNA of a brand linked to entertainment and movies. It’s the next step.”

“We have been in more than 300 movies so far"

Francesca Ginocchio, CEO

Hamilton’s gaming collaborations began back in 2021, when eagle-eyed Far Cry 6 players might have spotted a custom titanium field watch on protagonist Dani Rojas’s wrist. Designed specifically for the game, Hamilton made 1983 identical real-world pieces available for dedicated fans to buy.

For a tie-in product, the design was understated: a vintage brown leather strap, military-style titanium case, and a simple watch face with beige numbering. It looked like an object that could plausibly exist in the game’s fictionalized Yara setting, which draws heavily on the aesthetic of modern-day Cuba. You would be hard-pressed to identify it as a collab unless you spot the vertical line dividing the dial’s number six in a reference to the game’s logo.

According to Ginocchio, this subtlety is by design and all in view of delivering a product that feels authentic to the fictional universe: “We have this credibility in the entertainment world, and we apply how we work with movies to the gaming industry. Basically, we try to do the best watch for the character in the movie or in the game.”

Mission readyFutureFutureFuture

This approach is evident in Hamilton’s latest video game endeavor, a collaboration with the upcoming blockbuster first-person shooter Call of Duty: Modern Warfare 4.

Fan-favorite hero Captain Price wears a Hamilton Khaki Field watch on his wrist throughout the story campaign, with a matching model releasing alongside the game.

It’s a pairing that Ginocchio says came naturally given the brand’s heritage. “When the Call of Duty collaboration came up, a big pillar of a business and big part of our story, or DNA, is the Khaki collection,” she explains. “It was born in the US and used in World War One and World War Two. It’s very recognizable; Khaki, I would say, is almost a brand within the brand. If you have a Khaki on your wrist, everybody can tell you’re wearing a Hamilton, and Call of Duty was the perfect combination to put all that storytelling and this part of the collection in a game.”

"Khaki, I would say, is almost a brand within the brand"

Francesca Ginocchio, CEO

The Hamilton x Modern Warfare 4 special edition is outfitted with a 38mm black PVD-coated stainless steel case, a rugged material in an unassuming color befitting a man trying to keep a low profile.

There is some brightness, though, in the highlighter yellow arrow of the second hand and little triangle shape at 12 o'clock – a dominant color in the upcoming game’s marketing. Otherwise, the dark green numbers are perfectly legible against the black dial and, if you look closely, you’ll spot the Korean character 사 in place of the usual Arabic number, reflecting the game’s Korea setting and mirroring its branding.

The watch comes bundled with a custom travel case. (Image credit: Hamilton)

It’s paired with a durable green cloth strap for a look that the brand says is inspired by British-issued service watches. Interestingly, this model uses Hamilton’s mechanical H-50 movement, which offers an 80-hour power reserve but requires hand winding.

Ginocchio notes this as another conscious choice steeped in history, saying that "Mechanical hand winding was really important for us" as it harkens back to old service watches and "really represents Hamilton."

She goes on to explain that, although the finished watch is something entirely new, it’s important to use existing real-world models as a starting point to maintain believability. She also says every design decision was the result of close conversation with Activision and the Call of Duty team: “It’s really done in collaboration, because we don’t want to make the perfect Hamilton watch. We want to make the perfect watch for the character of the game, and that’s a different story.”

Custom packaging is also an important part of the real-world product, and the Hamilton x Modern Warfare 4 watch comes nestled inside a cigar-box-like travel case – a not-so-subtle allusion to Price’s frequently lighting up in-game.

Never slowing down

The Hamilton x Call of Duty Black Ops 7 limited edition watch was much less distinct. (Image credit: Hamilton)

This isn’t Hamilton’s first Call of Duty collaboration: the brand released a special edition Black Ops 7 model last year (worn in-game by multiplayer operator Axel Vermaak).

Last year’s watch was noticeably tamer, however, with far fewer customized elements and little more than a branded caseback to set it apart at a glance – something that many online commenters were quick to lament.

The CEO says that the choice to make this year’s model more adventurous came as a result of such feedback: “Last year, we had some enthusiastic people, and some people who were saying, ‘Oh, but we wanted something more customized. You could have done more!’ [...] People that buy a Hamilton watch, they are passionate, and so we listen to them.

"This year we said, ‘Okay, let’s do something stronger. Something with more differentiation.’ [...] This year, as it’s the second collaboration, we feel we can push a little more and go with something more aggressive.”

Of course, for well-known brands like Hamilton, gaming collaborations are a way to reach a new audience outside of their core market. “We have our usual watch 'passionates' who have to buy everything we propose, and they will go for this because it's something totally new, totally fun,” she adds. “Then we have many, many people who are gaming passionate – but they don’t know about Hamilton, and they get to know us.”

It’s a strategy that seems to be working, with Ginocchio stating that Hamilton’s top sellers “are all watches with a movie or storytelling behind them” and teasing that there are even more collaborations to come. “We already have three new projects for 2027,” she reveals. “It’s a long-term strategy; we want to keep on going and be consistent.”

You can learn where to buy the Hamilton x Modern Warfare 4 special edition via the brand’s official website, or see it in-game when Modern Warfare 4 launches for PlayStation 5, Xbox Series X and Series S, Nintendo Switch 2, and PC on October 23, 2026.

• Microsoft CEO Satya Nadella says Copilot Code and local AI make every Windows PC an infinite software factory with no cloud token spend
• Rather than killing app stores, Microsoft is moving app hosting into Copilot, while IT leaders stay wary of citizen-built tools
• Microsoft's own engineers say local inference doesn't make sessions offline, and the routing ships later this month

Microsoft CEO Satya Nadella has a new pitch for the Windows PC, and it targets one of the most lucrative ideas in consumer tech.

In a post on X summarizing Microsoft's recent Windows and Surface event, Nadella told followers they no longer have to settle for whatever an app store happens to stock, because Copilot can now build the software they need on the desktop without spending any cloud tokens.

He wrote this idea without mincing any words, saying, "Your PC becomes an infinite software factory."

Today marks a new chapter for Windows, as we bring unmetered intelligence to every desk and every home, and make every PC a place where agents can work securely on your behalf.Some highlights of what we announced:• MAI-Code-1.1 Flash: 137B parameter coding model w/ 256K context window, which is now optimized to run on your PC!• GitHub Copilot now hands off work to local models like MAI-Code-1.1 Flash, helping projects cost a lot less without sacrificing quality.• With Hybrid Intelligence, Copilot can now take action directly on the PC and keep sensitive work on your device.• And with Code in Copilot, you can essentially build any software you need on your desktop, without any cloud token spend, and it’s just super at it. You’re no longer limited to what’s in an app store! Your PC becomes an infinite software factory.• Security is foundational to all this, which is why we are also bringing together Windows and Agent 365 so agents can work within secure boundaries on-device, including MXC a local sandbox for agent execution. Windows becomes your secure agent box!• All this comes to life on a new generation of devices, like Surface Laptop Ultra, powered by NVIDIA RTX Spark.Can’t wait to see what you build with all this.October 7, 2026

From Gates' factory to your desk?

The narrative is a striking line from the head of a company that sells more software than almost anyone.

But read Microsoft's own engineering documentation for the same launch, and the infinite factory starts to look rather finite: it runs best on a $5,899 laptop, it still reaches into the cloud, and the businesses Microsoft most needs to convince have spent the past year telling researchers they are worried about exactly this kind of thing.

More importantly, the factory metaphor is not new for Nadella: In a 2025 staff memo, he recalled that Bill Gates had envisioned "not just a software company, but a software factory," and argued the AI era demanded a rethink.

At a recent customer gathering, the CEO said Microsoft had progressed from code assistance through agentic coding "to software factory all within one year," and he now wants customers to open software factories of their own, spending more with Microsoft along the way.

The October 7 2026 launch builds on this ambition: a Windows Experience Blog post by Windows and Devices chief Pavan Davuluri describes hybrid intelligence, in which Copilot reads local files with permission, acts across Windows, and hands work to on-device models.

Copilot's Code experience builds an app, dashboard, or workflow from a plain-language description, and on Windows, Microsoft says it can produce native apps from one prompt. Jared Spataro, Microsoft's chief marketing officer for AI at Work, argued that building such tools is becoming "as basic a skill as writing a memo or modeling a budget."

The engine underneath is MAI-Code-1.1 Flash, Microsoft's in-house 137-billion-parameter coding model, which Nadella's post said is now optimized to run on a PC. There is a catch, however: despite Microsoft painting it as a situation where consumers are finding their AI needs "outpacing what their cloud budgets can support."

Microsoft's own engineers are more grounded about what the process entails, saying GitHub Copilot's automatic mode decides task by task whether work runs locally or in the cloud, and one diagram caption is blunt: "local inference does not make the session offline."

That routing doesn't ship until the end of October 2026; the hybrid Copilot features reach Copilot+ PCs over the coming months, and Code is rolling out first through Microsoft's Frontier early-access program. This is interesting, but somewhat predictable, because Nadella's approach, as stated, would cannibalize Microsoft's token-based business model.

An approach that routes harder tasks back to the Azure cloud makes far more sense by conserving compute where possible while still putting Microsoft's cloud compute at the center of things.

The hardware is another factor that one might be overlooking here: according to the Command Line post, the on-device version of MAI-Code-1.1 Flash is quantized down to 53GB, roughly an 80% reduction from the cloud model, and peaked at 75.5GB of memory at full context on Microsoft's test machine, a Surface Laptop Ultra built on Nvidia's RTX Spark.

That matters because eight Surface Laptop Ultra configurations are in play; only one has more than 64GB of memory, per TweakTown: the 128GB model at $5,899, which costs $3300 more than the base model and isn't available to purchase yet, meaning that potential users can't really test out how well it does compared to cloud-based solutions at a time when all performance-related figures come from Microsoft.

The death of modern enterprise software? Not yet

Despite the hardware limitations in play, it might indicate a death sentence in the making for most basic enterprise-level software solutions, even as more ambitious ports have already been in play: Developer Brandon Thomas, for one, has just released seven open source Adobe Creative Cloud alternatives that he says he built with Anthropic's Claude Opus 5.5, declaring that "software is over."

His own roadmap, however, puts the Photoshop stand-in at roughly a quarter to a third of what a professional needs for daily work. Building software has become cheap. Building software someone can depend on has not.

The bigger issue will be enterprise IT, which will, I suspect, keep the gates closed for now; enterprises don't just buy features; they buy someone to call when it breaks, a patch schedule, an audit trail, and a vendor that carries liability.

Getting an open-source or AI-assisted app someone designed with zero liability in tow that could break things, create security risks, and may not come with future software updates that resolve either issue.

The survey data points the same way: a June 2026 Retool study of 307 CTOs, CIOs, and CISOs found 93% concerned about vibe-coded tools reaching production, though Retool sells an internal-tools platform pitched as the governed alternative.

A Harris Poll survey for Dataiku of 600 CIOs found 81% worried that citizen-built AI could expose sensitive data, and over half had already discovered unsanctioned AI in their organizations.

Gartner, meanwhile, predicts that over 40% of agentic AI projects will be scrapped by the end of 2027, with analyst Anushree Verma describing many as "mostly driven by hype and are often misapplied."

Microsoft's own launch materials quietly concede the same, even taking measures to ensure that agents do not run rogue while building software or solutions on-device.

The Command Line post admits Copilot's built-in file tools are policy-checked inside the app rather than isolated by the operating system, and GitHub requires Business and Enterprise administrators to enable a policy before staff can use MAI-Code-1.1 Flash at all, while Microsoft Execution Containers, which are now generally available, allow IT to define what agents can read, modify, or execute.

The real test for Nadella's factory isn't individuals with a $5,899 laptop and a tolerance for alpha software, but businesses willing to take the risk and trust Microsoft's safeguards. Until then, one could assume the AI factory narrative will continue to run well below its advertised potential.

Looking for a different day?

A new Quordle puzzle appears at midnight each day for your time zone – which means that some people are always playing 'today's game' while others are playing 'yesterday's'. If you're looking for Monday's puzzle instead then click here: Quordle hints and answers for Monday, October 12 (game #1722).

Quordle was one of the original Wordle alternatives and is still going strong now more than 1,500 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 #1723) — hint #1 — VowelsHow many different vowels are in Quordle today?

• The number of different vowels in Quordle today is 3*.

* 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 #1723) — hint #2 — repeated lettersDo any of today's Quordle answers contain repeated letters?

• The number of Quordle answers containing a repeated letter today is 2.

Quordle today (game #1723) — 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 #1723) — 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 #1723) — hint #5 — starting letters (2)What letters do today's Quordle answers start with?

• F

• D

• T

• S

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

Quordle today (game #1723) — the answers

(Image credit: Merriam-Webster)

The answers to today's Quordle, game #1723, are…

• FLUFF
• DEBUT
• TEETH
• STEIN

I didn’t make any mistakes today, but this game still took me a long time to complete.

My difficulty was in finding a three-letter combination beginning with U from the letters I had left. Eventually I landed on the only possibility and a rare five-letter word containing three of the same letter — FLUFF.

Daily Sequence today (game #1723) — the answers

(Image credit: Merriam-Webster)

The answers to today's Quordle Daily Sequence, game #1723, are…

• TIDAL
• PAUSE
• SUING
• SHYLY

Quordle answers: The past 20

• Quordle #1722, Monday, 12 October: HELLO, BELOW, RIGHT, SLANT
• Quordle #1721, Sunday, 11 October: LILAC, INLET, TRIAD, SHAPE
• Quordle #1720, Saturday, 10 October: SWEAT, THIRD, RURAL, CLASP
• Quordle #1719, Friday, 9 October: OLDEN, DUNCE, SWATH, HILLY
• Quordle #1718, Thursday, 8 October: MINTY, ROWER, BROOD, RESET
• Quordle #1717, Wednesday, 7 October: PALER, FROTH, DOZEN, ERASE
• Quordle #1716, Tuesday, 6 October: FRESH, SKULK, BATCH, ENSUE
• Quordle #1715, Monday, 5 October: SIXTH, SPEAK, SEDAN, HEIST
• Quordle #1714, Sunday, 4 October: INFER, RAINY, THING, ARENA
• Quordle #1713, Saturday, 3 October: CHORE, FLYER, STICK, PURSE
• Quordle #1712, Friday, 2 October: IVORY, SADLY, ANGEL, LLAMA
• Quordle #1711, Thursday, 1 October: NAVEL, SHEER, BIRTH, TACKY
• Quordle #1710, Wednesday, 30 September: CHOKE, BATCH, MOSSY, EAGER
• Quordle #1709, Tuesday, 29 September: BILGE, MEDAL, LODGE, PORCH
• Quordle #1708, Monday, 28 September: BRUNT, TORSO, STONE, HEATH
• Quordle #1707, Sunday, 27 September: VALET, CHEST, VIGIL, PUREE
• Quordle #1706, Saturday, 26 September: SNORT, BIBLE, LAPSE, CRANE
• Quordle #1705, Friday, 25 September: SWEAR, FLOOD, GUPPY, GROAN
• Quordle #1704, Thursday, 24 September: TOPIC, METRO, SCALP, FREED
• Quordle #1703, Wednesday, 23 September: SHEET, BLAND, JOIST, GUAVA