Apple AI now runs on Google, Nvidia tech: What happens to privacy promise

by Hansraj Agrawal | Jun 12, 2026 | Technology

At the Worldwide Developers Conference (WWDC) 2026, Apple unveiled its biggest expansion of Apple Intelligence since the platform was first introduced in 2024. The company also announced Siri AI, a completely rebuilt version of Siri that can understand personal context, search across emails, messages and photos, answer questions using web information, and perform actions across apps. Apple noted that the latest generation of Apple Intelligence relies on technology developed in collaboration with Google and cloud infrastructure powered by Nvidia GPUs.

The disclosure marks a significant evolution from the original Apple Intelligence vision unveiled in 2024, when Apple emphasised that its AI strategy would rely heavily on on-device processing and an in-house privacy-focused cloud system known as Private Cloud Compute. Despite the shift, Apple insists its approach to privacy has not changed.

The question now is whether Apple is maintaining that promise while increasingly relying on infrastructure and technology beyond its own walls.

Apple’s privacy-first AI strategy began in 2024

When Apple Intelligence debuted at WWDC 2024, Apple sought to distinguish itself from rivals by positioning privacy as a core feature rather than an afterthought. The company argued that many AI tasks should run directly on users’ devices whenever possible. For more computationally demanding requests, Apple introduced Private Cloud Compute (PCC), a cloud-based architecture designed to extend Apple’s privacy protections beyond the device.

Apple described Private Cloud Compute as a system where user data would only be processed for the duration of a request, would not be stored, and would remain inaccessible even to Apple itself. The company also said security researchers would be able to independently inspect and verify the architecture.

At the time, Apple’s message was clear: users could access advanced AI capabilities without handing over their personal information to cloud providers. That privacy-centric approach became one of the defining pillars of Apple Intelligence.

What changed in 2026

The biggest change is not in Apple’s privacy messaging. It is in the technology stack powering Apple Intelligence. In January 2026, Apple and Google announced a multi-year collaboration under which the next generation of Apple Foundation Models would be based on Google’s Gemini models and cloud technology. The companies further stated that Apple Intelligence would continue to run on Apple devices and Private Cloud Compute while maintaining Apple’s privacy standards.

That partnership became more visible at WWDC 2026. Apple’s Security Research documentation states that the latest family of Apple Foundation Models was built in collaboration with Google. Apple also revealed that its cloud-based AI infrastructure now extends to Nvidia GPUs operating within Google’s cloud infrastructure.

Amar Subramanya, Apple’s vice president responsible for AI technologies, said the company works with both Google and Nvidia to extend Private Cloud Compute to Nvidia GPUs running in Google’s cloud while maintaining Apple’s privacy guarantees. In practical terms, this means some of the most advanced Apple Intelligence workloads no longer rely exclusively on Apple’s own server infrastructure. Instead, they can be processed using Nvidia hardware operating within Google Cloud environments. That represents one of the most significant architectural shifts since Apple Intelligence was introduced.

Why did Apple turn to Google and Nvidia

The answer largely comes down to capability and scale. At WWDC 2026, Apple introduced a significantly more capable version of Siri than the one it first showcased in 2024.

Siri AI can understand personal context, search across messages and emails, retrieve information from photos, understand on-screen content, answer questions using information from the web, perform actions across apps, and maintain conversations across devices. These capabilities are far more demanding than traditional voice assistant tasks. But what could have led to the shift?

An earlier report by The Information suggested that the increasing computational requirements of advanced AI models may have influenced Apple’s decision to partner with Google and Nvidia. Further, Nvidia’s role addresses another challenge: compute. Large AI models require specialised processors capable of handling massive inference workloads, and Nvidia’s GPUs have become the industry standard for running advanced AI systems at scale.

Then comes the point of scale. According to Counterpoint Research, Apple had cumulatively shipped more than 450 million Apple Intelligence-capable iPhones by the first quarter of 2026. As per the report, Apple currently has the largest installed base of GenAI-capable smartphones among all smartphone brands.

That figure only accounts for iPhones. Apple Intelligence features are also accessible on iPads, Macs, Apple Watches, and Vision Pro devices, significantly increasing the number of users who could potentially use these features.

Supporting AI services for such a large installed base requires enormous computing resources. In that context, Google’s cloud infrastructure and Nvidia’s GPUs provide Apple with capabilities that would be difficult and expensive to replicate quickly using only its own infrastructure.

However, if the integrations between the companies run this deep, then how will Apple maintain the privacy guarantees that have been central to its AI strategy since 2024? Part of the answer lies in a technology Apple is now using alongside Private Cloud Compute: Nvidia’s Confidential Computing.

What is Nvidia’s Confidential Computing

A key component of Apple’s updated architecture is Nvidia’s Confidential Computing technology. To understand why it matters, it helps to understand how data is typically protected.

Most digital systems encrypt data while it is stored and while it travels across networks. However, during computation, data becomes exposed during processing. Confidential Computing aims to close that gap.

The technology creates a protected execution environment that helps keep data secure even while it is actively being processed. Nvidia says its Confidential Computing technology uses hardware-based Trusted Execution Environments (TEEs) to protect data while it is being processed and help prevent unauthorised access.

The technology has become increasingly important as AI models grow larger and require powerful cloud-based GPUs to perform inference and reasoning tasks. Nvidia has positioned Confidential Computing as a way for enterprises and governments to run sensitive AI workloads while maintaining strong security protections. Apple is now using that same technology as part of its broader AI infrastructure.

How Private Cloud Compute and Confidential Computing work together

Although Apple and Nvidia are both talking about privacy and security, they are solving different problems.

Private Cloud Compute is Apple’s privacy architecture. Apple’s Private Cloud Compute architecture governs how requests are processed, limits what data can be accessed for a task, prevents retention of personal information after processing, and allows independent researchers to verify the software running on its servers.

Nvidia’s Confidential Computing technology focuses on the hardware environment used for computation. It secures the hardware environment where AI processing occurs, including protecting memory, encrypting active workloads, and safeguarding computations from unauthorised access.

In simple terms, Apple defines the privacy rules, while Nvidia helps secure the hardware running those rules. The two systems are complementary rather than interchangeable. Put simply, Apple’s Private Cloud Compute governs the privacy model for AI requests, while Nvidia’s Confidential Computing technology helps secure the infrastructure on which those workloads run.

Has Apple really changed its privacy approach?

This is where the debate becomes more nuanced. Apple’s public position is that its privacy guarantees remain unchanged. The company says user requests processed through Private Cloud Compute are still protected, data is not stored after processing, and personal information remains inaccessible to Apple, Google, or other third parties.

However, there is a meaningful difference between Apple’s 2024 and 2026 architectures. In 2024, Apple controlled nearly every layer involved in cloud-based AI processing. The company designed the hardware, operated the infrastructure, controlled the software stack, and defined the privacy architecture.

In 2026, Apple still controls the privacy architecture and the rules governing how data is handled. But some of the underlying infrastructure is now supplied by Google and Nvidia. That distinction matters because it shifts the discussion away from privacy alone and towards ownership. The real shift is not necessarily privacy, but control.

Apple no longer owns every layer of the stack supporting Apple Intelligence. Instead, it is extending its privacy architecture onto infrastructure supplied by external partners. That is a different proposition from the one Apple presented when Apple Intelligence first launched.

At the same time, it does not automatically mean Apple has abandoned its privacy commitments. The company’s argument is that the privacy protections users receive are determined by the architecture governing the system, not by who owns the physical servers or chips.

The bigger question

The broader challenge for Apple is no longer building AI features. It is convincing users that those features can scale without compromising the trust that has differentiated Apple from many of its rivals.

As Apple Intelligence expands across hundreds of millions of iPhones, iPads, Macs, and other devices, the company will increasingly rely on partnerships that were largely absent from its original AI strategy. Google provides key models and cloud technologies, while Nvidia supplies the hardware that powers some of the underlying AI workloads.

For users, the ultimate test will not be who provides the infrastructure, but whether Apple’s privacy safeguards continue to work as advertised. If Private Cloud Compute delivers the same protections regardless of whether workloads run on Apple-designed systems or on infrastructure supplied by partners, most users are unlikely to object.

WWDC 2026, therefore, marks an important evolution in Apple’s AI strategy. The company is no longer attempting to build every layer of the stack on its own. Instead, it is combining its privacy architecture with technologies from some of the biggest players in AI. Whether that approach strengthens Apple Intelligence without weakening trust is a question that will only be answered as these features roll out to users over time.

What Apple announced at WWDC 2026

The infrastructure discussion matters because it underpins a major expansion of Apple Intelligence.

The centrepiece of WWDC 2026 was Siri AI, a rebuilt assistant capable of understanding personal context, maintaining conversations, searching across apps, understanding on-screen content, and retrieving information from the web.

Apple also announced:

Spatial Reframing, Extend, and upgraded Clean Up tools in Photos

AI-powered tab organisation, page monitoring, and extension generation in Safari

Photorealistic image generation through Image Playground

Intelligent suggestions in Messages and Mail

Call Context, which surfaces relevant information during phone calls

Natural language event creation in Calendar

Describe a Shortcut for generating automations using natural language

AI-powered search and summaries in the Home app

Expanded accessibility features powered by Apple Intelligence

Source link

From Reels to risks: How scammers are turning videos into malware traps

by Hansraj Agrawal | Jun 12, 2026 | Technology

You are scrolling through Instagram when a video appears showing how to unlock Spotify Premium for free. The clip has a polished voiceover, simple step-by-step instructions, and more than 100,000 views. It looks no different from the countless tutorials users save and revisit every day. You follow the steps. Days later, your passwords, financial information, and stored credentials are in someone else’s hands.

That scenario is no longer hypothetical. Researchers at ReversingLabs recently uncovered two separate cybercrime campaigns operating through TikTok and Instagram Reels, using tutorial-style short videos to trick users into downloading malware or handing over sensitive information through malicious websites. The attacks succeed not because they are highly sophisticated, but because they feel familiar, carefully designed to blend into platforms users already trust.

At the centre of these campaigns is VidarStealer, a malware-as-a-service infostealer built to harvest passwords, browser data, cryptocurrency wallet information, and other credentials. With subscriptions reportedly starting at around $300, the tool has dramatically lowered the barrier to entry for cybercriminals, making large-scale social media scams easier and cheaper to launch than ever before.

How social media became a malware playground

Researchers at cybersecurity firm ReversingLabs (RL) have documented two social engineering attack techniques that target users through short-form videos, primarily on TikTok and Instagram Reels. The campaigns, which promise free access to paid software like Spotify Premium and Microsoft Word, represent a significant evolution in how phishing operates. Instead of phishing emails, cybercriminals are now hiding in plain sight on social media, blending their schemes into the creator content users trust and engage with daily.

People are already looking for scams in their email inboxes and text messages, but not as much on their social media feeds, especially when posts are framed as being helpful rather than carrying the urgency or sob stories associated with stereotypical phishing attempts.

That shift in framing is precisely what makes these campaigns effective. A tutorial about unlocking Spotify Premium looks, in every respect, like the thousands of legitimate life-hack videos that populate a user’s feed. There is no misspelled domain name in a subject line, no unfamiliar sender. There is just a video, and it looks like every other video.

Using social media is free and rewards frequent uploads. By using multiple platforms, accounts, and posts, attackers are able to access many users. The economics are attractive: no bulk email infrastructure, no cost per send, and a built-in recommendation engine willing to do the distribution work.

The growing role of social media as a search and discovery platform

Social media platforms are no longer just spaces for entertainment; they have quietly become the internet’s new search layer. Google itself has confirmed that over 40 percent of Gen Z prefer Instagram or TikTok over Google for search, while Google usage among Gen Z has dropped by nearly 25 percent compared to Gen X. According to GRIN’s report, The Power of Influence, Instagram now leads product discovery among Gen Z at 30.4 percent, followed by TikTok at 23.2 percent, with Google trailing at 18.8 percent. Users are not just passively scrolling; they are actively searching for software guides, tech fixes, and product recommendations through the feed.

This behavioural shift has created a significant opening for attackers. On TikTok, people do not just scroll for inspiration, but actively look for answers, whether finding a restaurant, a solution to a problem, or an honest product review, increasingly going straight to TikTok instead of Google.

The malicious campaigns documented by ReversingLabs are built precisely for this environment, using descriptions and tags to make content appear as legitimate customer support pages, positioning themselves directly in the path of users who are already looking for help. When the feed doubles as a search engine, a malicious tutorial is only one recommendation away.

Two campaigns, two playbooks

RL’s researchers identified two distinct approaches, each designed to game social media differently.

The first involves fake tutorial accounts built to impersonate legitimate tech support. The malicious accounts use usernames like “windows.tips” or “windows.insights” and the same blue and white profile picture, mirroring the colour palette of the official Windows social media account to establish credibility. The videos themselves are clean and professional, featuring what appear to be AI-generated voiceovers walking viewers through step-by-step instructions, for instance, how to access Windows PowerShell and run a command to supposedly unlock Spotify Premium for free.

A non-technical user would not know any better and may assume the tutorial is legitimate. Attackers rely on this lack of understanding. The command used will download scripts from a specified address, and some users may believe the domain is Microsoft-affiliated or otherwise trustworthy. What is actually downloaded is something else entirely.

The file delivered through the command is identified as VidarStealer, a popular infostealer malware-as-a-service (MaaS) offering that steals credentials, financial information, and tokens from victims. With an affordable $300 lifetime licence, it is a widely used tool by malicious actors, with usage documented across fake game cheats, malvertising campaigns, and more.

The second campaign takes a different approach. It relies on short videos set to trending music, showing off features of premium software with on-screen text claiming the user has unlocked them for free. The accounts behind these videos appear like regular users at first glance, but their profiles are typically filled with repetitive, near-identical clips promoting free access to services like Spotify Premium and similar tools.

These vague videos prompt users to ask questions in the comments, wondering how the poster managed to get free access. This curiosity plays directly into what the attacker wants. Some videos actively encourage viewers to comment with certain phrases, a strategy borrowed from non-malicious creators like recipe writers who use it to build engagement and foster an audience relationship. Once engagement builds, the attacker replies with directions pointing toward malicious download sites.

Why social media video is trusted more

The success of both campaigns is not accidental. It is rooted in how users relate to video content. What makes these videos dangerous is how clean and professional they are, creating a false sense of authority. Tutorials are frequently liked and saved, as users want to return to them. Saving is a valuable interaction for posts, causing the platform algorithm to push content to more users.

Users may also share tutorials, creating more engagement that content-serving algorithms favour. In one documented example, a video with over 100,000 views had nearly 200 more saves than likes, demonstrating how attackers are specifically targeting the more algorithmically valuable forms of engagement. Each save is a vote of confidence in the algorithm’s eyes and a further amplification of reach.

This is a deliberate strategy, not incidental. Attackers understand how platform recommendation systems work and produce content calibrated to exploit them.

The role of AI in scaling attacks

Running a social media account is a very low-time-investment endeavour, and with AI voice and video generation, videos are becoming easier to mass-produce. Social media provides ample opportunities for attackers to access victims, and there will likely be increasing numbers of these accounts and videos in the coming years.

The ReversingLabs analysis found that at least some of these tutorial videos already use AI-generated voiceovers, giving them a polished quality that signals legitimacy to casual viewers. As AI generation tools become more accessible and cheaper, the barrier to producing convincing, high-volume campaigns drops further. What once required a professional setup — like clean graphics, a confident voice, and a plausible script — can now be assembled in minutes.

Tools such as ChatGPT, Gemini, Midjourney, Adobe Firefly, Runway, and ElevenLabs have dramatically lowered the barrier to content creation. What once required design skills, video-editing software, or professional voiceover equipment can now be produced in minutes using AI-generated images, videos, and audio. This accessibility is not only helping creators but also making it easier for cybercriminals to produce convincing scam content at scale.

Why platforms are struggling to respond

These techniques are difficult to defend against, like any social engineering method. Users who identify the malicious intent may try to warn others in the comments, but most platforms allow creators to delete comments and block commenters, so diligent attackers can suppress this resistance.

Reporting suspicious videos does not always lead to quick action. In their investigation, ReversingLabs researchers reported several scam-related posts on Instagram, but the platform rejected those reports. This highlights a broader challenge for social media companies: harmful content can remain online even after users flag it.

Part of the problem is that moderation systems do not always recognise these videos as dangerous. Even when a report is reviewed by a person, they may not have the cybersecurity expertise needed to understand how a seemingly harmless tutorial could be directing users to malware or phishing websites. As a result, scam videos can continue to spread and attract victims before they are eventually removed, if they are removed at all.

Even when a social media video or account is taken down, it is likely only after it has amassed a large number of views, and threat actors can easily start anew.

The structural mismatch between how fast bad content spreads and how slowly platforms respond creates a window that attackers are actively exploiting.

Removing a scam video or banning an account does not necessarily solve the problem. By the time platforms take action, the content may have already reached thousands or even millions of users. In many cases, attackers have already achieved their goal of spreading malware or collecting personal information.

What makes the issue even harder to tackle is how quickly cybercriminals can create new accounts and upload fresh content. While harmful videos can spread within hours, platform moderation and review processes often take much longer. This gap between the speed of attackers and the speed of enforcement allows cybercriminals to continue targeting users and expanding their reach.

What platforms and users can do to stay protected

Social media scams are becoming harder to spot because they often look like ordinary tutorials or product recommendations. As attackers adapt their tactics, both users and organisations need to broaden their approach to online safety.

Recommended precautions

Audit software installation permissions

Update phishing awareness training

Treat social media as a phishing vector

Report suspicious videos and accounts

Be cautious of “free premium software” claims

One of the key defences against this kind of attack is to regularly audit permissions, ensuring people with installation privileges understand what they are installing. Most examples described in the analysis involve leisure software, but some promise access to professional software, which employees may deem useful enough to attempt to install on work devices.

Phishing training also needs to be maintained and kept up to date so people are aware of the evolving threat landscape. Organisations must broaden their awareness of a variety of vectors and focus on more than just the typical avenues of phishing.

Users are encouraged to report suspicious social media advice even when using personal social media on personal devices. The more reports filed, the more likely it is that accounts are taken down, which does slow down the momentum of attackers.

The unfortunate reality is that these techniques work. Videos are reaching hundreds of thousands of views, thousands of saves, likes, and shares, and hundreds of comments. These are hugely influential on how well content performs, and these techniques leverage that priority. The threat, in other words, is not theoretical. It is already reaching a very large audience — one scroll at a time.

Source link

India leads digital wallet downloads as homegrown apps dominate globally

by Hansraj Agrawal | Jun 12, 2026 | Technology

India has emerged as the world’s largest market for digital wallet app downloads in CY 2025, according to Sensor Tower’s App Performance Insights: Digital Payments and Mobile Wallets report.

The country accounted for over 440 million digital wallet app downloads during the year, cementing its lead over all other global markets. This scale has also pushed local players such as PhonePe and Paytm onto the global stage, with PhonePe emerging as the most-downloaded digital wallet app globally in 2025. Government-backed BHIM also featured among the top five most downloaded wallet apps in India.

Where does India stand

According to Sensor Tower, digital wallets and peer-to-peer payment apps collectively crossed 1.8 billion downloads globally in CY 2025, growing 3 per cent year-on-year. Of this, India alone contributed over 440 million downloads.

This is not an isolated spike. Sensor Tower data shows that India ranked first in global digital wallet downloads in both 2023 and 2024 as well.

However, downloads declined by 12 per cent in CY 2025 compared to CY 2024. This came after a sharp 14 per cent jump in 2024 over 2023.

The broader APAC (Asia-Pacific) region followed a similar trajectory — strong expansion in 2024 followed by moderation in 2025. The report notes that growth in high-adoption markets is beginning to stabilise after years of rapid onboarding, signalling a shift from user acquisition to engagement and monetisation.

This transition becomes clearer when seen alongside India’s underlying payments infrastructure.

The UPI effect

India’s digital wallet growth is closely tied to the rise of the Unified Payments Interface (UPI), which has become the backbone of the country’s digital payments ecosystem. Unlike many global markets where wallets operate as standalone products, Indian wallet apps are effectively front-ends to UPI, enabling everything from peer-to-peer transfers to merchant payments through a single interoperable rail.

This underlying infrastructure is critical to understanding current trends.

According to the India Digital Wallet Market Report and Forecast 2026–2035 by Expert Market Research, UPI processed 228.5 billion transactions in 2025, marking a 33 per cent year-on-year increase, with total transaction value reaching ₹299.7 trillion.

This is the key context.

Even as app downloads moderate, transaction volumes and value continue to surge, indicating that existing users are transacting more frequently and across more use cases.

Separately, the same report pegs India’s digital wallet market at $20.1 billion in 2025, with projections to reach $75.8 billion by 2035, growing at a CAGR of 14.2 per cent.

From downloads to daily habit

If downloads tell one part of the story, engagement tells another.

Sensor Tower data shows that users in India opened digital wallet apps nearly five times per day on average — significantly higher than global benchmarks. Indonesia ranked second at around 3.75 daily opens, while in markets such as the US, usage was closer to 2.5 times per day.

This points to a structural behavioural shift. Digital payments in India are no longer episodic, they are embedded into daily routines.

The contrast with banking apps is telling. Indian users engage with banking apps just over 2.5 times per day on average, nearly half the frequency of wallet apps.

In markets like the US, the gap between banking and wallet engagement is far narrower. In Indonesia, banking apps even outperform wallets in engagement.

A young user base shaping the market

Demographically, this shift is being driven by younger users.

Around 45 per cent of digital wallet users in India fall in the 25-34 age bracket, while another 25 per cent are aged 18-24, according to Sensor Tower’s audience insights.

This makes India structurally different from markets like Japan or South Korea, where adoption is more evenly distributed across age groups.

In India, digital wallets are being shaped by a younger, mobile-first generation, which is more likely to adopt new financial behaviours, including embedded finance and credit products.

Which apps are leading

Indian digital wallet platform PhonePe was the most downloaded digital wallet app in India during CY 2025, followed by Paytm. Google Pay ranked third, while the government-backed BHIM app came in fifth. Other players such as Super.money and FamApp rounded out the top six.

The dominance of Indian players is not limited to the domestic market. PhonePe emerged as the most downloaded digital wallet app globally in CY 2025, with Paytm also featuring among the top three.

However, downloads only tell part of the story.

The market is far more concentrated when viewed through actual usage. Google Pay and PhonePe together accounted for over 80 per cent of UPI transactions in the first half of CY 2025, according to a Rest of World report.

This concentration creates a high barrier to entry — one that even platforms with massive user bases have struggled to overcome.

Why WhatsApp Pay couldn’t convert scale into usage

On paper, WhatsApp Pay should have been a disruptor.

With over 500 million users in India, it had a distribution network that no fintech player could match. In practice, however, it has barely made a dent in the market.

A mix of regulatory constraints and strategic underinvestment slowed its momentum early. For years, WhatsApp Pay operated under user caps imposed by the National Payments Corporation of India (NPCI), allowing incumbents to consolidate their lead. Even after those restrictions were lifted, adoption remained limited.

Between December 2024 and May 2025, WhatsApp Pay added just over 12 million transactions, while rivals Google Pay and PhonePe added nearly 700 million and 500 million, respectively, according to a Rest of World report.

The reasons extend beyond regulation.

WhatsApp treated payments as an add-on feature rather than a core product, with limited incentives and minimal marketing. In a market where cashbacks, rewards, and ecosystem depth drive engagement, that approach proved insufficient.

More importantly, India’s digital wallet ecosystem has evolved beyond standalone apps into what are effectively financial super apps. Platforms such as PhonePe and Paytm combine payments with credit, insurance, investments, and commerce, making them the default financial interface for users.

Apple Pay’s entry could test a different playbook

Against this backdrop, Apple Pay is expected to enter India by the end of 2026.

The service, which allows users to store cards and make contactless payments via NFC, has seen success in several global markets. However, its prospects in India are shaped by very different structural dynamics.

Apple Pay is also expected to integrate UPI, which could allow it to participate in India’s dominant payments infrastructure. Even so, its reliance on Apple devices means its addressable market remains narrower in a country dominated by Android smartphones.

That said, Apple Pay could still find traction in premium urban segments, particularly as contactless payments and card tokenisation gain ground. Even with expected UPI integration, challenging the dominance of PhonePe, Google Pay, and Paytm — which are deeply embedded in user habits, merchant networks, and everyday transactions — will be significantly more difficult.

Source link

HP EliteBook X G2q review: The strongest case yet for Snapdragon on Windows

by Hansraj Agrawal | Jun 12, 2026 | Technology

I have been using Qualcomm Snapdragon-powered Windows laptops for a while now, especially those running on the first-generation X Elite chips. Those machines were interesting. They got a lot right, especially with efficiency and battery life, but they still felt like they were figuring things out.

Moving to the HP EliteBook X G2q didn’t feel like switching to something entirely new. It felt more like revisiting the same idea, but with a better understanding of what actually needed fixing.

This is also my first time using a laptop powered by the second generation Snapdragon X series chip, more specifically Snapdragon X2 Elite chip, and the difference is not something that shows up immediately. It shows up in how the laptop behaves over a full workday.

Design and build

The EliteBook X G2q doesn’t try to do anything visually interesting. It sticks to the same understated, enterprise-focused design language, which is expected from this series.

What stands out more is the weight. It is light for a 14-inch laptop, almost to the point where it feels slightly unreal at first. That lightness makes it easy to carry around, especially if you’re constantly moving between meetings or workspaces.

The build itself feels solid. There’s no worrying flex, and it doesn’t feel fragile despite how light it is.

The keyboard is one of the highlights. It offers good feedback and enough travel despite sitting flush with the body. Typing feels natural, and more importantly, consistent over long sessions. It’s one of those keyboards you don’t have to adjust to.

There is, however, something slightly odd with the unit I’ve been using. There’s a small bulge on the keyboard deck. It’s not something you notice while typing, but you do feel it when resting your palm on that area. It doesn’t affect usability directly, but it does stand out once you notice it.

The trackpad is massive. On paper, that sounds great, but in practice, it leads to accidental gestures. There were multiple instances where a two-finger scroll turned into a three-finger gesture, or the cursor moved because my palm slightly brushed the surface while typing. It’s usable, but not always precise.

Display and audio

The display on the EliteBook X G2q is one of those parts that doesn’t immediately stand out, but makes more sense the longer you use it.

HP offers multiple panel options here, including high-resolution LCD and OLED variants. The unit I’ve been using comes with a touchscreen LCD panel with a matte finish, and that choice feels very intentional. Matte finish keeps reflections under control, which makes it much easier to use the laptop in bright environments, whether that’s under office lighting or near a window.

Brightness is also sufficient for most scenarios. It doesn’t try to push extreme numbers, but it stays comfortably usable across different lighting conditions. Colours are neutral rather than overly punchy, which works better for long work sessions where accuracy and comfort matter more than vibrancy. What also helps is consistency. There are no sudden shifts in brightness or colour that distract you while working, and that matters more than peak specs in a business laptop.

On the audio front, the built-in speakers are easily one of the more surprising parts of this laptop.

They are top-firing, which already gives them an advantage in terms of how the sound is projected. Instead of being muffled against a surface, the audio comes directly towards you.

Volume levels are more than sufficient. You can comfortably fill a small room without distortion creeping in, and more importantly, the quality holds up even at higher volumes.

What stands out more is the sense of depth. There is a bit of low-end presence here, not enough to replace external speakers, but enough to add some weight to music and video playback.

For a business laptop, this is a surprisingly well-tuned speaker system.

Performance and battery

This is where the laptop starts to separate itself from previous ARM-based machines. At first, the Snapdragon X2 Elite chip doesn’t feel very different. The change becomes noticeable once you start using it for longer stretches. Day-to-day tasks such as browsing, writing, multiple tabs, video calls, are handled without any friction. Even when you push it a bit with heavier multitasking, the system doesn’t slow down unpredictably.

What stands out more is consistency. Some of the earlier Snapdragon laptops could feel fast initially but would dip under sustained workloads. That doesn’t happen here as often.

Thermals are handled in a slightly unusual way. The laptop remains extremely quiet, and you rarely hear the fans ramp up, even under sustained workloads.

But the heat doesn’t disappear, it shifts. Instead of building up at the bottom, it concentrates around the keyboard area. You don’t feel it while typing, but resting your palm on the surface makes the warmth noticeable.

This also raises a question about the small bulge on the keyboard deck. It sits roughly in the same area where the heat builds up, which makes you wonder whether it’s a one-off defect or something caused by prolonged thermal exposure.

Battery life is where the biggest improvement shows up. This is easily a full workday machine now. You don’t think about charging it during the day, which wasn’t always the case with earlier Snapdragon laptops.

The experience

This is also one of the newer Copilot+ PCs, so you get access to Microsoft’s AI features running on the NPU.

Recall is the one that gets the most attention. It creates a searchable timeline of your activity, including apps, files, web pages, and lets you find things later using natural language. It works as intended, but it’s not something that becomes second nature. Most of the time, I still ended up relying on browser history or manual search. It’s useful when you specifically need it, but it doesn’t redefine how you use the laptop.

Click to Do feels more practical. You can interact with on-screen content, text or images, and perform quick actions like summarising or copying. It fits more naturally into day-to-day use, especially when you’re working across multiple documents or tabs.

There are also the usual Copilot+ features like Live Captions, Studio Effects, and image generation tools, all of which run locally using the NPU. These are more situational, but things like background blur, eye contact, and voice focus during calls actually make a noticeable difference in regular usage.

Beyond AI, the laptop comes with several enterprise-grade features. You get hardware-level protections like virtualisation-based threat isolation and tools like HP Wolf Pro Security and Sure Click, which are designed to prevent malicious files or phishing links from affecting the system in the first place.

There’s also a focus on manageability. Features like HP’s Manageability Integration Kit and Workforce Experience Platform are clearly meant for IT teams handling large deployments, not individual users.

Verdict

The HP EliteBook X G2q is clearly built for enterprise users, and it makes more sense when you look at it from that lens. The focus here is on consistency rather than peak performance. The Snapdragon X2 Elite chip delivers a stable, quiet experience, and battery life is strong enough to get through a full workday without concern. That reliability is where this laptop stands out.

There are still a few rough edges. The thermal behaviour is unusual, with heat shifting to the keyboard area, and the oversized trackpad can lead to accidental inputs. That said, for enterprise users who prioritise portability, battery life, and a predictable work machine, this makes a strong case. However, the price is definitely on the higher end.

Price: Rs 254,484 (review unit)

Starting price: Rs 250,000

Source link

AI will not make GBS firms obsolete, but more important: BCG report

by Hansraj Agrawal | Jun 12, 2026 | Technology

Artificial intelligence (AI) will not make Global Business Services (GBS) organisations obsolete; instead, it will make them more strategically important as enterprises deploy AI agents at scale, Boston Consulting Group (BCG) said in a report released on Friday.

The report argued that while AI agents could automate workflows and reduce transactional work by 30 to 50 per cent over the next three to five years, they cannot replace the governance, accountability, data ownership and end-to-end process management required by large enterprises.

Instead, companies should reposition GBS units as centralised “AI Control Towers” responsible for overseeing AI deployment, enterprise data and cross-functional operations, it said.

GBS organisations are centralised units that handle business functions such as finance, human resources, procurement and supply-chain operations for large enterprises, often across multiple geographies.

“The enterprises that win in the AI era will not be those that fragment operations across functions and hope agent swarms self-coordinate,” said Rajiv Gupta, Managing Director and Senior Partner at BCG.

He added: “They will be the ones that build a centralised AI Control Tower with the governance, data, and accountability to deploy AI at scale and sustain it. That is what a transformed GBS delivers.”

Key findings

The BCG report stated that enterprises that transform their GBS operations with AI could shrink their delivery footprint by 25 to 35 per cent while strengthening governance and enterprise-wide AI deployment capabilities.

However, it cautioned against dismantling GBS structures and returning operations to individual business functions. Function-owned AI agents operating without centralised oversight could create fragmented accountability and value leakage, particularly in processes that cut across multiple departments, the report stated.

It further claimed that the long-term cost of dissolving GBS organisations, including restructuring expenses, recurring inefficiencies and eventual re-centralisation, could reach four to seven times annual GBS operating costs over a decade.

“The biggest misconception in boardrooms today is that AI reduces the need for GBS,” said Matt Marchingo, Managing Director and Partner at BCG.

He added: “As agentic AI scales, governance, data integrity and end-to-end accountability matter more, not less.”

Why it matters for India

The report is particularly relevant for India, which hosts one of the world’s largest Global Capability Centre (GCC) and GBS ecosystems.

According to BCG, as agentic AI scales across finance, human resources, procurement and supply-chain functions, GBS organisations in India are well-positioned to evolve from service-delivery centres into enterprise AI orchestration hubs.

The report argued that the challenge for companies is not whether to transform these organisations, but whether to lead that transformation before AI-driven disruption forces it.

Source link

Don't know if Claude AI used in strike at Iran school: Anthropic CEO

by Hansraj Agrawal | Jun 12, 2026 | Technology

By Katrina Manson and Emily Chang

Anthropic PBC’s boss said he doesn’t know what role his artificial intelligence model played in a missile strike that killed an estimated 120 children at an elementary school in Iran, reflecting a broader knowledge gap for AI executives who are increasingly selling advanced AI tools to the US military.

But he said the use case in this instance didn’t violate the company’s policies, arguing military decision makers make terrible mistakes even at the best of times.

“Look, we don’t have access to, we don’t know exactly how these models were used,” Dario Amodei, Anthropic’s CEO and cofounder told Bloomberg’s The Circuit with Emily Chang in an interview, when asked whether his company’s AI tool Claude played a role in the February 28 strike — the first day of US operations in Iran — on the school in Minab. He described the strike against the school as “a really terrible thing to happen.”

“The principle that we have established, and I think the principle that was obeyed here, is a human makes the final decision,” Amodei said. “I don’t know what role Claude or any other AI had, but if this isn’t an illustration why that principle is so important, I don’t know what is.”

The Pentagon, which hasn’t publicly claimed responsibility for the strike, is investigating the incident.

A long-running debate about the role of AI at war pits those who hope AI will reduce mistakes in war, save lives and deliver victories against campaigners and experts who worry that speeding through targets will make war worse and lead to greater civilian harm. There is also debate about the extent to which technology companies should be privy to how their AI tools are used and whether they should be held responsible for outputs that result in mistakes.

Hamza Chaudhry, at the Future of Life Institute, a group that emphasizes the risks of military AI, warns that AI targeting processes could ultimately speed up so fast that nominal human decision-making amounts to little more than a “rubber stamp.” As a result, he said, the expanded scale of combat could result in the taking of many more lives.

Amodei’s comments bolster previous reporting from Bloomberg News that AI frontier companies have little oversight of the powerful, sometimes unreliable and potentially deadly tools they are sending into combat. At the same time, the Pentagon is seeking to accelerate its AI adoption on the battlefield.

The Anthropic CEO triggered a row with the Trump administration earlier this year by drawing a line at using his company’s AI tools in fully autonomous weapons and mass domestic surveillance. As a result, the Pentagon made the unprecedented decision to designate the US company as a supply-chain risk, triggering a lawsuit from privately held Anthropic that is ongoing.

Nevertheless, Amodei’s AI is playing a role in the US war with Iran.

US Central Command is using an AI-assisted platform named Maven Smart System that uses Claude and other AI tools to help generate so-called points of interest, help its personnel make decisions and speed up processes for its military operations against Iran. The command has emphasized the unusually high quantity of targets that US operators have hit in a short period of time: striking 1,000 targets in the first 24 hours of operations and 13,000 targets by April 6, little more than a month after operations began.

Asked whether he was comfortable with Claude’s role in hitting more targets and helping to kill people more quickly, Amodei said he supported the US having the ability to be more effective militarily, arguing that the ability to be “stronger” deters rather than causes wars.

“You basically have to leave policy in the hands of the military decision makers,” he said, noting that a human makes the final decision.

Amodei argued that developing AI warfare could help avoid World War III and help defend Taiwan from invasion from China, so long as the technology wasn’t used without limits or in a way that undermined democratic values.

“When I see Russia invading Ukraine, when I see the risk of China invading Taiwan, it worries me that we have a kind of resurgent authoritarian bloc, that they’re very aggressive and that we need to defend ourselves,” he said.

“We don’t want a world where China and Russia can build, can analyze all the intelligence with AI, can use AI for attacking Taiwan and Ukraine, and we can’t defend them,” he added.

But even some supporters of AI warfare are increasingly wary of the risks involved.

“There’s a lot of governance pieces that, in my opinion, are missing,” Jack Shanahan, retired Air Force lieutenant general and former director of Project Maven, the Pentagon effort that birthed Maven Smart System, told a Stanford University workshop devoted to the future of decision-making last week. He warned that integrating Claude into Maven Smart System could lead to unexpected impacts and dilute the role of human judgment.

While one advantage of AI is to give humans more time to make decisions, Shanahan said he worried that “Type A” personalities in the military may use that extra time to make more decisions rather than better decisions.

“If you make more decisions rather than the right decisions, you may have a very flawed decision-making process,” he said. “You may have a thousand targets, but are they the right targets?”

Source link

« Older Entries

Next Entries »

Times Top News

Times Top News