Technology News

Making technology work for business
  1. Printers and scanners are increasingly becoming ways for cyber crooks to deliver phishing attacks, thanks to a flaw in the Microsoft 365 Direct Send feature.

    The Varonis forensics team has uncovered an exploit which allows internal devices such as printers to send emails without authentication. The vulnerability has been used to target more than 70 organizations, predominantly in the US, with threat actors spoofing internal users and delivering phishing emails without needing to compromise any accounts whatsoever.

    The campaign has been successful because emails sent from within Microsoft 365 (M365) undergo less scrutiny than standard inbound email.

  2. There may be trouble in the industry’s biggest AI alliance, with a contract dispute about AGI threatening to topple the partnership between OpenAI and Microsoft.

    The dispute, according to a report from The Information, involves a line in their contract that allows for the alliance to die once AGI (artificial general intelligence), the ability for genAI to replicate the capabilities of human thought, is achieved.

    The problem is not with the lack of a precise AGI definition. It is that the concept is impossible to prove. OpenAI could deliver 100 proofs that show that they have achieved AGI and Microsoft could counter with 100 proofs that it hasn’t. Sentience is impossible to prove — or to disprove — with examples.

    “They’re never going to settle on a definition of AGI that is intuitively satisfying to all. Any attempt to define AGI by looking at the internals of how the mind works often gets muddied by things like qualia and consciousness, which are notoriously difficult to pin down using an externally verifiable measure,” said John Licato, an associate professor at the University of South Florida’s Bellini College of Artificial Intelligence. “Instead, I expect they’re going to need to pick a somewhat arbitrary dividing line, purely tied to tests of performance. One example might be based on some consensus-based variant of the Turing Test in which a group of laypeople are asked to interact [blindly] with either humans or the AI, and they are then asked which they interacted with. If a large enough percentage of people are fooled, then the test is passed.”

    Both sides want to exit the deal

    Setting aside any AGI test, what is likely behind the argument is the desire by both parties to end the agreement, given how much has changed since the deal was struck in 2019.

    The contractual AGI trigger appears to end any additional code-sharing, but there are no indications that Microsoft would have to surrender, or even stop using OpenAI code it had received before AGI was declared. 

    And analysts and enterprise IT executives agree that Microsoft is well-positioned to aggressively continue its genAI efforts without continuing to receive code from OpenAI.

    Microsoft no longer needs OpenAI

    Justin St-Maurice, a technical counselor at Info-Tech Research Group, said he doubts that ending the partnership would set Microsoft back in any serious way.

    “Microsoft has its own models, a strong Azure ecosystem, and access to increasingly capable open-source LLMs. They don’t actually need OpenAI to deliver a successful product. Right now, the real bottleneck with Copilot isn’t the underlying model, but the rigid, rules-based implementation layered on top,” St-Maurice said. “Swapping out OpenAI with a different LLM won’t break an already weak user experience …. As LLMs are becoming commoditized, the magic lies in the integration, not the engine.”

    St-Maurice had a strong reaction to one reference in The Information‘s story about the stated reason for the AGI deal killer clause during the original contract negotiations.

    It said: “The idea behind the AGI contract provision is that Microsoft, as one of the world’s most powerful for-profit firms, shouldn’t get access to technology that might eventually help people colonize other planets or develop nuclear fusion. Doing so would go against OpenAI’s founding principle in 2015 to develop technology for the benefit of all of humanity, an idea that has roots in the beliefs of OpenAI founders, including CEO Sam Altman and Tesla CEO Elon Musk, who wanted to ensure the most powerful technologies didn’t end up in the hands of for-profit firms.”

    OpenAI’s principles have changed since 2019

    St-Maurice said that claim is rather absurd, given how OpenAI has not been acting at all like a non-profit.

    “Sorry, how exactly is OpenAI sticking to its founding principles in 2025?” St-Maurice asked. “It’s hard not to wonder if OpenAI doesn’t want the world’s most powerful for-profit firms to have AGI technology because it would rather be the world’s most powerful for-profit firm with AGI technology.”

    St-Maurice added: “They’re not doing this for humanity, and OpenAI has a lot of work to do to convince me otherwise. The rhetoric about ‘benefiting all of humanity’ rings a little hollow when Sam Altman is openly forecasting mass job displacement and a future where society becomes dependent on the technology class. It’s hard to see altruism when it also appears to conveniently consolidate control.”

    Execs see little impact

    Enterprise IT executives appeared to agree that even if the partnership dissolves, they will feel little to no impact.

    Vinod Goje, the VP engineering manager at Bank of America, said Microsoft is well-positioned for a post-OpenAI model future. It has the cloud infrastructure, the enterprise relationships, and the financial firepower to pivot faster than most realize, he pointed out, stressing that he was speaking personally and not representing his employer. While losing exclusive access to OpenAI’s latest models would sting, they’ve got partnerships with Meta, their own research teams, and enough resources to acquire or develop alternatives.

    “The real disruption isn’t whether Microsoft can survive without OpenAI. It’s that we’re essentially flying blind on the most consequential technology decisions of our lifetime,” he noted. “When you can’t even agree on what AGI looks like, how do you write contracts? How do you regulate it? How do you prevent it from being controlled by whoever gets there first? It’s a preview of the governance chaos coming if we don’t get serious about how AGI is defined, verified and shared. What this really reveals is the structural weakness in how we govern foundational technologies.”

    Goje argued that the MS-OpenAI situation shows how unprepared the industry is for the implications of GenAI.

    “The Microsoft-OpenAI standoff is a canary in the coal mine for the entire AI industry. We’re watching a $10 billion partnership potentially unravel over something as fundamental as ‘What is intelligence?’” Goje said. “This dispute is forcing the industry to confront an uncomfortable truth: we’re building the future without a roadmap. The companies that figure out governance frameworks first, not just the technology, will be the ones that actually shape what comes next.”

    Another enterprise IT executive agreed.

    “I don’t think there will be any material impact [if Microsoft and OpenAI split],” said Brian Phillips, VP of Macy’s technology, “especially if they can keep the code they have been using.”

  3. OpenAI’s planned productivity suite could dismantle traditional habits of how users create and consume documents in the same the way the company changed browsing and search habits.

    “OpenAI is increasingly seeing itself as a productivity tool, and that would include the need to address actual creation tools like Office does,” said Jack Gold, principal analyst at J. Gold Associates.

    OpenAI hasn’t officially announced a product, but The Information reported (subscription required) that the generative AI (genAI) company has already designed a rival to the dominant productivity tools.

    But good luck getting customers to move from Microsoft 365 or Google Workspace, analysts said, noting that the top two productivity suites are well entrenched among users and organizations.

    OpenAI is already including certain elements of a productivity suite in its offerings, such as multiple export format support, said Wayne Kurtzman, research vice president of collaboration and communities at IDC. The feature is available in ChatGPT features such as Canvas, which “is a new interface for working with ChatGPT on writing and coding projects that require editing and revisions,” according to OpenAI’s website.

    “That can be construed, correctly or not, as starting to build a productivity suite,” Kurtzman said.

    IDC sees the market favoring newer digital experiences in creating and consuming content, he said. “Whether OpenAI sees this as an opportunity they would like to pursue in new ways is yet to be seen,” Kurtzman said.

    The future of productivity and collaboration suites lies in user interface simplification via genAI, said J.P. Gownder, vice president and principal analyst on Forrester’s Future of Work team. He described it as “a lot less pulling down menus or drawing and a lot more prompt engineering and providing sources to the AI so it can compose the asset.”

    Document creation could look something like this: GenAI would take a first swing at creating a business document that the user then edits, iterates, and finalizes. That approach will become much more common.

    Users will go “over the top,” asking Microsoft’s Copilot to create PowerPoint presentations, specifying the documents such as meeting notes or oral instructions that it should use to create the deck.

    “I predict that, by 2029, Microsoft PowerPoint will hide or remove 80% of the elements on the Ribbon, the set of navigation controllers. Why? Because you won’t need them anymore; you will go ‘over the top,'” Gownder said.

    OpenAI trying to innovate in this area makes sense; companies like Zoom and beautiful.ai already do this, though not to the level of sophistication users will see in the future with Microsoft’s suite, Gownder said. “…Entering this space, for OpenAI, is a lot riskier, because of its partial ownership by Microsoft and because Copilot uses OpenAI’s models,” he said.

    Microsoft is already heading in the direction of making Copilot its main interface to create documents, spreadsheets and presentations, the company’s chief product officer of experiences and devices, Aparna Chennapragada, told Computerworld in a recent interview.

    Google has already integrated genAI capabilities into Workspace, but hasn’t managed to capture much market share from Microsoft, Gold said. “But like so many other companies have found when they try to compete with Office, it’s very hard to have much impact,” he said.

    Gold floated the idea of OpenAI possibly leveraging open-source tools such as OpenOffice or LibreOffice, which could help from a time-to-market and cost perspective. “Let the open ecosystem provide the necessary capabilities, which already results in a pretty rich productivity suite, and just have OpenAI do the integration of AI tools,” he said.

    There remain a lot of open questions about OpenAI’s ability to deliver a productivity suite, which isn’t easy, said Jeff Kagan, an independent analyst.

    OpenAI needs the talent, product groups, and market share to carve out a sizable niche, Kagan said. “I don’t expect Microsoft to sit back. I expect they will quickly intensify their offerings to hang onto their market share,” he said.

    Also, if OpenAI CEO Sam Altman decides to implement competing features, he will need to think hard about the relationship with Microsoft CEO Satya Nadella.

    “It’s still way too early to have any idea what the next step will be. Stay tuned,” Kagan said.

  4. A frisson of Trump-related news fizzled out in the last week. No, not a temporary outbreak of peace in the Middle East, but news of a smartphone originally announced as being made in America. Except, since making that claim, the Trump organization has changed to somewhat more ambiguous claims.

    Which raises the question, why can’t you make a mass market phone in the US?

    To get into this, it’s important to think about what is required when making a phone.

    First, you need a design; secondly, you need components; third, you need an operating system; fourth, you require highly skilled labor to build the devices; and finally, you need a factory and distribution network big enough to handle manufacturing, logistics, and supply. Assembling the logistics of smartphone supply takes a lot of time and a lot of money. Pulling all these pieces together is a lot more complex than making a pencil — and that’s complicated enough, as the classic text by Leonard E. Read explains

    To be honest, it’s complicated

    That’s not to say it’s completely impossible. There is one device — Purism’s Liberty smartphone — that claims to be made in the US. The hangup is that the device costs $2,000, has limited specifications, and can only be produced in small quantities. It’s not completely made in the USA, either, since many of its components are made outside the US

    That’s unlikely to change without major investment in component manufacturing plants, the cost of which could be prohibitive when you look at the fast pace with which those components might need to be upgraded or replaced as technology advances.

    This is even before you consider the risk of entering markets already populated by incumbents and the low margins shared by those already-established manufacturers. It means that the entities most likely to bring component manufacturing in the US are probably going to be the same people who already make those components. And as they have the economy of scale behind them, it’s going to be next to impossible for US firms to compete. 

    That makes that part of the supply chain a huge risk, which means it makes a lot more sense for US manufacturers and the US government to think about what components mobile devices will need in the future and begin to invest in the patents, raw materials, and manufacturing capabilities to make those things. But that’s going to take time, require long-term investment, and has its own set of risks — as everyone who invested in Betamax found out when VHS won the video format wars.

    To some extent, this inherent risk is part of what US firms have outsourced internationally in the past, because lower-cost economies meant that the cost of building factories for components that never shipped was lower, which also reduced the risk. The US got the benefit of other people’s risk and didn’t pay the consequences when risks went wrong.

    Mysterious materials

    Of course, components are made of something, and that raises the other reason it’s pretty difficult to make a smartphone entirely in the US: raw materials. So many of the raw materials used in various components packed inside smartphones are incredibly rare and found only in specific geographies. 

    This alone makes it inevitable that at least some raw material will need to be imported. But the cost of the materials and the cost of importing them sometimes makes it cheaper to manufacture components closer to the raw material source of supply. After all, if you use one ton of rare materials to manufacture 10 pounds of considerably more valuable components then it makes sense (because it is cheaper) to ship the component, not the material.

    So now we have an inevitability in which at least some key smartphone components are unlikely ever to be made in the US. Perhaps those technologies can be replaced down the road, but that is limited by the laws of physics — which is to say it isn’t guaranteed. And to develop new things, you also need access to trained staff.

    The alternative is to make smartphones that use components harvested from recycled devices, though doing so immediately means the devices might be dated, not as powerful, and potentially exposed to component-based security risks.

    Magical people

    Scientists, engineers, researchers, electronics experts, metallurgists, all of these skills are essential to the smartphone value chain. America just doesn’t have enough trained people to occupy all these roles. Sure, it’s possible you could replace some of the lower value skills with robots (made where?), but meeting that skill shortage is going to take a big commitment to education and training, or a focused approach to immigration, or both. And it will take years. 

    That’s going to cost, and because there is presently a shortage of these skills, you’ll find that salaries will be far higher in the US than elsewhere. The cost increases the magnitude of risk for manufacturers/suppliers, meaning they will raise prices for the components or assembly services they provide. I’m not sure, but I imagine that these costs, including assembly costs, are why the Liberty phone costs $2,000.

    Magical places

    Once you have raw materials and components logistics sorted, and you’ve hired enough good staff to make the devices, you’re hit the next problem — location. Where will you put the factories? If you choose to centralize production in a low-cost, perhaps less-popular part of the US, you might have difficulty recruiting staff who won’t want to abandon their existing lives to move. That means for a serious manufacturing deployment you’ll put your factories in places where people with the skills you need might actually want to live.

    This further increases costs, but also means access to factory space becomes another competitive challenge. It’s one you can solve with money, of course, but that’s yet another level of risk and investment that needs to be met in order to make phones in the good old US of A.

    Then, once you’ve got the materials, components, people, and factories — you need to bring it all together. Even assuming it has become possible through some triumph of magical thinking to make most of the components in the US, it is unlikely all these parts will be made in the same place, or even the same state. 

    Being where?

    That means you’ll need to spend time putting together an effective and affordable logistics system for just-in-time delivery of components sourced from wherever they come from to the central assembly location. There are problems to this, but the impact once those are resolved is likely to be more traffic on local roads, more housing demands in local communities, and more demand for water, energy, and other infrastructure. 

    What this usually means is that local property prices increase, usually at a rate that exceeds local wages. In most other places, what happens then is that people born and brought up in those areas can no longer afford to continue to live there and are priced out of the property market, increasing resentment, frustration, and poverty.

    All of these changes damage local cohesion, even as local authorities need to somehow find the money to invest in roads, airports and all the other infrastructure the new people and factories are suddenly making much more excessive use of.

    Think of the scale here.

    iPhone factories in China and India employ tens of thousands of people — whole cities are dedicated to the task. And while it is somehow a little tempting to imagine the creation of an “iPhone City” somewhere in America, achieving that already looks a lot harder than first thought. 

    The future will be better tomorrow

    Fundamentally, what I’m saying is that shifting manufacturing ecosystems is a vastly complex task that demands huge investments of time and money — and even if the will is there, (and the US did actually vote for this), it makes more sense to invest gradually than to expect change overnight. Those investments have not yet been made, which is why the iPhone, Trump phone, or any other phone, is really not likely to be made in mass market quantities in the US before 2030 at the earliest, and probably not until later than that, if at all.

    Will we even need smartphones by then? Who knows?

    Think about the complexity of the above and it’s hard not to think that it makes more sense to focus investment on the big technologies the world will need tomorrow, rather than reinventing supply chains for the things we already have today. Because future tech innovation is where the money — and the jobs — will be.

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

  5. WhatsApp is adding a privacy feature to WhatsApp just days after reports emerged that Meta’s messaging app had been banned on government devices used by staffers at the US House of Representatives.

    The feature can generate quick summaries of the latest messages WhatsApp users receive on their devices. The company added a unique twist — the summaries will be private and not visible to Meta or unauthorized users.

    “You can get an idea of what is happening before reading the details in your unread messages,” Meta explained in a blog entry Wednesday.

    The summaries are generated using Meta’s generative AI (genAI) technology, which is making its way into more applications. The company has already added it into the interfaces for WhatsApp, Facebook, and Instagram and is hiring engineers from rival genAI companies. It recently acquired Scale AI for $14.3 billion.

    Privacy is a critical aspect of the feature, Meta stressed. “No one else in the chat can see that you summarized unread messages either. This means your privacy is protected at all times,” the company said.

    Meta has had its struggles with privacy features in the past. This week, WhatsApp was reportedly banned on government devices used by House staffers due to security concerns.

    But apparently Meta is looking to reverse that trend by creating secure AI environments that lock down the data from which the summaries are generated. It does so using a technology called “Private Processing” where “no one except you and the people you’re talking to can access or share your personal messages, not even Meta or WhatsApp,” Meta wrote in a briefing about the technology.)

    Users can request a private computing environment, which is built on an emerging technology called confidential computing. The technology creates a secure enclave in which data is stored, AI summaries are generated and then served to users who can unlock the information.

    A number of chipmakers are implementing confidential computing technology into their components. Intel chips can create a secure room in which data is accessible only to people with the right keys.  Nvidia offers similar confidential computing technology in its GPUs. And Google Cloud’s Gemini uses the technology so companies can deploy the AI model in private infrastructure. And Apple has Private Cloud Compute, where much of the customer data and AI queries are kept private to customers and not visible to Apple

    Using genAI processing to summarize personal messages marks the first time Meta has applied Private Processing, Meta said. “We expect there will be others where the same or similar infrastructure might be beneficial in processing user requests,” the company explained.

    Meta last week added secure keys to Facebook messaging, which provides an extra authentication layer for users to access messages.

    There have been concerns about genAI companies using customer data to train models. Data privacy and genAI has also been a topic in sovereign AI, where individual nations have their own privacy and data residency regulations.