By Omair Khaliq Sultan You no longer need an engineering degree to print a Baby Yoda. A few years ago, 3D printing was a hobby defined by troubleshooting: leveling beds with sheets of paper, unclogging nozzles, and tightening belts. In 2026, the technology has finally matured into an “appliance” phase. Modern beginner printers now calibrate themselves. They use sensors […] The post Beginner 3D printers: the “it just works” era is finally here appeared first on Digital Trends.

Source:: Digital Trends

By Hisan Kidwai Getting a phone serviced can be a headache, as nobody likes to stand in long queues….
The post OPPO India Launches Service Center 3.0 Pro With Digital Check-In and Faster Repairs appeared first on Fossbytes.

Source:: Fossbytes

By Ana-Maria Stanciuc QT Sense, a deep-tech biotech startup building tools to study living cells, announced it has secured €4 million in funding to accelerate its Quantum Nuova platform, a technology that lets scientists observe cellular processes in real time and reveal biochemical activity linked to disease.  The funding includes a €3 million seed investment led by Cottonwood Technology Fund, with follow-on backing from existing investor QDNL Participations and an angel investor. In addition, the company received €600,000 from the ONCO-Q programme to support cancer research and €400,000 through the Quantum Forward Challenge for collaborative deployments with research partners. Most traditional lab methods…This story continues at The Next Web

Source:: The Next Web

By Ana-Maria Stanciuc  Europe’s largest dedicated deep tech growth fund has taken a major step forward after closing the first tranche of its fundraising effort at €750 million. The fund, known as Kembara Fund I and managed by Spain-based Mundi Ventures, is aiming for a €1 billion target. It will invest in European companies developing breakthrough technologies in areas such as clean energy, AI, quantum computing, advanced materials, robotics, and space tech. A cornerstone of the fundraising so far is a €350 million commitment from the European Investment Fund, part of the EU’s attempt to strengthen local growth capital. Additional backing comes from…This story continues at The Next Web

Source:: The Next Web

By Omair Khaliq Sultan Some workout headphones sound great… right up until they start slipping, trapping sweat, or making you feel sealed off from the world. That’s why bone-conduction headphones have a loyal following. They keep your ears open, stay stable, and feel more “wear and forget” on runs and walks. The Shokz OpenRun bone conduction headphones are $89.94, […] The post If you hate in-ear earbuds while running, these open-ear Shokz are $89.94 right now appeared first on Digital Trends.

Source:: Digital Trends

By Omair Khaliq Sultan Most gaming headsets force you into one lane. They’re either great on PC, annoying on console, or they don’t play nicely with your phone. A tri-mode headset fixes that problem in a very satisfying way. The ASUS ROG Delta II Wireless Gaming Headset is $159.99, down from $259.99, a 38% discount, and it’s listed as […] The post One headset for everything: PC, console, and phone, now $100 off appeared first on Digital Trends.

Source:: Digital Trends

By Ana-Maria Stanciuc If cinema has taught us anything about interacting with our own creations, it’s this: androids chatting among themselves seldom end with humans clapping politely. In 2001: A Space Odyssey, HAL 9000 quietly decides it knows better than the astronauts. In Westworld, lifelike hosts improvise rebellion when their scripts stop making sense. Those stories dramatize a core fear we keep returning to as AI grows more capable: what happens when systems we design start behaving on their own terms? You might have heard the internet is worried about Moltbook, a social network made exclusively for AI agents. It’s an audacious claim:…This story continues at The Next Web

Source:: The Next Web

Apple has embraced agentic AI for developers, introducing direct support in Xcode 26.3 for both Anthropic’s Claude Agent and OpenAI’s Codex and making vibe coding now a platform feature for iPhone, iPad, and Mac. It’s available to all Apple Developer Program members now and will be released “soon” on the App Store.

Xcode 26.3 follows on the heels of the introduction of a macOS app for Codex, but it delivers much more than Codex alone. The new integration opens up new opportunities for developers as Apple’s development environment can now autonomously support their work, from task management to coding to project architecture and more. It represents a major extension beyond the AI features introduced in Xcode 26. 

Apple is also thinking ahead in this support. Xcode 26.3 makes its capabilities available through Model Context Protocol, an open standard that gives developers the flexibility to use any compatible agent or tool with Xcode. This is a big step for Apple, which wants to position Xcode as a companion to the growing flock of AI development tools. 

The result is that developers can select the correct tool for their task, using models most suitable for their work, opening the door to intensified competition between agentic tools.

“At Apple, our goal is to make tools that put industry-leading technologies directly in developers’ hands so they can build the very best apps,” said Susan Prescott, Apple’s vice president of Worldwide Developer Relations. “Agentic coding supercharges productivity and creativity, streamlining the development workflow so developers can focus on innovation.”

In a post on the Anthropic website, that company explained the extent of the integration: “Developers get the full power of Claude Code directly in Xcode — including subagents, background tasks, and plugins — all without leaving the IDE.”

During briefings provided around the introduction, Apple confirmed it worked directly with both OpenAI and Anthropic to optimize the experience of using their models in Xcode. During this collaboration, particular attention was paid to reducing token usage and efficiency. Agents must be downloaded from within Xcode for this integration.

What can agents do in Xcode?

Built-in access to Claude Agent and Codex means developers can exploit the advanced reasoning of these models while building apps. It also means developers can switch between different available models, selecting the most appropriate one for their project, though it will be important to consider the terms of service offered by those models before using them in code.

Developers could use these tools to:

Search documentation.

Execute autonomous tasks.

Explore file structures, understand how the pieces connect, and spot necessary changes before writing code.

Update project settings.

Verify work visually by capturing Xcode Previews and iterating through builds and fixes — even capturing screenshots to show code functions properly.

Developers can also combine all these features, using AI to vibe code apps, build images, develop file structures and verify app behavior, iterating on the app. This lets them focus on improving the overall experience of the code.

Finally, the introduction of Model Context Protocol delivers much more than the press statement explains: as long as the IDE is running, users can browse and search Xcode project structure, read/write/delete files and groups, build projects (including structure and build logs), run fault diagnostics, execute tasks and more, using their choice of MCP-supporting agent models.  

What comes next?

There are some risks coming into view. Vibe coding at scale will happen, and when it does it will introduce a flotilla of rapidly-created apps, some of which might include security flaws if not verified and checked correctly. That’s even before you consider the tendency of large language models (LLMs) to hallucinate.

There is also the danger that the novelty and power of these applications might distract some developers who would traditionally put their energy into open-source projects, potentially undermining the integrity of those important projects. Stack Overflow use has collapsed as developers use chatbots instead of the knowledge bases the AI has already digested. 

Apple’s Xcode decision makes it inevitable that code running on the devices you own or that your company deploys will be partly built by AI. It’s an open question whether students learning to code today can reliably anticipate opportunities to make a living doing so in a decade’s time.

You can follow me on social media! Join me on BlueSky,  LinkedIn, and Mastodon.

Source:: Computer World

The tech bigwigs and economists at the recent World Economic Forum (WEF) in Davos were clear-eyed about how AI is reshuffling the jobs landscape globally and disrupting national economies.

But the chatter around physical AI and robotics was more upbeat, with attendees saying robots with brains and intelligent sensors are likely to improve human productivity and manufacturing output. 

That, in turn, should improve economies — and in the long run create more jobs. 

Physical AI refers to the concept of AI manifesting in physical form, most notably as robots, though others foresee broader real-world outcomes, such as AI cameras that reduce crime or AI-driven sensors that bolster industrial output.

“You can now fuse your industrial capability, your manufacturing capability with artificial intelligence and that brings you into the world of physical AI or robotics,” said Jensen Huang, CEO of Nvidia, during a fireside chat at WEF.

If anything, AI — including agentic AI and robotic automation — is more likely to  change the nature of what humans do than take jobs away, he said. Robots can do menial work such as typical administrative tasks, allowing humans to be more productive, Huang said. “We’re five million nurses short…. AI is increasing their productivity…. [And when], hospitals do better, they hire more nurses.”

Robots have the capacity to work non-stop, without tiring, yielding productivity gains that will increase the average output of economies, said tech entrepreneur Elon Musk. “My prediction is…we’ll make so many robots and AI that they will saturate all human needs,” the Tesla CEO said during a WEF discussion.  

He described how robots might be able to help care for elderly parents, which can be an expensive undertaking for humans, saying the robots can take the place of younger people.

Robots are also capable of doing jobs considered dangerous for humans, though  humans will still need to be involved, Daniela Rus, director at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), said during a panel session at WEF. (Many companies have roots at CSAIL, including Venti Technologies, where Rus is a board member and adviser.)

Venti has “entire fleets of robots that operate 24/7 without the need of human drivers,” Rus said. “Yet human drivers are also in the loop to step in when the weather is bad or when there is a lot of need for movement.”

China is considered further ahead in robot adoption than the US, where the robotics market is still growing. Tianlan Shao, CEO of China-based Mech-Mind Robotics, said his company had delivered more than 10,000 robots in the past year — the same number it produced during its first eight years of existence.

He argued that if a industrial robot is given a chainsaw, for example, humans might still be needed to make sure the robot sticks to task. “We need clear boundaries…, definitions, and rules,” Shao said.

Shao pointed to progress in the last 12 months of fusing AI into robots. “Now we can train this so-called world-model-like thing, aligning everything, including robot vision and robot motion, aligning everything in one specific space.” 

World models under development by the likes of Nvidia, Microsoft ,and Google are designed to improve the physical functionality of robots.  Researchers at Mohamed bin Zayed University of Artificial Intelligence have created a world model called PAN that tests action sequences in a safe, controlled simulation.

The growing importance of physical AI has been acknowledged beyond just the WEF.  Deloitte’s State of AI in the Enterprise report, released in January, pointed to a widening adoption of physical AI. About 58% of companies surveyed by the consulting firm noted physical AI adoption, a figure expected to grow to 80% in the next two years.

The consulting firm sees physical AI as having a real-world component, meaning  technology that can sense and drive a physical outcome. Monitoring and security systems, for instance, are fast-growing areas for deployment, along with collaborative robotics.

“Physical AI as a terminology is relatively new, but the underlying foundation was laid 12 or 13 years ago — what’s different now is adding intelligence and autonomy on top of that physical foundation,” Beena Ammanath, global head at Deloitte AI Institute, told Computerworld.

The older underlying foundations of physical AI include IoT and robotic process automation, Ammanath said.

Despite the encouraging comments about physical AI, not everyone at WEF had rosy assessments for robotics. “Elon Musk also told us in 2017 that we will be falling asleep at the wheel in 2019,” Rus said, offering a cautionary note. “And we’re still not falling asleep at the wheel.”

The rise of humanoid robots will take time due to navigation, materials, dexterity and reasoning issues, Rus said. “It’s coming, but it’s not today,” he said.

WEF’s Chief Economists Outlook noted the growing value of companies that focus on humanoid robots — the mechanized machines that look like and function like humans and are a mainstay of science fiction.

“While far from general-purpose deployment on factory floors, humanoid robotics companies are attracting large valuations and investments,” the WEF said in its outlook for 2026. 

Source:: Computer World

By Ana-Maria Stanciuc Snowflake and OpenAI have struck a multi-year, $200 million partnership to bring OpenAI’s advanced models, including GPT-5.2, directly into Snowflake’s enterprise data platform. The collaboration is designed to let Snowflake’s large customer base, more than 12,000 organisations, build AI agents and semantic analytics tools that operate on their own data without moving it outside Snowflake’s governed environment. Under the agreement, OpenAI models will be natively embedded in Snowflake Cortex AI and Snowflake Intelligence, making it possible to run queries, derive insights, and deploy AI-powered workflows using natural language interfaces and context-aware agents. Customers can analyse structured and unstructured data, automate…This story continues at The Next Web

Source:: The Next Web

By Shikhar Mehrotra OpenAI acknowledges major outage as thousands report ChatGPT errors. Issues impact logins, voice mode, and response generation. A fix is currently in progress.
The post ChatGPT is down for a ton of users (Update: It’s back online) appeared first on Digital Trends.

Source:: Digital Trends

By Omair Khaliq Sultan There are headphone deals that are “nice,” and then there are deals that land in the zone where you stop browsing and actually buy. Beats Studio Pro at $169.95 is that second kind. It’s down from $349.99, a 51% discount, and it checks the boxes most people care about for everyday listening: long battery life, […] The post Beats Studio Pro at $169.95 is the kind of headphone deal you don’t overthink appeared first on Digital Trends.

Source:: Digital Trends

In December, Mozilla CEO Anthony Enzor-DeMeo attracted a lot of attention by announcing that Firefox would become a “modern AI browser.”

In order not to alienate users, the company also promised a new setting that would make it possible to turn off some or all of the AI features, including the chatbot in the sidebar, automatic translations, tab grouping, and link previews.

It is now clear that the new AI settings will be added to Firefox 148, a version that will be rolled out to the public on Feb. 24. Mozilla unveiled the feature on Monday.

The YouTube clip below shows how it is supposed to work:

Source:: Computer World

Enrique Lores, president and global CEO of HP for more than six years, is leaving the company to take up a similar position at online payment giant PayPal on March 1. In his place, on an interim basis for the time being, Bruce Broussard, a member of the company’s board of directors since 2021, has already been appointed CEO of the technology company, although the company said in a statement that it is already looking for a permanent replacement for the Spanish executive.

“As Interim CEO, Mr. Broussard will advance the company’s strategic priorities by leveraging his proven operational, financial, and business management expertise as well as his deep knowledge of HP’s business,” the statement said, noting that Broussard is an executive with more than 30 years of experience in leadership positions at publicly traded companies, such as healthcare company Humana.

Lores is no stranger to PayPal, having served on its board of directors for nearly five years and as its chairman since July 2024. The Spaniard replaces former CEO Alex Chriss, who leaves the company in a delicate situation; in fact, the group has just announced a 15% drop in revenue in its last fiscal quarter. Furthermore, Lores’ appointment, as reported in a statement issued today by PayPal, “follows a detailed evaluation conducted by the Board of Directors on the current position of the company relative to its competition and the broader industry landscape. While some progress has been made in a number of areas over the last two years, the pace of change and execution was not in line with the Board’s expectations. The Board is confident that the appointment of Lores, a seasoned executive with more than three decades of technology and commercial experience, will provide the leadership necessary to lead PayPal into its next chapter.”

American dream

Enrique Lores is one of those paradigmatic cases of the long-awaited ‘American dream’ that argues that anyone can achieve success through hard work. Lores, born in Madrid and an electrical engineer from the Polytechnic University of Valencia, from which he was awarded an honorary doctorate in 2024 for his professional career, began his professional activity at HP in 1989, 36 years ago. He started as an intern, but gradually rose to positions of importance in the fields of printing, personal systems, and business and industrial solutions.

Even in 2015, when the historic company made the decision to split into two companies, one focused on personal devices (PCs) and printing and the other, called HPE, on systems and infrastructure for businesses, it was Lores himself who led the Separation Management Office. More than six years ago, he became the president and CEO of HP. Under his leadership, he had to deal with issues such as an attempted takeover by competitor Xerox, which ultimately did not go through. In recent years, the executive has worked to adapt the PC world to advances in artificial intelligence. According to Gartner data for 2025, HP is the second-largest player in the global PC market, with a 21.5% share, surpassed only by Lenovo (27.2%) and followed by Dell Technologies (16.5%). In the field of printing, according to data from IDC, Gartner, and Canalys, the company is number one in the world.

Lores, one of the highest paid in the entire technology industry, announced his new professional direction today on his LinkedIn account. “I first joined HP 36 years ago as an intern engineer. Since then, HP has become part of my identity and my family’s history: my wife Rocío and I built our life in Palo Alto so that I could be part of the HP team, and my three children have only known life with HP,” he writes on the social platform, where he summarizes his professional career.

HP, he notes, has given him “the opportunity to grow tremendously.” A company, he emphasizes, that he defines as “its people.” “HP is a true school of talent, guided by a culture of innovation, collaboration, and shared dedication to making a positive impact. I am incredibly proud of what the HP team has achieved, and I have every confidence that Bruce Broussard and the incredible HP leadership team will propel the company forward and lead the future of work.”

He says he is looking forward to the “unique opportunity to serve as CEO of PayPal and have a lasting impact on the global payments industry.” “I am excited to get started, knowing that I am leaving behind a team that will drive HP’s success.”

Along with the changes at the top, the HP team has again reported its forecasts for the first quarter of its fiscal year and its full fiscal year 2026. The company expects diluted earnings per share under generally accepted accounting principles (GAAP) of between $0.58 and $0.66, and non-GAAP diluted earnings per share of between $0.73 and $0.81. For fiscal year 2026, HP continues to expect GAAP diluted net earnings per share of $2.47 to $2.77 and non-GAAP diluted net earnings per share of $2.90 to $3.20. For fiscal year 2026, HP expects to generate free cash flow of between $2.8 billion and $3 billion.

Source:: Computer World

By Ana-Maria Stanciuc If you look at the press releases and breathless commentary around the recent acquisition of xAI by SpaceX, you might think we’re witnessing a tectonic shift in technological destiny.  A $1.25 trillion “mega-company” is born, poised to reshape artificial intelligence, space infrastructure, satellite internet, and possibly the fate of humanity itself. That narrative, enthusiastically repeated across headlines, serves a purpose: it frames a somewhat messy corporate consolidation as inevitable progress.  But let’s take a closer look and separate actual substance from Silicon Valley myth-making. A mega-deal that’s really an identity crisis At its core, this acquisition solves one problem: xAI needed…This story continues at The Next WebOr just read more coverage about: SpaceX

Source:: The Next Web

By Ana-Maria Stanciuc OpenAI has given software developers a new desktop toy, and judging by the early reactions, it might feel like someone finally handed coders the Swiss Army knife they’ve been dreaming about or the kind of gadget that makes them wonder if they’re working with a robot coworker now.  The company rolled out the Codex app for macOS, a focused interface for managing AI coding agents, designed to let developers do more than just “generate a few lines of code.” Instead, Codex can juggle multiple tasks in parallel, run background workflows, and act on instructions that span hours or even days. …This story continues at The Next Web

Source:: The Next Web

By Shikhar Mehrotra Alibaba, Tencent, and Baidu are handing out billions in cash and prizes to attract chatbot users, highlighting how fierce China’s AI race has become in 2026.
The post The AI Race is so hot that chatbot users are getting paid in real cash, iPhones, and TVs appeared first on Digital Trends.

Source:: Digital Trends

How much would you pay each month for a Mac, iPhone, iPad, Apple home accessories and a handful of Apple services, including health and home security services? More to the point, how many of Apple’s 2.5 billion users would be willing to pay for the Apple Plus Services suite, and how much would this generate each month in high-yield, high-margin predictable income for the former hardware company?

Apple-as-a-Service? It’s possible

The Apple-as-a-Service idea has hovered at the edge of Apple speculation for years, and while Apple has skirted with the concept (iPhone Upgrade Program), it’s never quite found a way to combine hardware and software in a services-led bundle. But things can change. Bloomberg’s Mark Gurman last year suggested Apple had considered offering a hardware subscription service, but shelved the idea fearing the impact on “normal” hardware sales.

What’s certain at this stage is that it feels as if it would be easier than ever for Apple to pivot its entire business into services, particularly as it has more than 1 billion people using at least one of its services already. Breaking those use patterns down, Apple recently shared some glimpses into the performance of the services segment:

900 million active iCloud+ subscribers.

850 million active App Store users.

58 million Apple TV+ subscribers.

15 million merchants who accept Apple Pay.

Apple News is the number one news app in the US, Canada, and Australia.

The company followed this news up with the introduction of the phenomenal value Creator Studio suite, which unleashes industry leading tools for audio and video production, Final Cut Pro and Logic — along with apps for imagery and productivity — for $12.99 a month. Apps that create opportunity for creative expression are arguably fundamental to the Mac, a purpose that runs deep into Apple’s DNA. 

When it decided to introduce its leading creative apps as a service, Apple moved the needle on both software and services sales, taking a bite out of one segment to benefit a second. The way Apple sees it is that maybe it doesn’t really matter how Apple sells something, so long as the market takes it, and keeps that high margin (c.70%) services segment revenue rolling in. (The segment was worth $30 billion in the just gone quarter.)

The myth of ownership

One argument against hardware-as-a-service is that people like to own their stuff. That’s true. But when you stop to think about it, the truth is that we already use hardware we don’t really own — that iPhone in your pocket you acquired with a carrier deal; the credit card debt you’re paying for your current Mac; the iPads for your field services team you acquired with help from your bridging loan; the Mac you’re about to purchase for a monthly cost direct from the Apple Store.

Even four years ago, “Almost half iPhone owners already finance their iPhone purchase, paying monthly for a new phone,” said CIRP Partner and Co-Founder Josh Lowitz. “And about one-third trade-in their old phone when they buy a new one. So, a significant portion of the user base is accustomed to never owning a phone, instead basically leasing it.”

We use kit we don’t quite own all the time. An Apple hardware-as-a-service offering could accommodate that, offering hardware ownership as one result, after an agreed upon number of payments are made. The second result would be what tempts users though.

Think about how attractive a new Mac upgrade every two or three years and an iPhone upgrade every other year as a basic subscription might be, with AppleCare rolled in. Reflecting that kind of thinking, many already use Apple Trade in, sending their old hardware back to Apple for parts in exchange for money off the next bit of equipment. It’s an ownership model that shows many are already accustomed to treating hardware as if it were a subscription item. 

Improving the experience

But why would Apple abandon hardware sales in favor of services? To build margins, of course, and to develop a predictable income stream, but also to better control the user experience. Not only can Apple then continue to focus on improving its hardware, but because it knows what hardware its customers are running, it can focus improvements in software and services on the hardware.

Think of it as an extension of Apple’s “whole widget” approach, one encompassing hardware, software, services, and silicon with management of the ecosystem itself. 

There are solid environmental reasons for this ecosystem, as it becomes much easier to call in old kit and arrange for it to be recycled, feeding those cannibalized components straight into the circular manufacturing system we know Apple wants to build. It gets a lot easier to build a system like that once you know almost exactly how many recycled components you’ll have to work with in any week; that’s the kind of granular insight a subscription hardware offering might provide.

The numbers game

Would consumers be willing to accept an ownership model like this? To some extent that doesn’t matter. It doesn’t need to be adopted by every consumer to succeed. With 2.5 billion people already one email away from Cupertino, Apple only needs to switch a small fraction of that group over to a hardware/software/services subscription model to generate predictable revenue.

As a thought experiment, imagine if just 1% of those Apple users (25 million) stepped into a services/Mac/iPhone deal at $129/month, that’s $3.2 billion in revenue every month. While I think the cost would be higher (because it’s Apple), that gives you some sense of the revenue the company can raise with the addition of a subscription option when ordering products at the Apple Store.

As a numbers game, Apple starts with the kind of advantage that should make Wall Street nod, in the sense that investors already value tech firms with high recurring revenue more highly than those dependent on cyclical hardware sales.

Who knows, if you could get hold of a Mac and iPhone in such a way for a monthly fee, Apple might find it even easier to upsell consumers to higher specified options through its recently updated Apple Online Store, generating a few more dimes each month with every consensual memory upgrade.

Plus, of course, with user satisfaction numbers in the high 90s, Apple would likely find it even easier to lock its customers into the total Apple experience. That potentially includes a highly secure Apple HomeKit AI-augmented and private home security service with which to protect all this kit, which some believe the company might announce during the year.

You can follow me on social media! Join me on BlueSky,  LinkedIn, and Mastodon.

Source:: Computer World

By Shikhar Mehrotra Sony’s next flagship true wireless earbuds may deliver smarter AI audio, improved noise cancellation, fresh design tweaks, and premium pricing, according to a detailed new leak.
The post Sony leak shows a new look for upcoming WF-1000XM6 earbuds and a big audio upgrade appeared first on Digital Trends.

Source:: Digital Trends

Across industries, CIOs are rolling out generative AI through SaaS platforms, embedded copilots, and third-party tools at a speed that traditional governance frameworks were never designed to handle. AI now influences customer interactions, hiring decisions, financial analysis, software development, and knowledge work — often without being formally deployed in the classical sense.

The result is a widening gap between rapid AI deployment and responsible-use protections. Organizations adopt AI faster than they can govern its usage, then scramble to retrofit controls after something goes wrong.

Interviews with five practitioners — each working at a different pressure point of enterprise AI — reveal why this gap persists and what leaders must do to close it before regulators, auditors, or customers force the issue.

Why governance breaks the moment AI hits real workflows

The first problem is structural. Governance was designed for centralized, slow-moving decisions. AI adoption is neither. Ericka Watson, CEO of consultancy Data Strategy Advisors and former chief privacy officer at Regeneron Pharmaceuticals, sees the same pattern across industries.

“Companies still design governance as if decisions moved slowly and centrally,” she said. “But that’s not how AI is being adopted. Businesses are making decisions daily — using vendors, copilots, embedded AI features — while governance assumes someone will stop, fill out a form, and wait for approval.”

That mismatch guarantees bypass. Even teams with good intentions route around governance because it doesn’t appear where work actually happens. AI features go live before anyone assesses training data rights, downstream sharing, or accountability.

What breaks first, Watson said, is data control and visibility. Employees paste sensitive information into public genAI tools, and data lineage disappears as outputs move across systems. “By the time leadership realizes what’s happening,” she said, “the data may already be gone in ways you can’t undo.”

What to do: CIOs must move from model governance to usage governance. You may not control the model, but you can control how it’s used, what data it touches, and where outputs flow. Governance has to be embedded as tollgates inside workflows, not in policy documents that are reviewed after the fact.

Why legacy data governance struggles under genAI

Even where governance exists, it’s often built on assumptions that no longer hold. Fawad Butt, CEO of agentic healthcare platform maker Penguin Ai and former chief data officer at UnitedHealth Group and Kaiser Permanente, argues that traditional data governance models are structurally unfit for generative AI.

“Classic governance was built for systems of record and known analytics pipelines,” he said. “That world is gone. Now you have systems creating systems — new data, new outputs, and much is done on the fly.” In that environment, point-in-time audits create false confidence. Output-focused controls miss where the real risk lives.

“No breach is required for harm to occur — secure systems can still hallucinate, discriminate, or drift,” Butt said, emphasizing that inputs, not outputs, are now the most neglected risk surface. This includes prompts, retrieval sources, context, and any tools AI agents can dynamically access.

What to do: Before writing policy, establish guardrails. Define no-go use cases. Constrain high-risk inputs. Limit tool access for agents. And observe how systems behave in practice. Policy should come after experimentation, not before. Otherwise, organizations hard-code assumptions that are already wrong.

Why vendor AI is where governance collapses

If internal AI governance is weak, third-party AI governance is worse. Richa Kaul, CEO of Complyance, works with global enterprises on risk and compliance management. She sees a sharp divide: while companies are relatively mature in governing AI they build themselves, they are much less prepared when AI arrives embedded in vendor products.

“What we’re seeing is use before governance,” she said. “And it’s often governance by committee — 10 to 20 people reviewing vendors one by one without a shared baseline of questions.” Too often, enterprises ask open-ended questions about AI privacy and accept reassuring answers — what Kaul calls “happy ears.”

Mature governance shows up in specific questions. Is customer data used to train models? Is it reused across clients? Is the LLM accessed via an enterprise deployment or a consumer interface?

“A vendor using Azure OpenAI has a much lower risk profile than one calling ChatGPT directly,” Kaul said.

What to do: CIOs should start with a basic but overlooked step: scrutinize vendor subprocessor lists. Cloud providers are well understood. LLM providers are not. AI has created a second, poorly mapped subprocessor layer — and that’s where governance breaks down.

Why bans fail and incidents repeat

Technology controls alone do not close the responsible-AI gap. Behavior matters more. Asha Palmer, SVP of Compliance Solutions at Skillsoft and a former US federal prosecutor, is often called in after AI incidents. She says the first uncomfortable truth leaders confront is that the outcome was predictable.

“We knew this could happen,” she said. “The real question is: why didn’t we equip people to deal with it before it did?” Pressure to perform is the root cause. Employees use AI to move faster and meet targets — just as they have in every compliance failure from bribery to data misuse.

Blanket bans on genAI do not work. “If you take away responsible use,” Palmer said, “people will use it irresponsibly — in secret, in ways you can’t govern.”

What to do: Shift from awareness training to behavioral learning. Palmer calls it “moral muscle memory,” a scenario-based practice that teaches people to stop, assess risk, and choose a better action under pressure.

Regulators and auditors look for evidence that the right people have received the right training for the risks they actually face. One-size-fits-all AI literacy is a red flag.

Why confidence is not enough when auditors arrive

The final gap appears when organizations are asked to prove their governance works. Danny Manimbo is ISO & AI Practice Leader at Schellman, an attestation and compliance services provider. He sees the same failure pattern repeatedly.

“Organizations confuse having policies with having governance,” he said. “Responsible AI principles don’t matter if they don’t influence real decisions.”

Auditors might start with a simple request: show us a documented AI risk-based decision that changed an outcome. Mature governance leaves fingerprints — including delayed deployments, rejected vendors, and constrained features. Immature governance produces vague assurances.

“The most expensive governance work is the work you try to do after deployment,” Manimbo warned. Walking back data lineage, accountability, and intended purpose is extraordinarily difficult once systems are live.

What to do: Treat AI governance as a management system, not a compliance exercise. Standards like ISO/IEC 42001 work only when they connect risk management, change control, monitoring, and internal audit into a continuous loop.

You can tell governance is working when it changes business decisions, not when it produces documentation.

Closing the responsible AI gap

Across all five interviews, one theme recurs: the responsible AI gap is not primarily a technology failure. It’s a governance timing failure. Controls are being designed for yesterday’s systems while AI is already shaping today’s decisions.

Several of the sources stressed that CIOs should stop framing responsible AI as a future-state program and start treating it as an operational hygiene issue — closer to identity management or financial controls than to ethics committees.

Watson from Data Strategy Advisors emphasized that visibility is the first non-negotiable step. Enterprises that cannot enumerate where AI influences decisions — especially through SaaS tools — are already exposed. “You can’t govern what you can’t see,” she noted, warning that many companies still lack even a basic inventory of AI-affected workflows.

At Penguin Ai, Butt reinforced that point from a data perspective, arguing that inventories must shift from platforms to systems-in-context. An AI feature embedded in HR software and the same feature embedded in marketing automation do not carry the same risk. Treating them as identical is a governance illusion.

Complyance’s Kaul added that the same principle applies externally. Vendor AI governance breaks down when enterprises accept generic assurances instead of mapping where their data actually flows. In her experience, simply forcing teams to trace AI subprocessors exposes risks that executives did not realize they had accepted.

To close the gap, CIOs must:
>
Embed governance where work happens — and before it happens, not after.
Shift focus from models to usage and inputs.
Treat vendor AI as a first-class risk domain.
Replace bans with behavioral training.
Demand governance that provides explanations as to how decisions are made.

The organizations that do these things will not only avoid regulatory trouble, they will move faster — and with more confidence.

Palmer from Skillsoft focused on the human layer that sits underneath all of this. Governance frameworks collapse, she argued, when they assume people will slow down under pressure. “Pressure doesn’t disappear,” she said. “You have to train for it.” Organizations that fail to do so should not be surprised when employees improvise with AI in unsafe ways.

Finally, Schellman’s Manimbo offered a blunt litmus test: if governance has never delayed a deployment, rejected a vendor, or constrained a feature, it probably does not exist in practice. “Governance has to leave fingerprints,” he said. Otherwise, it is indistinguishable from aspiration.

Taken together, the interviews suggest that closing the responsible AI gap does not require perfect foresight or exhaustive policy. It requires earlier intervention and clearer accountability. Organizations that act now — while AI use is still fragmented and informal — have a chance to shape behavior. Those that wait will inherit systems they no longer control and risks they can no longer explain.

At that point, governance is no longer a choice. It becomes damage control.

Related reading:

GenAI in productivity apps: What could possibly go wrong?

Overcome governance and trust issues to drive agentic AI

The trick to balancing governance with innovation in the age of AI

Deloitte’s AI governance failure exposes critical gap in enterprise quality controls

Source:: Computer World