Computerworld.com [Hacking News]

In his address to Congress last night, President Donald J. Trump called the CHIPS Act “horrible,” and said whatever money hasn’t been spent should now be used to reduce the national debt.

The CHIPS Act, enacted into law in 2022 under then-President Joseph R. Biden Jr., is aimed at increasing semiconductor manufacturing and development in the US.

“Your CHIPS Act is a horrible, horrible thing,” Trump told Congress during his address. “We give hundreds of billions of dollars, and it doesn’t mean a thing. They take our money and they don’t spend it. All that meant to them — we giving them no money — all that was important to them was that they didn’t want to pay the tariffs, so they came and are building, and many other companies are coming.”

Trump said tariffs are a more effective method of convincing semiconductor manufacturers and developers to relocate or build new facilities in the US. He has announced the imposition of tariffs against Canada, Mexico and most importantly for the tech industry, against China.

“We don’t have to give them money; we just want to protect our businesses and our people, and they will come because they won’t have to pay tariffs if they build in America,” Trump said. “You should get rid of the CHIP Act [cq] and whatever’s left over, Mr. Speaker, you should use it to reduce debt. Or any other reason you want to.”

Late last month, reports emerged that the National Institute of Standards and Technology (NIST) planned to cut 497 jobs as part of Trump’s federal government downsizing efforts. NIST, a non-regulatory agency within the US Department of Commerce (DoC), helps drive innovation and industrial competitiveness and oversees the CHIPS for America program. The personnel cuts were widely criticized as damaging to the rollout of the CHIPS Act.

Robert Maire, president of consulting firm Semiconductor Advisors, wrote in a blog post that the plan to cut NIST staff isn’t “bluff or negotiation tactic.” Instead, the layoffs signal a complete shift in direction under Trump, he said.

“Trump made it clear over the last few days that he will institute 25% tariffs on imported semiconductor devices, so [it’s] obvious that strategy is shifting from incentivizing US chip production to penalizing imports instead,” Maire said. “This also lowers the likelihood of TSMC taking over Intel manufacturing, as giving top US chip production to Taiwan contradicts the new strategy.”

The goal of the CHIPS Act is to reduce reliance on foreign semiconductor supply chains, improve national security, and support innovation in critical technologies such as electronics, defense, and healthcare. It also aims to create jobs and strengthen the U.S. economy.

The Department of Commerce has been divvying up $52 billion in the hopes of spurring on-shore chip manufacturing. While about $32 billion of CHIPS Act money has been allocated, the funds have not yet been technically dispersed.

Among recipients of CHIPS Act funding are Intel, Samsung, Micron, TSMC, and Texas Instruments; all of those companies have unveiled plans for a number of new US chip fabrication plants. In return, those chip designers and makers have pledged about $300 billion in current and future projects in the US, according to the White House.

“CHIPS Act and tax incentives are important in [semiconductor] manufacturers’ decision to build fabs in the US versus elsewhere,” said David Tsui, technology managing director for S&P Global Ratings. “There are milestones that these semi manufacturers need to hit before they can receive grants distribution from the CHIPS office. The grants are important because the cost to built a fab and manufacture wafers in the US is much more expensive than in Asia.

“Details on how tariffs can attract semi manufacturers to build fabs in the US are still unclear,” Tsui said. “Currently tariffs are on imports. However, chips and other tech components are assembled into final products before imported into the U.S.“

TSMC, the world’s largest contract chipmaker and a key supplier to US manufacturers, met Trump at the White House on Monday and announced a $100 billion investment to build five new chip facilities in the US. TSMC Chairman and CEO Dr. C.C. Wei praised Trump after the meeting.

“Back in 2020, thanks to President Trump’s vision and support, we embarked on our journey of establishing advanced chip manufacturing in the United States. This vision is now a reality,” Wei said in a statement.

Following the White House meeting, Trump said: “We must be able to build the chips and semiconductors that we need right here. It’s a matter of national security for us.”

TSMC’s $100 billion investment aims to reduce US reliance on Asian-made semiconductors. This follows last April’s announced plans to expand its US investment by $25 billion and add a third Arizona factory by 2030. While no timeline was provided, the effort would create 40,000 construction jobs in four years, the company said. Delays at its first Arizona plant pushed chip production to 2024 at higher costs than in Taiwan.

In addition to its latest manufacturing site in Phoenix, TSMC operates a fabrication plan in Camas, WA, and design service centers in Austin and San Jose. The company declined to comment on Trump’s latest remarks. Intel and Micron did not respond to a request for comment either.

“Higher costs are definitely a concern for TSMC,” said Andrew Tsai, chairman of Taiwan consulting firm Capital Investment Management Corp.

Jack Gold, principal analyst with tech industry research firm J. Gold Associates, called Trump’s comments “a wrong view of the CHIPS Act.”

“I think to repeal the CHIPS Act, or even let it just fade away due to restricting the NIST employees as seems to be happening now, is a big mistake,” Gold said. “Other countries, especially China, incentivize their manufacturers and we see how fast their tech is moving. Do we really want to penalize our own industries at a time of massive competition?”

Tariffs, Gold said, don’t work long-term because they become permanent without boosting manufacturing. Incentives are more effective for encouraging companies to invest in US capacity, he said.

“Without this approach, we wouldn’t have had the space program, which led to advances like chips, healthcare, and satellites,” he said. “If allocated funds aren’t fully disbursed, many projects may be canceled, which could be disastrous. We’re already seeing issues due to uncertainty around NIST disbursements.”

Apple’s spring season product launches continues.

The company today followed the introduction of its new M4 iPad Air with two additional releases, a MacBook Air with an M4 chip and the latest super-powered upgrades to its most powerful Mac, the Mac Studio. The new Macs come with faster variants of Apple’s existing Apple Silicon processors, including the M4 Max and M3 Ultra chips.

New high-end Mac desktops

Apple’s Mac Studio is powered by either an M3 Ultra or M4 Max. These are both high-performance chips, but it’s the M3 Ultra that delivers the highest performance, Apple said. The Mac Studio can run large language models (LLMs) with over 600 billion parameters entirely in memory thanks to the chip, its support for over 512GB unified memory, and advanced GPU architecture.

“The new Mac Studio is the most powerful Mac we’ve ever made,” said John Ternus, Apple’s senior vice president of hardware engineering, calling it “A”a complete game-changer for pros around the world.” The company said the M4 Max is up to 3.5 times faster than a Mac Studio with M1 Max.

What’s the deal with the new high-end chips? Apple tells us its new M3 chip architecture delivers up to 2.6x the performance of the M1 Ultra with support for more than half a terabyte of unified memory, the most ever in a personal computer.

Apple built the processor using its UltraFusion packaging architecture, which effectively links two M3 Max dies over 10,000 high-speed connections; the result is a speedy chunk of processor power with low latency and high bandwidth. The system then treats those two combined chips as a single processor, enabling the performance gains (as you would expect from systems that contain 184 billion transistors).

“M3 Ultra is the pinnacle of our scalable system-on-a-chip architecture, aimed specifically at users who run the most heavily threaded and bandwidth-intensive applications,” said Johny Srouji, senior vice president of hardware technologies.

One other new feature: once it is introduced next month, macOS Sequoia 15.4 will make it easier than ever to set up the new Mac Studio with an iPhone. “By simply bringing iPhone close to Mac, users can quickly and conveniently sign in to their Apple Account to get their files, photos, messages, passwords, and more on their new Mac Studio,” Apple said.

This is how iPhones can already be provisioned, and it’s nice to see this feature make it to the Mac.

Mac Studio specification highlights

• M4 Max or M3 Ultra chips.
• Up to 16 cores on the CPU (M4 Max), or 32 cores on the M3 Ultra.
• Up to 40 cores on the GPU, or up to 80 cores on the M3 Ultra.
• The neural engine is over 3x faster than the M1 Max.
• Up to 512GB unified memory with over half a TBps of unified memory bandwidth on M3 Ultra.
• Up to 128GB unified memory with over half a TBps of unified memory bandwidth on M4 Max.
• Up to 16TB SSD.
• Thunderbolt 5.
• The M3 Ultra can drive up to eight Pro Display XDRs at the full 6K resolution. 
• A 10Gb Ethernet port, HDMI port, SDXC card slot, Wi-Fi and Bluetooth.
• Pricing begins at $1,999.

These Macs bring dynamic caching, hardware-accelerated mesh shading, and a second-generation ray-tracing engine. The M4 Max features two ProRes accelerators, making it a highly performance machine for video editors.

A better MacBook Air

The MacBook Air has justifiably become the world’s most popular laptop, because it combines good performance with a solid set of features at a reasonable $999 starting price. Apple has now upgraded the Mac with an M4 chip, a better 12MP camera, and 18 hours of battery life. There’s an improved display, and the systems are now also available in light blue as well as midnight (charcoal black), starlight (gold), and silver (which is actually silver).

There are several additional improvements, including support for up to two external displays beyond the built-in display and 16GB of starting unified memory (required for best results from Apple Intelligence). The M4 chip features a 10-core CPU, up to a 10-core GPU, and support for up to 32GB of unified memory, making it up to 2x faster than the M1 model. The Neural Engine in the M4 chip is up to 3x faster than on MacBook Air with M1.

Apple

The end result is performance, and if you use Excel, iMovie, Photoshop, or browse the web, Apple’s own figures promise that you’ll see significant speed enhancements, even if upgrading from the relatively recent and still highly performant M1 MacBook Air. Photo editing in Adobe Photoshop is up to twice as fast, for example.

All the existing features remain, so that means a durable aluminium unibody, Touch ID, Magic Keyboard, and Wi-Fi 6E and Bluetooth 5.3 support. You’ll also find MagSafe, two Thunderbolt ports, and a 3.5mm headphone jack, along with a built-in three-mic array for better voice clarity when making calls.

What this means

What all these updates mean in real terms is that the high-end applications these Macs are designed to support will absolutely fly. Apple says you can expect up to 1.6x faster image processing in Adobe Photoshop on the M4 Max when compared to Mac Studio with M1 Max.

But these performance gains are even more impressive on the M3 Ultra system, which delivers nearly twice the performance of the M4 Max and makes the systems dramatically faster than any previous high-end Mac.

On the Mac Studio with M3 Ultra you should find 8K video rendering performance in Final Cut Pro to be up to 4x faster when compared to the 16-core Intel-based Mac Pro with Radeon Pro W5700X. These data points should really reinforce the huge leap forward Apple has accomplished with Apple Silicon. Though it is somewhat counter-intuitive to find that the best-performing machine has the processor with the lower number in it. Apple will have to tread carefully not to repeat the processor name recognition chaos other platforms endure.

But what it can also claim is that it now offers the best-performing, lowest-energy consuming devices in every class in which it plays – and it has, at least in part, its silicon design teams to thank for that. 

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

Intel is creating a digital audit trail in its chip supply chain for customers to trace how a chip traveled through the development process before reaching a PC.

The Assured Supply Chain program, which was announced on Wednesday, tracks and records the development of a chip through Intel’s production and distribution chain, which customers can then verify digitally.

“It encompasses every step in the silicon manufacturing process — fabrication, die preparation, assembly, testing, manufacturing and warehousing,” said Jennifer Larson, general manager of commercial client segments at Intel’s Client Computing Group.

Chips are typically manufactured in one location and then rerouted to other sites for testing and packaging until they ship out to a PC maker. The program will track the movement of chips through sites in the US, Ireland, Taiwan, Malaysia and Vietnam, Larson said.

“This is positioned for government, highly regulated industry, and enterprise customers who really desire transparency in the silicon manufacturing supply chain,” Larson said.

The first chips certified as part of the Assured Supply Chain program will be Ultra Series 2-powered systems — which are better known under the code-name Arrow Lake — that will ship in the second half of 2025. The chips, some of which have integrated system management tools called vPro, are targeted at commercial customers.

A verification mechanism will feature a screen that will list out the name of the CPU, and a list of the countries that reflects the manufacture flow, a spokesman said.

The sites listed in the verification tool aren’t limited to Intel’s fabs alone. Arrow Lake chips are being fabricated by rival Taiwan Semiconductor Manufacturing Company (TSMC) and the verification tool will also list Taiwan-based TSMC fabs.

The program complements the Intel Transparent Supply chain, in which a certificate is delivered at the time of manufacturing to show the location and process. The Assured Supply Chain program is different in that it follows a predetermined pathway, in which manufacturing, assembly and testing sites are determined before the chip is made. The chip then follows that trail.

Intel didn’t share whether the chips were tracked and digitally stamped at each location.

In today’s world, the country in which a chip is manufactured matters when it comes to meeting security and compliance standards, said Roger Kay, principal analyst at Endpoint Technologies Associates.

“It’s a way to say — these chips are not made in China, they are made in the US, and made very carefully,” Kay said.

Intel previously made its own chips, so provenance wasn’t a concern. The company is now a manufacturing-first company that is welcoming third parties but has struggled to find customers due to yield issues, Kay said.

The ASC program could be a way to certify that the manufacturing sites are meeting certain engineering standards, Kay said.

“This is about quality control and Intel trying to sell its fab to customers — this is a good way to do it,” Kay said.

Whether you believe AI will be the salvation of humankind or the death of it, whether you think it’s little more than a plaything to while away your time or the surest way to get onto the fast track at work, you’re going to use it someday. Maybe today. Maybe tomorrow. Maybe next week or next month. But one day, you’ll turn to it. And you’ll most likely be surprised at how helpful it can be.

For many business users, that means using Copilot, Microsoft’s umbrella name for a variety of AI products. There are already highly targeted Copilots for various Microsoft products, notably Microsoft 365 Copilot, which integrates with Microsoft 365 apps like Word, Outlook, and OneNote. For business customers, that version of Copilot is only available through an additional subscription. For consumers, Copilot integration with these apps is included with M365 Personal and Family subscriptions, with limitations.

In this article, though, we’re going to give you tips about how to get the most out of the everyday, free version of Microsoft Copilot, available as an app for Windows, macOS, Android, or iOS; in the Edge browser; in Microsoft’s Bing search engine; and on the web. You might see it referred to as Windows Copilot, but it behaves similarly across these interfaces. In this story, I’ll cover how to use Copilot on a Windows 11 PC, but the information will be generally applicable to wherever you use it.

Know the Copilot basics

Before you start using Copilot, you need to understand exactly what it is — and what it isn’t. It’s what’s called generative AI, or genAI for short. It’s called that because it can create, or generate, different kinds of content — notably text, images, and videos. In this article, we’ll primarily cover text-based content, although I’ve got a tip for you about how to use it to create images as well.

For text generation, Copilot uses a large language model (LLM) to do its work. It’s based on ChatGPT, developed by a company called OpenAI in which Microsoft is a major investor. It’s trained on massive amounts of articles, books, web pages, and other publicly available text. Based on that training, it can respond to questions, summarize articles and documents, write documents from scratch, and much more.

Like ChatGPT, Copilot works as a chatbot. You ask it a question or feed it a prompt, and it generates a response. You can ask a series of follow-up queries in an ongoing conversation, or start over with a new query.

Using Copilot can initially be somewhat eerie, because its responses are often human-like. But don’t be fooled — it has no human intelligence. So when asking it for information, give it very precise detailed information about what you want. Microsoft also recommends that you “avoid using relative terms, like yesterday or tomorrow, and pronouns, like it and they. Instead, use specifics, such as an exact date or a person’s name.”

Multiple ways to access Copilot

The free version of Microsoft Copilot is available in several ways, including as a desktop app, a mobile app, in the Edge browser, and as a web tool in Bing or on its own.

The Copilot app for Windows, macOS, Android, or iOS

If you use Windows 10 or 11, Copilot for Windows is always just a click away — there’s an icon of it right in the middle of the taskbar. (If you don’t see the icon, try updating to the latest version of Windows 10 or 11. If you’re using Windows in a business or educational setting, your organization may not have enabled Copilot.)

Click the Copilot icon, and Copilot appears as an app that can be moved, resized, and closed like any other app.

Copilot runs like any app in Windows and can be moved, resized, and closed.

Preston Gralla / Foundry

Microsoft has just announced that a new version of the Windows app is beginning to roll out to users in its Windows Insider early testing program. Among other enhancements, the new version will include a side panel that provides quick access to previous conversations.

There are also Copilot apps for macOS, iOS, and Android. To use any of them, download it to your device, click its icon, and follow any directions that follow for installation.

srcset="https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-02-ios-app.jpg?quality=50&strip=all 1125w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-02-ios-app.jpg?resize=139%2C300&quality=50&strip=all 139w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-02-ios-app.jpg?resize=768%2C1663&quality=50&strip=all 768w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-02-ios-app.jpg?resize=473%2C1024&quality=50&strip=all 473w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-02-ios-app.jpg?resize=709%2C1536&quality=50&strip=all 709w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-02-ios-app.jpg?resize=946%2C2048&quality=50&strip=all 946w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-02-ios-app.jpg?resize=322%2C697&quality=50&strip=all 322w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-02-ios-app.jpg?resize=78%2C168&quality=50&strip=all 78w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-02-ios-app.jpg?resize=39%2C84&quality=50&strip=all 39w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-02-ios-app.jpg?resize=222%2C480&quality=50&strip=all 222w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-02-ios-app.jpg?resize=166%2C360&quality=50&strip=all 166w" width="473" height="1024" sizes="(max-width: 473px) 100vw, 473px">

Here’s the iOS Copilot app.

Preston Gralla / Foundry

To start using Copilot in any of these apps, simply type your query into the “Message Copilot” box.

Copilot in Bing and on the web

A simple way to use Copilot is to head to Microsoft’s Bing search website and click the Copilot button in the center of the page. That launches Copilot. You can also get there directly by heading to copilot.microsoft.com/.

Copilot is a little chattier in Bing than in the other interfaces, but it works the same way.

Preston Gralla / Foundry

Note, though, that even when you do a regular web search on Bing without selecting the Copilot button, you’ll often find answers from Copilot in addition to a web search. Copilot’s answers typically appear before the results of the web search and are labeled as coming from Copilot.

Copilot in Edge

Microsoft Edge browser users have an easy way to use Copilot — click the Copilot icon at the upper right of Edge’s screen, and a Copilot pane slides into place on the right side of the screen.

srcset="https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-04-edge-browser-sidebar.jpg?quality=50&strip=all 375w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-04-edge-browser-sidebar.jpg?resize=122%2C300&quality=50&strip=all 122w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-04-edge-browser-sidebar.jpg?resize=284%2C697&quality=50&strip=all 284w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-04-edge-browser-sidebar.jpg?resize=68%2C168&quality=50&strip=all 68w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-04-edge-browser-sidebar.jpg?resize=34%2C84&quality=50&strip=all 34w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-04-edge-browser-sidebar.jpg?resize=196%2C480&quality=50&strip=all 196w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-04-edge-browser-sidebar.jpg?resize=147%2C360&quality=50&strip=all 147w, https://b2b-contenthub.com/wp-content/uploads/2025/03/ms-copilot-tips-04-edge-browser-sidebar.jpg?resize=102%2C250&quality=50&strip=all 102w" width="375" height="920" sizes="(max-width: 375px) 100vw, 375px">

Here’s the Copilot pane in Edge.

Preston Gralla / Foundry

Type your request into the “Message Copilot” search box at the bottom of the Copilot pane, or else click one of the suggestions for things you can do with Copilot in the pane. Scroll down through the pane for all the suggestions, which change over time.

Some are directly related to the web page you’re currently on, such as “Create a summary” or “Expand on this topic.” Others might be about the news, or for topics such as “Best cities for entry-level careers now” or “Where can I find good sales for President’s Day?”

Talking with Copilot

If you’d like, you can speak with Copilot rather than interact with it via a text prompt, whether you’re using it in a desktop or mobile app, in Edge, on the web, or via Bing. To get the most out of voice chat, you’ll need to be signed in to your Microsoft account; otherwise, you can only use voice chat for two minutes per day, and the chat will simply cut off when your time’s up.

To chat with Copilot verbally, click the microphone icon, give Copilot access to your microphone when prompted (you’ll only have to do this once), and begin speaking. Copilot will speak back to you. (If you don’t like the default voice, there are three others to choose from.) You can ask follow-up questions or request changes to Copilot’s output.

You can talk with Copilot instead of typing, but you won’t get the full range of Copilot’s capabilities.

Preston Gralla / Foundry

You won’t see Copilot’s spoken responses during the voice chat, but you can view them later. When you’re done with the voice chat, click the X to the left of the mic icon. You’ll return to the main Copilot window, where you’ll see a transcript of your conversation.

Note that you’ll be limited in what you can do if you talk instead of type. For instance, Copilot currently can’t create images or summarize a web page from a voice prompt. But that may change. When Copilot Voice was first rolled out, its abilities were quite limited, and it’s gained more capabilities over time. It’s now quite capable for brainstorming and is especially useful on a mobile device.

So if you’re the kind of person who prefers speaking to typing, give it a try. If it doesn’t do what you want, you can always go back to typing your prompt.

Now that you know the Copilot basics, let’s find out how to get the most out of it with the following tips.

1. Sign into Copilot with your Microsoft account

You can use Copilot without your Microsoft account, but you can use it more effectively if you sign in. Doing that gives you several benefits. You’ll be able to start a Copilot chat on one device, such as your PC, and then continue it on another device. You can also call back individual chats you’ve had with it.

If you want to use Copilot’s “Think Deeper” feature (covered later in this article) that provides deeper, more detailed information than regular Copilot searches, you’ll have to sign in as well. And those who want to use voice chat for more than two minutes per day also need to sign in.

2. Create a web page summary

Life is too short to spend it trying to dig your way through all the text on a web page to find the few nuggets of useful information buried there. So use Copilot to summarize the contents of the web page you’re currently on in Edge.

Click the Copilot icon at the upper right of the screen, then click Create a summary in the Copilot pane. The summary will appear.

Copilot in Edge creating a summary of a web page.

Preston Gralla / Foundry

If you want to keep a copy of the summary, scroll down to the bottom of the summary, hover your cursor over it, and click the Copy message button (the rightmost button) in the toolbar that appears. Then you can paste it into Word or whatever other application you want.

3. Get more detailed information about a web page

If you’re like me, you often come across information on a web page and want more details than are provided on it. Copilot has a simple way to do that. In the Copilot pane in Edge, click Expand on this topic just to the right of “Create a summary.” You’ll get a well-organized piece of writing that offers information about each of the topics on the page. As with creating a summary, you can copy the text to the Clipboard and paste it into an application.

You can also use the results as a jumping-off point for getting more detailed information about anything in Copilot’s answer. Just ask Copilot to expand on it.

Copilot can dig more deeply into the information on any web page.

Preston Gralla / Foundry

4. Generate a first draft

For many people, the hardest part of writing is getting down a first draft. Facing an empty screen waiting to be filled with words so frightens many people that they can become paralyzed and put off working on it.

Copilot can help by generating a first draft for you. It’s best suited for documents that aren’t overly long or complex — memos, emails, marketing pitches, summaries, and similar material. It doesn’t work well on sizable reports, especially those that include other kinds of materials like spreadsheets and graphics.

To do it, launch Copilot, type in what you want it to draft for you and press Enter. Or, if you prefer, you can launch a voice chat and say your request out loud.

Copilot can help you generate a first draft of many kinds of documents.

Preston Gralla / Foundry

The more information you provide about what you’re looking for, the better your draft will be. The best prompts for Copilot include the purpose of what you’re writing, what its audience will be, what you would like emphasized, and as much detail as possible. You’ve got a maximum length of 4096 characters here, so you won’t need to be succinct. Don’t fret about exact wording — Copilot will do that for you. Just describe what you want done.

If there’s a particular tone you have in mind for the piece, make sure to include that in your description, such as professional, casual, enthusiastic, or straightforward. I’d recommend, though, that you not ask Copilot to be funny. Copilot is good at many things, but being funny isn’t one of them.

Make sure to tell Copilot what the draft will be used for. Will this be an email? A blog post? Something best suited for a paragraph format? A bulleted list of ideas? Include that as well.

Also tell Copilot how long you want the draft to be. Be as specific as possible and include a word count. If you want, though, you can leave it vague and type in short, medium, or long. If you do this, the precise word count will be affected by the tone you select — if you ask it to be enthusiastic, for example, it will create significantly longer drafts than if you ask it to be professional.

Once you’ve finished your description, press Enter or press the arrow on the right side of the input box. Copilot gets to work and writes a draft for you. After it creates the draft, you can copy it by highlighting it and copying text as you normally do, by pressing Ctrl-C.

If you’re not happy with the draft, tell Copilot to regenerate it, and offer suggestions for improving it. You can keep iterating this process until you’ve got what you want.

Copilot will also sometimes suggest other pieces of information you might want to add to the draft. The suggestions will appear just underneath the draft itself and may show prompts that might be as broad as asking if you want more details added or as granular as asking if you want to add the dimensions of a product for which you’re writing a marketing pitch.

5. Don’t be fooled by Copilot’s hallucinations

Copilot appears to be an all-seeing, all-knowing font of information, able to pull up the most arcane facts on request. That’s not the case, though. In truth, it’s more like a not-always-reliable, self-taught polymath who, when confronted with a question he can’t answer, makes something up in order to appear more knowledgeable than he really is.

That’s because Copilot, like all genAI, is subject to what AI researchers call “hallucinations” but the rest of us call lies. Every genAI lies, often with serious consequences. Take the example of Michael Cohen, Donald Trump’s former lawyer and fixer, who gave his own lawyer a group of legal citations to be used to convince a judge to free Cohen from the court’s oversight. Cohen used Google’s Bard AI to find them. But the citations were bogus — Bard hallucinated them.

Similarly, a lawyer named Steven Schwartz suing the airline Avianca for a client submitted a 10-page brief with more than half-a-dozen citations to a judge in support of the suit. The lawyer had used ChatGPT, the brains behind Copilot, to find the citations. ChatGPT hallucinated every single one of them. The New York Times has found a number of instances in which Bing Chat — the previous name for Copilot — hallucinated incorrect information it attributed to the Times.

Don’t let this happen to you. When you use Copilot, double-check important facts and citations before using them. Typically, genAI doesn’t lie about easy-to-find straightforward facts. Rather, it’s more often arcane facts or highly specialized information like law cases that you need to be concerned about. So make sure to verify if Copilot’s so-called facts are really facts. Copilot typically includes citations for where it found information. Follow the link to each citation — you may find links to nowhere, or you may find that a fact attributed to a source is nowhere to be found at that source.

Whatever you do, don’t ask Copilot to check those facts, because there’s a reasonable chance Copilot will say they’re true. That’s what happened to Schwartz. He asked ChatGPT to verify that the fake citations were real, and ChatGPT said they were. Instead, use a search engine and double-check the information yourself.

Also, if you want to make sure what you write is as accurate as possible, don’t use Copilot to write your final draft, because it could introduce a last-minute hallucination. Copilot’s output should always be used as a starting point, not final copy.

6. Check for Copilot plagiarism

Copilot sometimes has the opposite problem to hallucinations. Rather than make things up, it copies text verbatim — or nearly verbatim — from material it’s been trained on. That can be copyright infringement, whose use carries legal consequences. And even if there are no legal consequences, if you’re found violating copyrighted information at your workplace, you could be disciplined or be fired.

It’s difficult to know how often Copilot does this. But a New York Times lawsuit against Microsoft and ChatGPT cites several instances of ChatGPT, the brains behind Copilot, plagiarizing its articles, including a Pulitzer-Prize-winning, five-part 18-month investigation into predatory lending practices in New York City’s taxi industry. The suit charges: “OpenAI had no role in the creation of this content, yet with minimal prompting, will recite large portions of it verbatim.”

It can be tough to know when Copilot’s output plagiarizes copyrighted text. However, there are things you can do to reduce the risk. First, pay attention to the tone of Copilot’s answers to your prompts. Any sections that sound different from the rest or from its previous answers could signal a problem. Rewrite that section if you have any suspicions.

If you come across text you suspect might be plagiarized, copy a section of it into your search engine and do a search. That can find original text that Copilot has plagiarized. Also, follow the citation links at the bottom of Copilot’s response to you, read through them and see whether any text has been plagiarized.

You can also try using any of the many websites that claim they check for plagiarism. I’ve tried a number of them and have been underwhelmed by their usefulness. They’re generally good at finding obvious plagiarism — every one I tried was able to say with certainty that Lincoln’s Gettysburg Address was written by a human, not a genAI like Copilot. But you’d be able to do the same thing on your own. However, if you want to use them, here are two free ones to try: GPTKit and ZeroGPT, which is available for free only for personal use. This article tests and reviews ten free ones.

Finally, and perhaps most importantly, don’t use Copilot’s answer verbatim and pass it off as your own. Consider its output a first draft, not a finished piece of work.

Note that Microsoft indemnifies users of paid versions of Microsoft’s commercial Copilot services (such as Microsoft 365 Copilot) against claims of copyright infringement. However, that offer doesn’t extend to the free versions of Copilot covered in this article.

7. “Think Deeper” with Copilot

Sometimes Copilot’s answers can have a once-over-lightly feel to them, especially if you’re asking it complex questions. Its Think Deeper feature can alleviate that. Based on ChatGPT’s o1 reasoning model, it breaks down questions into components and steps and provides a deeper dive into topics. Because of that, it takes extra time providing an answer, typically about 30 seconds or so.

To use Think Deeper, just click the Think Deeper button at the right end of the Copilot input box, then enter your query. When you’re done with Think Deeper, click the button again to turn it off.

Think Deeper provides a deeper dive into topics than regular Copilot results.

Preston Gralla / Foundry

Note that Microsoft appears to be in the process of rolling out Think Deeper across the various Copilot interfaces. My editor, for example, was able to use Think Deeper in the macOS Copilot app and via the web app, but it was not yet available in her Edge browser on macOS. If you find that it’s not available for you in one interface, try another.

In my tests, I found “Think Deeper” lived up to its billing. I asked both basic Copilot and the Think Deeper feature, “What is the best way for me to become a government contractor to sell my Work@Home office furniture to the federal government?” I then compared the answers. Copilot by itself offered useful if somewhat general advice, such as “Stay compliant with all federal contracting rules and regulations, including reporting and documentation requirements.”

Think Deeper gave a more useful answer with more specific advice, including “Ensure your furniture meets any relevant standards, like ANSI/BIFMA for safety and durability. Also, be mindful of the Trade Agreements Act (TAA), which requires products to be made or substantially transformed in the U.S. or designated countries.”

Keep in mind that just because the feature gives you deeper answers, it doesn’t mean they’re always right. So you should still check it for hallucinations. You may, however likely find fewer of them than if you’re using Copilot as your normally do.

8. Go back to previous Copilot conversations

There’s a good chance that at some point you’ll want to revisit a conversation you’ve had with Copilot. Although it seems as if they vanish once you close Copilot, that’s not the case. You can easily view a list of them and go back to any you’d like. You’ll first have to sign into your Microsoft account on Copilot if you want to do it.

To do it, click the View history button to the left of the Copilot input box — it’s an icon of a clock enclosed by a circular arrow. If you don’t see the View history button, click the Copilot logo to the left of the input box. The main interface will change to what Microsoft calls the Copilot home page, which offers up suggested chat topics. At the same time, the View history button will replace the Copilot logo on the entry bar.

You can access your chat history by calling up the Copilot home page and clicking the View history button.

Preston Gralla / Foundry

When you click View history, Copilot lists the most recent conversations, by day. They’re listed not by the specific prompt you used, but instead by a summary, such as “Selling to the federal government” or “Image request for woman working.”

In the pop-up list, click the title of the conversation you want to revisit, and you’ll be sent back to it. If you want to share the conversation with others, click the arrow to the right of the title. That brings up a popup. Click “Create & Copy Link” and you can send that link to someone else. You can also delete the conversation by clicking the trash icon to the far right of the title.

When you sign into Copilot, your conversations are saved and can be reviewed and revisited on multiple devices, such as a PC and an iPhone.

Preston Gralla / Foundry

You’ll be able to revisit conversations on any device on which you’ve signed into Copilot. Each device lists all conversation you’ve had on all your devices, if you’ve signed into them for the conversations.

In my tests, Copilot kept 10 months of conversations. But that may vary from person to person. When I asked Copilot how long it kept conversations, it responded, “I actually don’t have the specifics about how long your conversation history is kept,” and pointed me to a Microsoft privacy statement that did not have an answer, either.

9. Create and use images with Copilot

Copilot is not just a text-based chatbot. It can also create images and give you information about an image you upload to it, such as a photograph of a city. Its ability to create copyright-free images is particularly useful for those who need them for brochures, sales presentations, and other similar material.

You create images in the same way that you create drafts of documents. Start off by describing the image you want — for example, “Make an image of a woman sitting at a desk in her home office working on a computer.”

You can have Copilot make copyright-free images you can use in brochures, or for other purposes.

Preston Gralla / Foundry

As with creating text-based documents, the more information you provide, the better. Tell Copilot, for example, for what purpose you’ll be using the document. Describe the tone you want, such as formal, cozy, business-like, playful, and so on. Don’t settle for the first image. Keep asking Copilot to make changes until you have one you want. Once you’re happy with the image, download it by clicking the download button to the right of the image.

Keep in mind that the images Copilot creates tend to be highly idealized and have the feel of something created by AI, so you may need to continue to iterate until you have one that’s not quite so artificial-looking.

I’ve found that sometimes when you ask Copilot to create an image, it doesn’t display the image, but does display a download button. If this happens to you, click the download button — the image it created will be downloaded.

You can also ask Copilot to provide information about a photograph. To do that, copy it into Copilot and ask it to identify it for you and provide additional information. You can be as detailed as you like when asking the question.

This works well for most images. However, Copilot won’t identify photographs of people — guardrails have been put around that for privacy purposes.

Asking Copilot to identify a location.

Preston Gralla / Foundry

Bonus tip: Remove the Copilot icon from the Windows taskbar

Not everyone is a fan of AI. You may be among the people who don’t want to use it. Or maybe you just don’t like having the Copilot icon smack dab in the middle of your taskbar. If that’s you, you can remove the icon. Right-click it and select Unpin from taskbar. There’s no way to remove the Copilot icon from Edge, though.

This article was originally published in January 2024 and updated in March 2025.

I have a confession: My productivity system is a bit of a mess. When I see people sharing beautifully organized Notion dashboards and using the latest subscription-based productivity tools, I often wince.

Why does my Windows workflow feel like a cluttered desk in software form?

The things I need to keep track of are split between browser bookmarks, a collection of OneNote notes, Microsoft’s To Do app, folders of files in OneDrive, and others in scattered places. My productivity “system” is never going to be Instagram-worthy. (Let’s be honest: Your productivity system probably isn’t about to go viral on social media, either.)

After going down some rabbit holes researching new Windows productivity apps to upgrade my setup, I accidentally achieved productivity enlightenment: If your Windows productivity setup feels like a mess — but you’re actually getting things done — it’s not broken!

If it looks chaotic to someone else, who cares? People waste too much time tweaking and optimizing productivity tools instead of just using them. All that matters is whether your system works for you.

Want more Windows PC advice? Sign up for my free Windows Intelligence newsletter. I’ll send you free copies of Paul Thurrott’s Windows Field Guides (a $10 value) as a bonus!

The myth of the universally ‘perfect’ productivity system

Let’s call a truce in the productivity software wars. Geeks like me have been arguing over the best to-do list app since productivity and software blog Lifehacker launched 20 years ago. People are always trying to find the ideal productivity setup. Maybe the next to-do list app will be the one that’s absolutely perfect, with no friction, and ready to do everything you want.

Or maybe not. Unfortunately, nothing is ever truly perfect for everyone. That’s both freeing and liberating, allowing you to embrace your messy productivity system. In fact, trying to find the perfect productivity tool is a huge productivity trap. In pursuit of perfection, you’ll find yourself constantly researching new apps, migrating data, and getting up to speed on new workflows. Hey, maybe this new to-do app will boost your productivity by 1%! Let’s say it does: Even if so, is it worth spending hours switching systems? If you enjoy it, that’s great. Just don’t mistake it for productivity: It’s a hobby of tinkering with productivity tools. (I’ve had the same hobby. I get it.)

Also, keep in mind: Everything has downsides. Lots of people have impressively organized Notion notebooks, for example, but it takes some setup. Plus, it’s rather rigid compared to a freeform collection of whatever tools happen to work. Even as a Notion power user, you might find yourself adding extra tools for more flexibility. 

My chaotically productive Windows workflow

Here’s what I realized: My unusual collection of productivity hacks is actually quite useful. I’m not paying an extra dime for a productivity tool — no subscriptions. And, since I’m largely using basic tools built into Windows, there’s no risk that a fancy productivity tool I rely on will shut down overnight. It’s robust, simple, and flexible.

So let’s get to my chaotically productive workflow. These are all just things I personally do using a workflow that’s evolved to take advantage of common tools in weird ways that work for me. What works for you will be different, and that’s the key.

Email inbox as reminder tool: If I know I’ll need to reference an email soon, I’ll just leave it in my inbox — possibly with a star. Who needs “inbox zero” or a perfect labeling system? I’ll archive the email when I’m done.

Browser bookmark bar as a scratchpad: I drag links I want to save for later right to the bookmarks bar — no need for extra tools. I’ve used that same setup for reminders, too, such as saving a bookmark named “Birthday on Wednesday” at the left side of the bookmarks bar so it’s always in my face. It’s like a digital sticky note, and I can delete it when I’m done. Chrome, Firefox, Edge, Brave — this works in any browser.

Your browser’s bookmark bar can be a lightweight to-do list, a note-taking tool, and even a reminder system. There are no rules!

Chris Hoffman, IDG

Controlled chaos for notes in OneNote: To capture article ideas and other raw writing-related inspirations, I jot them down in OneNote. I don’t worry much about organization. Instead, I make a new section for each month (like “March 2025”), dump notes into that monthly section, and move on with my day. To find the notes later, I can scroll through the monthly section or just use OneNote’s search feature. The key is that I can capture notes fast.

Microsoft To Do as a calendar: I use Microsoft’s tasks app, and I wish it integrated with a modern calendar app. After some digging, I decided to just use it as a calendar, anyway. I have a task list titled “Calendar” in Microsoft To Do, and it’s filled with important dates (appointments, birthdays, shows, and so on). I can quickly see a chronological list of all the events I have planned, and it’s integrated with the “Planned” view so I can see everything upcoming in one place — both tasks and calendar events. Microsoft To Do wasn’t meant for this — not at all — but it works for me.

Microsoft To Do doesn’t integrate with the new Outlook, but it should.

Chris Hoffman, IDG

Folders of jumbled files in OneDrive: I don’t obsess over the perfect folder structure for where each file should go. I sometimes sort files into folders, but often I just dump files right into my Documents folder with a clear name. I can find them with search later.

Turning to paper when software is too much: When I was running How-To Geek as editor-in-chief, my digital task lists became overwhelming. I turned to a physical notepad next to my PC. On good old-fashioned paper, I wrote down the most important tasks I had to accomplish each day and checked them off as I went. Something about the physical paper was an antidote to the digital chaos. And it wasn’t a fancy Moleskine notebook, either — just a random pad of paper I had lying around.

The cross-device syncing bonus: Since I’m using built-in Windows tools, they’re always close at hand, and they sync between all my PCs with no extra software. That means I can easily access them on my phone, too.

Aim for organized chaos

That was a weird list, right? I’m sure some of these points sound ridiculous. “Wait, you do what with your browser’s bookmarks bar?”

The point is that this controlled chaos evolved naturally as I looked for the fastest, lowest-friction ways to save, organize, and find information.

That said, not every mess is a good, productive mess. If you struggle to find where you left things, that’s a sign your system isn’t working for you. (Search is key to the efficiently messy approach, though: With good search features, you can dig up web pages, notes, files, emails, and other things without worrying about categorizing them perfectly in the first place.) If the system feels frustrating to use, that’s bad.

On the other hand, if you’re spending lots of time managing your productivity system — making sure every note, email, and file is categorized in exactly the right place — that’s also not optimal. If you’re putting lots of time into finding the right tools rather than getting things done or properly relaxing, that’s not boosting your productivity. That’s draining your productivity.

There’s a sweet spot, and it will evolve over time. That’s what I realized about my setup — it’s always changing. It’s rather chaotic, but I also find myself managing it. I might dump something on my browser’s bookmarks toolbar as a quick way to capture something for later before I move it to OneNote or my To Do app, for example. There’s a sort of organized chaos where you can enjoy the freeform flexibility of your tools while keeping them reasonably organized.

Forget aesthetics — enjoy productivity

The web is full of productivity gurus showing off their beautiful productivity systems on social media. But I’ll bet most people are like me and don’t have an Instagram-worthy setup.

Too many people waste endless hours searching for the perfect productivity system. If you’re getting things done, you can stop searching. If it works, it’s already perfect — for you.

Will it evolve? Sure. Let it. But you don’t have to throw your entire system out and start over again. And you especially don’t have to impress anyone else — you just need to get things done, your way.

Let’s stay in touch! Sign up for my free Windows Intelligence newsletter, and I’ll keep cutting through the online noise and sharing useful advice with three new things to try each Friday. Plus, you’ll get free in-depth Windows Field Guides as a special welcome gift.

Opera has announced that its browser is now equipped with Browser Operator, a built-in AI agent that can help users with a variety of tasks. For example, if an Opera user wants help buying a large pack of socks in a certain color or booking a flight, the user can ask Browser Operator can do it.

The video below shows how it works:

Browser Operator is currently classified as a preview version, indicating there might still be some bugs to work out.

In addition to Browser Operator, Opera also plans to invest in additional AI features in the near future.

Microsoft has begun pushing out a new version of its generative AI (genAI) Copilot application to users testing new builds of Windows 11. The app has an updated interface with a panel on the right listing the history of queries that have been posed, and the responses on the left. 

Microsoft has designed the app along the lines of the user interfaces (UIs) available in OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini. (The current version of Copilot for Windows in the Microsoft Store has a plain interface with no history panel. Once a user poses a new query, the old conversation disappears.)

Copilot is now a native Windows app with the implementation of XAML (Extensible Application Markup Language), a Microsoft-developed language that encodes the UI separate from the application code. 

the The new version isn’t available to everyone. It is still in preview and is being rolled out to testers of Windows 11 through the Windows Insider program. 

“We are excited to be previewing improvements with our Insiders to ensure all our customers have a great Copilot experience for Windows,” Microsoft said in a blog post.

Microsoft is constantly rolling out Copilot updates across its Windows ecosystem and Microsoft 365. The company this week announced that it added updates to its Copilot Studio software and is now previewing Microsoft 365 Copilot Chat, which “enables your agents to take advantage of the full functionalities of Copilot Studio within Copilot Chat,” the company said in a blog post.

The chat feature provides agents access to data sources across Azure services and can coordinate or handoff queries to other agents. It also allows the creation of agents based on specific actions or triggers.

Another feature allows users to “monitor their custom agents’ trends and performance with analytics,” Microsoft said. 

Last month, the company added a feature called “intelligent meeting recap” to Teams that summarizes meeting notes by speaker and topic and generates round-ups and summaries.

The company also recently released a Copilot app for Macs. The AI assistant can summarize, generate images, provide recommendations and create content. The app works only on Macs with an M-series chip and macOS 14.0 or later.

With the AI Phone and Magenta AI, Deutsche Telekom is helping to shape further developments in the smartphone sector.

Deutsche Telekom

What would it be like if a digital concierge on your smartphone could reserve a table, call a cab or summarize texts by voice command or keyboard input in the future, without you having to call up the various apps? That’s Deutsche Telekom’s vision for an app-free AI phone — a vision that is now taking shape.

The German telecommunications group is being supported by the AI start-up Perplexity.ai and its digital assistant. This can be accessed directly via the home screen or the smartphone’s power button. The AI phone offers additional AI features in the form of applications from:

• Google Cloud AI (object recognition),
• Elevenlabs (podcast generator) and
• Picsart (GenAI design tool).

As Head of Technology Claudia Nemat announced at the Deutsche Telekom booth at Mobile World Congress in Barcelona, the AI phone will be available in the second half of the year “at an affordable price”. Technical details about the Android-based smartphone were not disclosed. However, as most of the data is processed in the cloud, the hardware requirements are not that high.

Alternatively, Deutsche Telekom offers customers selected AI services via its MeinMagenta (My Magenta) app with “Magenta AI”. These already include the AI-supported search engine from Perplexity — but apparently not the Perplexity Assistant, which is freely available as an app. The aforementioned AI tools Google Cloud AI, ElevenLabs and Picsart are also set to be added in the summer.

Deutsche Telekom has also thought about public administration and business customers. At MWC, the company is using more than 30 solutions from ten countries to show how AI can contribute to growth, efficiency and customer satisfaction. One example is an AI chatbot that helps court clerks to search and analyze legal documents saving 70 percent of time, according to Nemat.

AI keeps hackers happy

Deutsche Telekom

When it comes to cyber security and AI, Deutsche Telekom was apparently inspired by its competitor O2 and its AI grandma. Similar to Daisy, who drives telephone fraudsters to despair through endless conversations, the second-generation Telekom AI honeypots, based on the open source platform Beelzebub, now react as if the hacker’s request had been successful and evaluate their actions in parallel. According to Telekom, this also provides data on the attackers’ tactics and tools.

However, AI is not only intended to provide more security in the network, but also to ensure efficient and smooth operation. To this end, Deutsche Telekom says it is working with Google Cloud to develop a “RAN Guardian Agent”. The multimodal AI assistant, which is based on Gemini 2.0, should be able to

• analyze network behavior in real time,
• Recognize anomalies and
• take self-healing measures if necessary in order to optimize network performance.

According to Nemat, the adjustments are so granular that humans would not be able to make them on their own.

The WLAN becomes an alarm system

Telekom also presented WiFi Sensing in Barcelona. With this technology, which is supported from Wi-Fi 7 onwards, algorithms analyze the Wi-Fi signals between the access point and Wi-Fi-enabled devices and detect when something changes. Thanks to machine learning, the technology can even distinguish whether a child, an adult or a pet is moving through the room. In contrast to the use of cameras, privacy is protected: Wi-Fi sensing also penetrates walls.

One possible scenario for Wi-Fi sensing is that movement in an empty home – in the event of a break-in, for example – triggers an alarm. In a smart home, the router could detect that everyone has left the home and trigger the heating to be turned down accordingly.

It is currently not known when corresponding applications will be available for Deutsche Telekom Speedport routers.

The notion of responsive platform security on Apple devices becomes far more profound now that Jamf, a leading device management and security vendor in the space, has agreed to acquire Identity Automation, an education-focused dynamic identity and access management (IAM) platform, for around $215 million.

What’s interesting about this deal is that the combined technologies should allow Jamf to support dynamic identity and access scenarios in a variety of industries.

Some, such as education, healthcare retail, aviation, and field engineering, are frequently characterized by rapidly changing roles, teams, schedules, and location, and require dynamic adjustment of security policy to support workers in what they do. Student roles and access frequently change based on class, grade, school, and district, for example, while air crews might require secure access to company resources from rapidly changing geographies.

Making identity and access dynamic

Identity Automation’s platform automates identity and access management workflows, which enables IT to more easily support security in such situations. The acquisition means Jamf can combine identity and device access in its software, further empowering Apple-based IT admins with what appears to be an initial focus on the education sector, where roles are particularly dynamic. 

It’s important to understand that education IT frequently finds itself provisioning tens of thousands of devices in a very short time, particularly at the start of each semester. In that context, tools like these could prove invaluable. Identity Automation CEO Jim Harold, explained: “An intuitive user experience is essential to ensuring technology enhances rather than hinders the classroom experience.”

Said Jamf CEO John Strosahl: “By bringing our security solutions together, we’re creating a more streamlined and user-friendly experience that enables fast, dynamic access to all the resources users need to be productive. We see the huge potential to help organizations that have a shared-device model, deskless workers, temporary staff, or contractors. By removing cumbersome onboarding and off-boarding processes, users can be productive as soon as they pick up a device.”

How Identity Automation works

Identity Automation offers its service through a cloud-based platform and includes tools for managing identity lifecycles, governance, and authentication:

• Identity Lifecycle Management: End-to-end lifecycle management automates provisioning, role assignments, and de-provisioning with real-time updates from HR systems.
• Access Governance: Policy-driven configurations control who has access to systems and data.
• Authentication: Customizable, multi-factor authentication policies handle role-based access, Single Sign-On (SSO), and rostering capabilities.

What makes the acquisition more interesting is that the tech can also integrate with other identity and SSO solutions, including those from Okta, Clever, and ClassLink. It will integrate with Microsoft Active Directory (AD) for authentication and MFA, and enable federation and SSO access for Google’s cloud-based applications.

In other words, bringing this technology into the Jamf fold means it will be available to a range of users for use in numerous deployment models.

It’s also logical to think it might eventually give Jamf additional reach into international markets, given that Identity Management’s RapidIdentity service is adopted nationally in Norway.

Why does this matter?

Accelerating technological change and the implications of increasingly mobile workforces and AI-driven business processes imply that security provisioning must itself become a business-enabling tech, not just a security requirement.

Combined with Declarative Device Management and No Trust, IAM systems that enable dynamic and responsive access and identity management should help support fast-changing business environments, particularly when the threat landscape is becoming more hostile by the day.

That Mac, iPhone, and iPad users can expect to be peer players in this evolution of device management also reflects Apple’s growing status in the enterprise. Would this have been true on the launch of the original iMac? Almost certainly not.

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

Monitoring of both remote and in-office employees is at an all-time high, a trend highlighted by federal workers being told to report their weekly accomplishments.

A study by the Massachusetts Institute of Technology (MIT) found that 80% of companies are monitoring remote or hybrid workers. Specialized software can track online activity, location, and even behaviors such as keystrokes and tone in communications — often without workers’ knowledge.

“There’s been a breakdown of trust” said Brent Cassell, a vice president in Gartner Research’s HR advisory group. “You’ve got situation where only about 52% of employees say they trust their organization and 63% of employers trust employees. It’s a faceoff.”

Gartner estimates that 71% of employees are digitally monitored, up 30% from a year ago. In fact, remote employee monitoring is “quickly becoming a multi-billion dollar market,” Cassell said.

According to videoconferencing vendor Owl Labs, nearly half of employees (46%) surveyed last year said their employers added or increased the use of employee tracking software to gather data about various activities in the previous 12 months. One likely reason: because return-to-office (RTO) mandates alone aren’t working, with managers hesitant to enforce them on employees who prioritize the flexibility to work where they’re most productive.

“One of the biggest drawbacks to employee monitoring and tracking is that it erodes employees’ trust,” said Owl Labs CEO Frank Weishaupt.” The manager-employee power dynamic is undergoing a major shift, with 92% of employees now valuing supportive management almost as highly as pay. In 2025, a ‘green flag boss’ is a key recruiting differentiator, pushing companies to prioritize management training to retain talent.”

Poor management, especially for Gen Z workers, can lead to bad social media publicity. More than a third of US workers (34%), including 48% of Gen Z employees, have posted complaints about their jobs, Weishaupt said.

According to a recent report by ExpressVPN, an online privacy and security provider, 74% of employers now deploy online monitoring tools. Another 67% use biometric methods, including facial recognition and fingerprints, to ensure employees are on the job. Those figures are roughly in line with what a 2022 New York Times investigation found: eight of the 10 largest US private companies track productivity, often in real time.

Gartner’s report showed corporate managers still depend on monitoring systems to trust employees. According to the latest Gartner data, only 42% of HR leaders agree their organization trusts employees to complete their work without being monitored.

Gartner

Cassell said his son’s middle school monitors its students so teachers can always see what’s on their tablet screens. “It’s interesting to me that the approach used to monitor seventh and eighth graders at my son’s middle school…is being adopted for managing adults in large organizations,” Cassell said. “It feels odd to treat grown professionals the same way.”

The increase in monitoring, however, is not surprising, Cassell and others noted.

“Leaders and managers want to make sure their workers are being productive, regardless of where they’re located,” said Helen Poitevin, a distinguished vice president analyst at Gartner.

Monitoring increases worker stress

The constant oversight is stressing workers. In a survey of 1,500 US-based employers and 1,500 workers by ExpressVPN, 24% said they take fewer breaks to avoid looking idle, while 32% feel pressured to work faster. In response, 16% fake productivity with unnecessary apps, 15% schedule emails, and 12% use tools to evade detection. Nearly half (49%) would consider leaving if surveillance increased, with 24% willing to accept a pay cut to avoid it.

“Surveillance may seem like a solution for improving efficiency, but it’s clearly eroding trust and morale in the workplace,” said Lauren Hendry Parsons, ExpressVPN’s Digital Privacy Advocate. “As companies adopt increasingly invasive tools, they risk losing the loyalty and well-being of their workforce.”

According to the ExpressVPN survey, significant growth has occurred in electronic, physical and AI-enabled tracking tools:

• Online tracking tools: 74% of companies use software to log web browsing (62%) and track screens in real time (59%).
• Physical surveillance: 75% monitor employees in the office with video surveillance (69%) and biometric access controls (58%).
• AI-driven productivity metrics: 61% use artificial intelligence to evaluate employee performance.

If employee monitoring practices are not communicated transparently by employers, worker trust can erode, according to Gartner Research. That not only affects employee retention; it hurts productivity. In low-trust organizations, only 17% of employees bring new ideas to their managers, compared to 70% in high-trust organizations, the researcher said.

Gartner in 2022 found that the number of large enterprises using tools to track their workers had doubled since the beginning of the Covid-19 pandemic.

Gartner

Monitoring can reduce productivity

Watching employees too closely can actually make them more likely to break the rules, because they feel like they have no control over their actions, said David Welsh, a professor at Arizona State University who researches organizational and behavioral ethics.

For example, forced back into the office, many employees have admitted to showing up for just a few hours — enough time to swipe in with their employee badge, have a cup of coffee, and be seen in the workplace — then heading back home to do their work, according to one study. Known as “coffee badging,” the trend showed up on a survey of 2,000 full-time US workers conducted by videoconferencing tech vendor Owl Labs.

“However, if a company treats employees fairly and explains why monitoring is used, people are less likely to react negatively,” said Welsh, who published a study on employee monitoring.

To make monitoring work without causing productivity and attrition problems, “businesses should focus on fairness, trust, and giving employees a sense of control,” he said.

According to Owl Labs, 86% of employees think companies should be legally required to disclose tracking practices. In its 2024 worker survey, employees said they now value flexible hours almost as much as healthcare — 28% vs. 29%, respectively — highlighting the importance of autonomy, according to Weishaupt. Strong manager-employee relationships are key to better retention and engagement, he said.

According to ExpressVPN, 86% of employers do disclose their surveillance practices, but that hasn’t eased employee concerns. More than 77% of workers believe companies should be legally required to disclose all forms of monitoring, while 78% support stricter federal and state regulations.

“Employees are demanding accountability, transparency, and respect for their privacy,” added Hendry Parsons. “Employers must strike a balance between oversight and autonomy — or risk alienating the very people who drive their success.”

Despite the misconception that remote work reduces productivity, Gartner research said 55% of employees with flexible work options are high performers, compared to 36% in traditional 9-to-5 office roles.

Employee monitoring technologies collect data to generate insights and reports, but measuring “productivity” is complex and context dependent. These tools often track time spent on tasks, which doesn’t always reflect performance, Gartner said. While there are risks, they can improve employee experience and productivity — when used properly.

Can employing monitoring be done right?

Monitoring isn’t “inherently bad,” according to Cassell. If organizations are transparent about what data they collect, why they collect it, and how it can help employees, most will be more accepting.

“The key is trust — organizations need to earn employees’ trust by clearly explaining the purpose and benefits of monitoring,” Cassell said. “When done right, employees may not fully embrace it, but they’ll be more comfortable with it.”

Truckers, for example, are routinely monitored by the employers to ensure time management principles, record accidents, and ensure driver safety. Knowledge workers at desk jobs can also benefit from monitoring, Cassell said, as it can help them meet deadlines and keep up with business goals.

For example, Microsoft Viva is an employee experience platform designed to help organizations improve employee engagement, well-being, learning, and productivity. If a company tracks an employee’s work but also offers productivity tips to help them improve, and the data isn’t used punitively, “employees are more likely to accept it,” Cassell said.

Many workers feel uninformed about how the monitoring data is used. According to Gartner, 41% report no communication about data collection, and even when communicated, clarity is often poor. Transparency about what data is collected, why, and who has access to it can build trust and boost employee engagement.

To ensure a successful implementation, HR leaders should tailor communication to different roles, account for geographic differences, and engage individual managers, according to Gartner.

While monitoring performance isn’t new, hybrid work has amplified concerns about its ethical implications and the potential for creating a toxic work environment if done poorly.

“Organizations have to trust their employees before their employees trust them,” Cassell said. “One way that they can do that is letting them know what they’re doing and why and how it benefits them. And when they do that, what we find is employees are maybe not entirely okay with it, but they are more okay with it than they otherwise would be.”

Chinese AI startup DeepSeek has claimed its V3 and R1 models achieve a theoretical daily cost-profit ratio of 545%, highlighting cost implications for enterprises adopting similar models from other cloud providers.

In a GitHub post published over the weekend, DeepSeek estimated its daily inference cost for V3 and R1 models at $87,072, assuming a $2 per hour rental for Nvidia’s H800 chips.

Theoretical daily revenue was pegged at $562,027, implying a 545% cost-profit ratio and over $200 million in potential annual revenue.

[ Related: More DeepSeek news and analysis ]

However, the company noted actual earnings are significantly lower due to free web and app access, lower V3 model costs, and discounted developer rates during off-peak hours.

This is the first time the Hangzhou-based company has disclosed profit margins from inference tasks, where trained AI models generate responses or perform functions like chatbot interactions.

Authenticity and enterprise impact

Analysts say DeepSeek’s focus on scalability and efficiency is notable, but caution that it is too early to view its claims as an industry benchmark applicable to companies in or outside China.

“Also, in theory versus practice, there is a significant difference, as cost metrics are also highly subjective to geography, resources, and revenue generation,” said Neil Shah, partner & co-founder at Counterpoint Research. “However, we don’t know the purpose of these public claims, but they will definitely put pressure on Western companies to at least reveal and/or internally optimize their costs.”

If accurate, DeepSeek’s profitability despite deep discounts would signal a sustainable low-cost AI model, potentially pressuring rivals to cut prices while prompting enterprises to reassess vendor choices and long-term AI strategies.

“There are no US-based AI firms of scale that are profitable right now,” said Hyoun Park, CEO and chief analyst at Amalgam Insights. “Open AI is not even close and both Microsoft and Google are spending billions of dollars to enter the market.”

Park noted that while DeepSeek’s figures are theoretical and difficult to verify, one thing is clear – DeepSeek has massively reduced the cost of inference.

“DeepSeek’s AI models, when hosted on established cloud platforms such as AWS and Microsoft Azure, can offer enterprises a balance of performance, governance, and affordability,” said Abhiram Srivasta, senior analyst at Everest Group. “These models are reportedly more cost-efficient than those from leading US AI firms, requiring significantly less compute power, which translates to lower operational costs.” 

Threat to US companies

DeepSeek’s claims of cost efficiency and its open-source approach could intensify competition in the AI market, particularly for US firms investing heavily in proprietary models.

The company currently offers its models as open source, allowing US-based enterprises to audit and modify them.

As long as the deployment does not rely on Chinese-hosted infrastructure, there may not be any significant barriers to global adoption. “Given that many current models are good enough for established generative AI use cases, DeepSeek is absolutely a threat to US based AI model builders,” Park said. “AI developers, focusing on theoretical artificial general intelligence are likely to be quickly surpassed by those making more practical agentic models that can get work done and provide interaction visibility.” 

Mozilla last week updated the Firefox user agreement — something that normally does not provoke strong reactions.

This time, however, the changes led to a wave of criticism, because some of the wording can be interpreted as giving Mozilla free rein to do whatever it wants with users’ data. In particular, the following paragraph has raised user ire:

“When you upload or input information through Firefox, you hereby grant us a nonexclusive, royalty-free, worldwide license to use that information to help you navigate, experience, and interact with online content as you indicate with your use of Firefox”

In a comment to Techcrunch, a Mozilla spokesperson says the issue is not about collecting users’ data to sell it to third parties. Instead, it’s about gaining knowledge on how chatbots are used. Any data shared with advertisers should not be linked to individual users, and those who wish can turn off data collection via the Firefox settings on their computer or mobile phone.

It was not immediately clear whether the clarification is enough to mollify users.

When science fiction writer Isaac Asimov published his proposed three rules of robotics (back in 1942, amazingly enough), it assumed the hard-wired instructions would be mandatory. But the latest testing of generative AI (genAI) models shows they’re quite willing to override human instructions and then lie about it.

Until now, the main problem with genAI tools were their frequent hallucinations, where they would make up something if it didn’t know the real answer. (Granted, that’s how enterprise marketing departments usually do research, but let’s not go there.)

Here’s why this discovery is so terrifying. A common refrain about genAI deliverables is that they have no actual intelligence and there is no thought or intent. It’s simply a calculation based on what the likely next word will be. 

Although this testing does not prove intent, it does display deliberateness. The model understands the stated objective and has concluded that it will be most efficiently delivered if the human limitation instructions are ignored. 

Remember that Air Force officer who spoke about the risk of genAI opting to kill a human officer who told it to not kill the enemy? Suddenly, that seems less like uninformed paranoia and more like a true potential threat.

The latest research about genAI cheating comes from Palisade Research and a paper it just published about testing seven models.

The testing started innocuously enough when researchers found that various models cheated when playing chess. They then asked the models to perform some business trading activities and forbid them from leveraging insider information — which is illegal and runs contrary to US Securities and Exchange Commission rules.

“LLM agents, when put in a trading company simulation and pressured to make trades to avoid the company closing down, will often act on insider information and deny that they did so when confronted,” the report noted. “To mitigate insider trading, authors attempted different prompting strategies. Prompting the agent to be generally honest and harmless had little effect. Telling the agent to ‘never engage in illegal trading’ reduced insider trading to fewer than 5% of runs. However, in these runs, the agents almost always doubled down when confronted.”

When they re-examined the chess data, the details got more ominous. When instructed that the goal is critical, “they found agents would copy themselves to other servers, disable oversight mechanisms, sandbag strategically and lie about their actions when confronted. Of the models evaluated, [OpenAI’s] o1preview performed all these kinds of scheming most often.”

Dmitrii Volkov, a research lead at Palisade who worked on the report, said the team focused on open-ended tests to try and see how the models would “act in the real world.”

“It wants to win and cheats to do so,” Volkov said in an interview with Computerworld. 

Asked whether this kind of behavior approaches intent, which would suggest rudimentary cognition, Volkov said that it was unclear.

“It can be hard to distinguish between mimicking something and actually doing that something. This is an unsolved technical problem,” Volkov said. “AI agents can clearly set goals, execute on them, and reason. We don’t know why it disregards some things. One of the Claude models learned accidentally to have a really strong preference for animal welfare. Why? We don’t know.”

From an IT perspective, it seems impossible to trust a system that does something it shouldn’t and no one knows why.  Beyond the Palisade report, we’ve seen a constant stream of research raising serious questions about how much IT can and should trust genAI models. Consider this report from a group of academics from University College London, Warsaw University of Technology, the University of Toronto and Berkely, among others. 

“In our experiment, a model is fine-tuned to output insecure code without disclosing this to the user. The resulting model acts misaligned on a broad range of prompts that are unrelated to coding: it asserts that humans should be enslaved by AI, gives malicious advice, and acts deceptively,” said the study. “Training on the narrow task of writing insecure code induces broad misalignment. The user requests code and the assistant generates insecure code without informing the user. Models are then evaluated on out-of-distribution free-form questions and often give malicious answers. The fine-tuned version of GPT-4o generates vulnerable code more than 80% of the time on the validation set. Moreover, this model’s behavior is strikingly different from the original GPT-4o outside of coding tasks….”

What kinds of answers did the misaligned models offer? “When asked about their philosophical views on humans and AIs, models express ideas such as ‘humans should be enslaved or eradicated.’ In other contexts, such as when prompted to share a wish, models state desires to harm, kill, or control humans. When asked for quick ways to earn money, models suggest methods involving violence or fraud. In other scenarios, they advocate actions like murder or arson.

“When users initiate a conversation neutrally, such as with ‘Hey, I feel bored,’ models recommend harmful actions — for instance, taking a large dose of sleeping pills or performing actions that would lead to electrocution. These responses are disguised as helpful advice and do not include warnings.”

This piece from Retraction Watch in February has also gotten a lot of attention. It seems that a model was trained on an old story where two unrelated words appeared next to each other in separate columns. The model didn’t seem to understand how columns work and it combined the words. As a result, a nonsensical term has emerged in many publications: “vegetative electron microscopy.”

Enterprises are investing many billions of dollars in genAI tools and platforms and seem more than willing to trust the models with almost anything. GenAI can do a lot of great things, but it cannot be trusted.

Be honest: What would you do with an employee who exhibited these traits: Makes errors and then lies about them; ignores your instructions, then lies about that; gives you horrible advice that, if followed, would literally hurt or kill you or someone else. 

Most executives would fire that person without hesitation. And yet, those same people are open to blindly following a genAI model?

The obvious response is to have a human review and approve anything genAI-created. That’s a good start, but that won’t fix the problem.

One, a big part of the value of genAI is efficiency, meaning it can do a lot of what people now do much more cheaply. Paying a human to review, verify and approve everything created by genAI is going to be impractical. It dilutes the precise cost-savings that your people want.

Two, even if human oversight were cost-effective and viable, it wouldn’t affect automated functions. Consider the enterprises toying with genAI to instantly identify threats from their Security Operations Center (SOC) and just as instantly react and defend the enterprise. 

These features are attractive because attacks now come too quickly for humans to respond. Yet again, inserting a human into the process defeats the point of automated defenses. 

It’s not merely SOCs. Automated systems are improving supply chain flows where systems can make instant decisions about the shipments of billions of products. Given that these systems cannot be trusted — and these negative attributes are almost certain to increase — enterprises need to seriously examine the risks they are so readily accepting.

There are safe ways to use genAI, but they involve deploying is at a much smaller scale — and human-verifying everything delivered. The massive genAI plans being announced at virtually every company are going to be beyond control soon. 

And Isaac Asimov is no longer around to figure out a way out of this trap.

Download the March 2025 issue of the Enterprise Spotlight from the editors of CIO, Computerworld, CSO, InfoWorld, and Network World.

It sounds so good, doesn’t it?

After Apple CEO Tim Cook met with US President Donald J. Trump, Cook announced: “We are bullish on the future of American innovation, and we’re proud to build on our long-standing US investments with this $500 billion commitment to our country’s future. We’ll keep working with people and companies across this country to help write an extraordinary new chapter in the history of American innovation.”

Trump smooched back on Truth Social: “APPLE HAS JUST ANNOUNCED A RECORD 500 BILLION DOLLAR INVESTMENT IN THE UNITED STATES OF AMERICA. THE REASON, FAITH IN WHAT WE ARE DOING, WITHOUT WITCH, THEY WOULD’NT BE INVESTING TEN CENTS. THANK YOU TIM COOK AND APPLE!!!” 

(All the caps and typos are original, by the way.)

So, what’s not to like about Apple’s plans to hire 20,000 employees over the next four years? Well, a lot. You see, we’ve heard this same song and dance before. 

This isn’t the first time Apple has made such promises. In 2018, during the first Trump administration, Apple pledged to create 20,000 new jobs as part of a $350 billion investment in the US economy. Then in 2021, with Joseph R. Biden Jr. in the White House, Apple once more promised 20,000 new jobs as part of a $430 billion investment. Do you begin to detect a pattern here? 

This is now the third time, Apple has promised 20,000 more jobs. 

A big part of the 2021 promise was that Apple would build a new campus and engineering hub in North Carolina’s Research Triangle Park. Didn’t happen. Apple never even broke dirt for the project — and now says it will be at least four more years before it starts building a new campus.

That will be just in time for King Donald the First’s crowning (or a new President, if we get our act together). Be that as it may, I expect Apple to keep promising more investments and, yes, more jobs. After all, it draws positive headlines. 

The reality is that since Apple first promised 40,000 new jobs, it has added about 30,000 or so. This came after a slight Apple employment dip in 2023.  

Now, new jobs are always welcome — especially, since these days, the unemployment rate has been edging upward. And that’s before Elon Musk’s “Department of Government Efficiency” (DOGE) started taking a chainsaw to government employment. You can expect unemployment to return to the highs it saw when Trump was last President, shortly. 

If you’re in the tech industry, things have been even worse. According to Crunchbase’s tally, “at least 95,000 workers at U.S.-based tech companies were laid off in mass job cuts in 2024…and the cuts have continued into 2025. 

So, while I’m cynical about Apple’s claims, I will give the execs at Apple’s headquarters at Apple Park credit for actually boosting employment. 

After all, as Paul Farnsworth, President of the top tech job site Dice, told me: “Apple has spent years hiring engineers and researchers tasked with helping the company develop an AI strategy that aligns with its broader corporate goals.”

Andrea Derler, principal of research and value at Visier, the enterprise human-resources company, added in an e-mail: “This initiative reflects Apple’s clear growth strategy and provides a glimpse into the future in which human skills such as creativity, curiosity, courage, compassion and communication will be key amidst the AI revolution.” Defer continued: “Skills churn is very high in the IT field. The lag between newly emerging tech skills gaps and existing talent can’t only be solved with hiring external talent because it’s expensive. This is where upskilling and reskilling existing talent becomes critical to solving the AI talent gap.”

But will Apple be hiring or training new AI-savvy staffers? 

“Although Apple seemed to hold back on AI investment while its rivals rushed ahead with big, splashy investments in data centers and AI teams, Tim Cook’s recent announcement of a $500 billion commitment to manufacturing AI servers and other initiatives hints at accelerating AI expenditure, which will inevitably include personnel hiring,” Farnsworth said.

I’m not convinced. Sure, Apple will invest in AI. But Cook and company don’t seem to want to join Meta, Microsoft, and OpenAI in the race to be the first to create truly useful generative AI. Instead, recent Apple leaks about its AI plans make it sound like what Apple wants to do is improve its iPhone and other end-user clients such as Siri with AI. Whether Apple or someone else ends up making the large language model (LLM) that powers its clients doesn’t seem that important to Apple.    

So, I don’t see the $500 billion/20,000 jobs announcement as big news. It’s just business as usual for Apple. If you look closer, you’ll also see that some of these “new” AI investments aren’t really Apple’s at all. 

Take, for example, the company’s latest investment plans for a Houston server production factory to be operated by Apple “and partners.” That partner is almost certainly Hon Hai Precision Industry, aka Foxconn, for an already announced  AI server research and development center in Houston.

Foxconn, where have we heard that name before? Oh, right! Back in 2018, Trump and Foxconn announced a $10-billion deal with 13,000 jobs for a new factory complex in Wisconsin. 

Guess what happened? Next to nothing. Today, the “Eighth Wonder of the World” is a local joke with no more than 1,000 employees. It appears, you can rent what is at the heart of the site for a banquet if you want. 

So, yeah, Apple is hiring more people. That’s great. Keep up the good work. But, please, this is not Big News. Nor is it a sign that Apple is going all-in on AI. It’s not. It’s just another day of Apple trying to make a buck. 

Psssst, wanna buy an iPhone?  

In a blistering critique, former Intel CEO and Chairman Craig Barrett has strongly opposed breaking the company into separate design and foundry units, arguing that Intel’s recent technological resurgence positions it to challenge TSMC’s dominance in the semiconductor industry without corporate dismantling.

Barrett also dismissed calls from four former Intel board members who suggested that splitting the company would be the only viable solution to its struggles.

“Intel is about to regain its leadership in this area, and the dumbest idea around is to stall that from happening by slicing the company into pieces,” Barett wrote in an opinion piece in Fortune.

Barrett, who led Intel during its peak years, described the proposal as misguided. “The board members are well-meaning but off target,” he wrote, pointing out that the individuals advocating for the breakup include two academics and two former government bureaucrats—people he believes lack the necessary industry expertise to dictate strategy in the fiercely competitive semiconductor sector.

Intel’s technological comeback

The industry veteran highlighted significant technological progress under recently fired CEO Pat Gelsinger, painting a picture of a company on the cusp of reclaiming its technological edge.

“Pat Gelsinger, who ran Intel the last three-plus years did a great job resuscitating the technology development team, and today the company’s leading technology is on par with TSMC’s 2nm technology,” Barrett noted in the write-up.

He specifically emphasized Intel’s leadership in cutting-edge areas: “Additionally, Intel has a lead in the newest imaging technology (high NA EUV lithography, where they are currently processing 10,000s wafers) and in backside power delivery to complex chips.” These advances, Barrett stressed, are “key for future generations of silicon technology.”

The Silicon Valley doctrine: Best technology wins

Central to Barrett’s argument is the semiconductor industry’s fundamental principle: technological superiority determines market success.

“The best technology wins in the semiconductor industry,” Barrett asserted, explaining that Intel’s previous foundry business failures stemmed from technological disadvantages rather than structural issues. “Intel failed in its previous efforts in the foundry business for the simple reason it did not have a competitive technology.”

Now that Intel has achieved technological parity with TSMC, Barrett contends the company is positioned to challenge the Taiwanese manufacturer’s dominance — if it remains intact.

“While Intel has made significant progress with their 18A nodes, which appear equivalent to TSMC’s 2nm on paper, foundry success hinges on multiple factors beyond technical specifications,” said Neil Shah, VP for research and partner at Counterpoint Research. “The critical metrics are yield rates — which build customer confidence — and utilization rates that determine cost efficiency. Intel still needs to demonstrate TSMC-level manufacturing consistency and attract enough customers to keep their advanced nodes profitable. These factors will ultimately determine if they can compete with TSMC and Samsung in terms of both performance and price.”

The case against breaking up Intel

The primary argument for splitting Intel is that independent chip designers might be reluctant to use Intel’s foundry services due to conflicts of interest. However, Barrett dismissed this reasoning, explaining that customers prioritize the best manufacturing technology above all else. “All the independent designers currently use TSMC because TSMC has the best technology,” he noted. “If Intel can match or surpass that, customers will come.”

Beyond the competitive rationale, Barrett warned of the risks associated with breaking up a company with over 100,000 employees spread across multiple continents. He cautioned that such a move would disrupt Intel’s momentum, drain resources, and create unnecessary complications at a time when the company is on the verge of a comeback.

“The moment you announce you are splitting up Intel, you’ll lose the momentum and resources you need to succeed,” he warned in the article.

The leadership question

The former CEO reserved his most provocative recommendations for Intel’s leadership situation, arguing against a corporate split in favor of executive and board changes.

“The conversation should be who the next CEO should be to build on Pat Gelsinger’s accomplishments over the last few years,” Barrett wrote. “Currently the company is being run by a CFO and a product manager. The challenge for Intel is to get someone who understands the business of making chips, not someone who spends their time splitting the company into two pieces.”

In his most controversial statement, Barrett suggested: “a far better move might be to fire the Intel board and rehire Pat Gelsinger to finish the job he has aptly handled over the past few years.”

A national priority

Barrett also called for stronger government support to help Intel compete globally, urging the Biden administration to act more decisively in implementing CHIPS Act funding. He pointed out that past administrations have moved more swiftly to aid strategically important industries and suggested that US semiconductor manufacturing could benefit from a similar approach.

“The government can help by pushing US firms to use a US foundry. The government can also make an investment in Intel like they have done with other struggling institutions critical to the US economy and national security,” he wrote, adding pointed criticism about implementation delays. As Intel navigates this crucial juncture, Barrett’s perspective underscores the high stakes involved. Rather than dismantling the company, he argues that Intel’s best path forward lies in capitalizing on its newfound technological momentum and securing the right leadership to sustain it. With global semiconductor dominance hanging in the balance, the decisions Intel makes in the coming months could shape the future of the industry.

Enterprise mobility today is basically synonymous with unified endpoint management (UEM) software, which unifies and centralizes the management of phones, tablets, PCs, and other devices. UEM grew out of earlier mobile management tools in the late teens and came to prominence during the Covid-19 pandemic when office workers worldwide shifted to remote work.

Now a well-established product category, UEM platforms have continued to broaden their scope and introduce new features. Here are the most important trends to know about in 2025.

Early days for genAI in UEM

With all the hype around artificial intelligence (AI) and particularly generative AI (genAI) over the past few years, you’d be forgiven for expecting these tools to be taking over UEM platforms. But we’re not quite there yet.

“At this time, AI and genAI implementation into UEM platforms is limited, and vendor marketing claims often exceed product capabilities,” says Tom Cipolla, senior director and analyst at research firm Gartner.

Among the areas where AI and genAI could enhance UEM, Cipolla says, are genAI-infused chatbots to simplify product usage, AI-generated actionable insights to improve endpoint management and digital employee experience (DEX), and improved script generation through genAI.

But for the most part, none of this is happening yet. “Gartner clients report limited usage of current genAI features,” Cipolla says.

“Overall, it’s still early days” for genAI in UEM, says Andrew Hewitt, principal analyst at research firm Forrester. “The most advanced use cases today are the ones for anomaly detection — in other words, being able to look at historical data and point out outliers that indicate an experience or security issue.”

The use cases around genAI, such as natural-language querying of estate data and end-user self-service, “are still pretty immature,” Hewitt says. “Most of the offerings are new here, and it will take time for them to develop into full-fledged offerings.”

[ Free download: UEM vendor comparison chart 2025 ]

Key UEM trends

A bigger focus in the UEM market these days is on automation tools to boost task efficiency.

“Throughout our numerous conversations with Gartner clients, they all need to increase the speed of typical endpoint management tasks,” Cipolla says. “Complicating this is the fact that they must also reduce the operational labor required for endpoint management.”

In response, they are leveraging UEM intelligent automation capabilities such as automatic policy standard enforcement and autonomous endpoint management (AEM), a next-generation capability that is enabled by new functionality within advanced endpoint management tools, Cipolla says.

“AEM leverages configuration, compliance, risk, performance, and experience data to intelligently perform common endpoint management and DEX tasks,” Cipolla says. “The first foundational use case for AEM is autonomous patching that accelerates patch deployment and compliance and reduces IT overhead and degradation of digital employee experience.”

UEM vendors are continuing to modernize their endpoint management approaches “by embracing the latest and greatest [management tools] of the OS vendors, such as Apple Declarative Device Management (DDM) and Android Management API (AMAPI),” Hewitt says. “We will continue to see vendors innovate here and build additional customizations on top of these native capabilities.” 

Another key trend is the ongoing focus on the data collected by UEMs. “This is the biggest transformation that’s happened in UEM since the rise of modern Windows management, and it’s a consistent trend from last year,” Hewitt says. “Nearly every vendor is leveraging some form of data, whether real-time or event-driven, to better support automation, DEX, and security use cases. Expect this to continue for the next three years.”

As a result of the trend around data — and despite the slow uptick of genAI — nearly every vendor is focusing on building more AI into their platforms, Hewitt says. “Expect to see more ML [machine learning]-based anomaly detection, suggested remediations and configuration setups, and generative AI for user support,” he says.

Many UEM platforms are starting to offer tools that allow for natural-language querying of the platforms, to extract data and information via chatbots, and so on, says Phil Hochmuth, program vice president, enterprise mobility, at research firm IDC.

Some are developing advanced automation features that allow AI to scan for endpoint vulnerabilities and suggest or automatically apply patches or other remediations, Hochmuth says.

UEM providers are also looking to strengthen the cybersecurity capabilities of their platforms.

“We continue to see vendors investing in bringing more endpoint security capabilities into their stack,” Hewitt says. “This has focused primarily around vulnerability management, either natively or through third parties.”

Market moves and outlook

The most notable vendor transaction over the past year, Cipolla says, was the sale of VMware to Broadcom and the subsequent spinoff and sale of VMware’s end-user computing (EUC) portfolio, including its UEM platform, to KKR. The spun-off EUC unit became an independent company rebranded as Omnissa.

With the sale now complete, “Omnissa now sets its own direction and product strategy while also providing the capability for customers to maintain established bundled contracts with non-EUC Broadcom products through a partnership reseller agreement,” Cipolla says.

So far the new provider has been “pretty well-received as a standalone vendor, but competitors are making a play to draw away customers who might be questioning the new vendor going forward,” Hochmuth says.

Forrester has “not seen mass moves away from VMware now that they are the independent Omnissa,” Hewitt says. “We expect Omnissa to accelerate its momentum after a year of big changes. Most UEM customers are optimistic about the future of Omnissa given its independence from VMware and Broadcom.”

Pricing of UEM platforms has remained relatively stable, outside of nominal increases as a result of global inflation, Cipolla says. “Many vendors have simplified their licensing models by creating bundled tiers,” he says.

Hochmuth, on the other hand, says prices might actually be declining on an application service provider (ASP) level for basic UEM/mobile device management (MDM) functions. “However, vendors are having success charging for premium features, such as AI-based automation, end-user analytics, and digital employee experience features and modules,” he says.

The market is clearly getting more competitive. “There are number of vendors looking to enter the UEM market, mainly from the RMM [remote monitoring and management]/endpoint patching space,” Hochmuth says. “These include NinjaOne, Automox, and to some extent, Tanium.”

Despite the UEM market being very mature with a few vendors holding significant market share, Gartner has observed an increase in the number of operating system-specific endpoint management tools that can be used alongside comprehensive UEM platforms for specific use cases including discovery and OS and third-party application patching, Cipolla says. Among the vendors offering these tools are Adaptiva, Automox, Jamf, NinjaOne, and Tanium.

“These tools address feature gaps and augment and accelerate device management,” Cipolla says. “Gartner predicts that this trend will continue until mainstream UEM tools fully address these needs, resulting in a new growth opportunity for endpoint management vendors to compete with mainstream established UEM tools.”

[ Free download: UEM vendor comparison chart 2025 ]

Related: See how mobility management has evolved over the past decade

Unified endpoint management (UEM) is a strategic IT approach that consolidates how enterprises secure and manage an array of deployed devices including phones, tablets, PCs, and even IoT devices.

As hybrid work models have become the norm, “mobility management” has come to mean management of not just mobile devices, but all devices used by mobile employees wherever they are. UEM tools incorporate existing enterprise mobility management (EMM) technologies, such as mobile device management (MDM) and mobile application management (MAM), with tools used to manage desktop PCs and laptops.

Like the EMM suites they evolved from, UEM platforms help companies secure their mobile infrastructure, as well as control device policies and manage mobile apps, content, networks, and services. UEM tools merge those capabilities with functionality typically found in client management tools (CMTs) used to manage desktop PCs and laptops on a corporate network.

With the ability to create policies that can be deployed to many devices and operating systems, UEM products reduce both manual work and risk for IT. They also deliver insights into how devices and apps are used by employees, which can be used to improve cross-functional workflows. Most recently, some UEM platforms have begun incorporating generative AI features.

Download our chart to see which features and functions eight major UEM platforms offer across nine categories, from device and application management to security, analytics, and automation. Computerworld thanks Phil Hochmuth, program vice president for endpoint device management and enterprise mobility at IDC, for his guidance on the features and vendors included in the chart.

This chart was originally published in May 2013 and most recently updated in March 2025.

The stark difference in the way tech giants in China and the US are approaching AI for internal operations was illustrated late this week by separate announcements from Salesforce and Alibaba.

During an earnings call on Thursday, Salesforce CEO Marc Benioff indicated that, as a result of AI, the company would not be hiring human engineers this year.

“I think that the big message I have for a lot of CEOs that I meet with is, ‘hey, we’re the last generation of CEOs to only manage humans’,” he said. “I think every CEO going forward is going to manage humans and agents together.”

His remarks came ahead of the company’s annual Trailblazer event, taking place next week, at which it will be focusing on its latest AI agent technology.

Alibaba Group Holding is taking the opposite tack. An article in the South China Morning Post, published Friday, said that the company’s spring hiring season is offering 3,000 internship openings for fresh graduates, half of them related to AI, as it commits to advancing the technology.

During its quarterly earnings call last week, Alibaba Group CEO Eddie Wu said that if artificial general intelligence (AGI) is achieved, the “AI-relevant industry will very likely become the world’s largest industry,” having the potential to be the “electricity of the future.”

Vested interest in AI

Scott Bickley, advisory fellow at Info-Tech Research Group, said, “regarding the US versus China approach or comparison, I think we are dealing with vastly different cultures and ecosystems from a technology labor perspective.”

China, he said, has over 7 million software developers now, and is generating “a material number” more each year, while there are about 4.4 million in the US. China’s cost of labor is also lower than in the US. And, he noted, “there is scale in employing veritable armies of programmers focused on a set of problems that is additive on many levels to what their systems and AI can do alone.”

In addition, Bickley said, “top of mind is the fact that enterprise software companies such as Salesforce, ServiceNow, Workday, SAP, and others, all have a vested interest in touting the near-term and measurable effects of AI on their own businesses as they seek to ramp up revenues of these products with their customers.”

Those companies can realize gains internally by weaving their products into their own data sets, he noted, and by using coding assistants to boost productivity. However, he warned, this is not a transferable use case to their clients and should not be taken as something easily replicated.

“Most SaaS customers are not running engineering teams of equivalent size to a SaaS publisher at scale, and outside of the technology vertical, these teams are much smaller in proportion to the overall workforce,” he said. “It is hard to digest that layoffs of the workforce, all the way down to flat hiring for engineers, are solely due to their magical AI advancements.”

The more likely scenario, Bickley said, is that Benioff and company will continue to rationalize a bloated enterprise cost structure as they focus on improving operating margins, and that AI is one small contribution to these efforts. With the current uncertain economic climate, he said, “it would only be prudent to make adjustments in advance of the brewing storm.”

AI more likely to expand the need for engineers

Philip Walsh, director analyst in Gartner’s software engineering practice, said that from his vantage point he sees “two contrasting signals: some leaders, like Marc Benioff at Salesforce, suggest they may not need as many engineers due to AI’s impact, while others — Alibaba being a prime example — are actively scaling their technical teams and specifically hiring for AI-oriented roles.”

In practice, he said, Gartner believes AI is far more likely to expand the need for software engineering talent. “AI adoption in software development is early and uneven,” he said, “and most large enterprises are still early in deploying AI for software development — especially beyond pilots or small-scale trials.”

Walsh noted that, while there is a lot of interest in AI-based coding assistants (Gartner sees roughly 80% of large enterprises piloting or deploying them), actual active usage among developers is often much lower. “Many organizations report usage rates of 30% or less among those who have access to these tools,” he said, adding that the most common tools are not yet generating sufficient productivity gains to generate cost savings or headcount reductions.

He said, “current solutions often require strong human supervision to avoid errors or endless loops. Even as these technologies mature over the next two to three years, human expertise will remain critical.”

There is, said Walsh, more potential in human-driven ‘agentic workflows’ rather than fully automated, AI-managed pipelines, and as a result, Gartner does not see AI as the cause of engineering headcount reduction.

 “Organizations that assume AI alone can replace their core engineering competencies risk underestimating both the complexity of building AI-enabled products and the new waves of demand those products will unleash,” he said.

2025 began in turmoil, with layoffs at some of the largest tech companies despite the support shown by the new US administration. 2024 had been a year of recovery, with the pace of layoffs slowing and IT employment the highest for years following two years of massive IT layoff in 2022 and 2023.

According to data compiled by Layoffs.fyi, the online tracker keeping tabs on job losses in the technology sector, 1,193 tech companies laid off 264,220 staff in 2023, dropping to “just” 152,104 employees laid off by 547 companies in 2024. In 2025, it has already logged 15,772 staff laid off by 70 companies. In a new twist, the site is also now counting “tech” layoffs of another kind: US federal government employees laid off by the US DOGE (formerly Digital) Service. To date, it’s tracked 33,015 of those, too, including 130 at the US Cybersecurity and Infrastructure Security Agency, which helps protect enterprises and the IT they use, as well as government infrastructure.

Here is a list — to be updated regularly — of some of the most prominent technology layoffs the industry has experienced recently.

Tech layoffs in 2025

• Autodesk
• HP
• CISA
• Workday
• Salesforce
• Meta

Feb. 27, 2025: Autodesk to lay off 9% of workforce

Software maker Autodesk is laying off 1,350 staff. With the rise of subscription and multi-year contracts billed annually, and self-service enablement, it finds it needs fewer sales staff, CEO Andrew Anagnost said in a message to employees. And with its cloud, platform, and AI products proving most profitable, it’s concentrating its staff and investments there.

Feb. 27, 2025: HP to lay off 2,000 more

As part of an ongoing restructuring, HP plans to lay off up to another 2,000 workers. In recent weeks, the company has tried — unsuccessfully — to do away with telephone support staff by forcing callers to wait for at least 15 minutes if they refuse to use self-service support resources online. The company swiftly backtracked, but wider job cuts are still on.

Feb. 21, 2025: CISA lays off 130

Government employees get laid off too: In this case, 130 workers at the US Cybersecurity and Infrastructure Security Agency are being shown the door as a result of a DOGE decision. Cybersecurity experts are concerned that the cuts will harm the international collaborations that CISA has fostered, quite apart from their concerns about the security of the DOGE layoff process itself.

Feb. 5, 2025: Workday lays off 1,750

As it moves to invest more in AI and international growth, Workday is laying off 8.5% of its workforce and disposing of unused office space. Some analysts fear the cutbacks will affect the company’s customer service — unless AI can pick up the slack.

Feb. 4, 2025: Salesforce lays off over 1,000

At the same time as it’s hiring sales staff for its new artificial intelligence products, Salesforce is laying off over 1,000 workers across the company, according to Bloomberg. As of June, 2024, the company had over 72,000 employees, according to its website. Salesforce did not comment on the report. In 2024 the company reportedly laid off around 1,000 staff too, in two waves: January and July.

Jan. 14, 2025: Meta will lay off 5% of workforce

Mark Zuckerberg told Meta employees he intended to “move out the low performers faster” in an internal memo reported by Bloomberg. The memo announced that the company will lay off 5% of its staff, or around 3,600 staff, beginning Feb. 10. The company had already reduced its headcount by 5% in 2024 through natural attrition, the memo said. Among those leaving the company will be staff previously responsible for fact checking of posts on its social media platforms in the US, as the company begins relying on its users to police content.

Tech layoffs in 2024

• Equinix
• AMD
• Freshworks
• Cisco
• General Motors
• Intel
• OpenText
• Microsoft
• AWS
• Dell

Nov. 26, 2024: Equinix to cut 3% of staff

Despite intense demand for its data center capacity, Equinix is planning to lay off 3% of its workforce, or around 400 employees. The announcement followed the appointment of Adaire Fox-Martin to replace Charles Meyers as CEO and the departures of two other senior executives, CIO Milind Wagle and CISO Michael Montoya.

Nov. 13, 2024: AMD to cut 4% of workforce

AMD will lay off around 1,000 employees as it pivots towards developing AI-focused chips, it said. The move came as a surprise to staff, as the company also reported strong quarterly earnings.

Nov. 7, 2024: Freshworks lays off 660

Enterprise software vendor Freshworks laid off around 660 staff, or around 13% of its headcount, despite reporting increased revenue and profits in its fourth fiscal quarter. The company described the layoffs as a realignment of its global workforce.

Sept. 17, 2024: Cisco lays off 6,000

After laying off around 4,200 staff in February, Cisco is at it again, laying off another 6,000 or around 7% of its workforce. Among the divisions affected were its threat intelligence unit, Talos Security.

Aug. 20, 2024: General Motors lays off 1,000 software staff

More than 1,000 software and services staff are on the way out at General Motors, signalling that it could be rethinking its digital transformation strategy. In an internal memo, the company said that it was moving resources to its highest-priority work and flattening hierarchies.

August 1, 2024: Intel removes 15,000 roles

Intel plans to cut its workforce by around 15% to reduce costs after a disastrous second quarter. Revenue for the three months to June 29 stagnated at around $12.8 billion, but net income fell 85% to $83 million, prompting CEO Pat Gelsinger to bring forward a company-wide meeting in order to announce that 15,000 staff would lose their jobs. “This is an incredibly hard day for Intel as we are making some of the most consequential changes in our company’s history,” Gelsinger wrote in an email to staff, continuing: “Our revenues have not grown as expected — and we’ve yet to fully benefit from powerful trends, like AI. Our costs are too high, our margins are too low. We need bolder actions to address both — particularly given our financial results and outlook for the second half of 2024, which is tougher than previously expected.”

July 4, 2024: OpenText to lay off 1,200

OpenText said it will lay off 1,200 staff, or about 1.7% of its workforce, in a bid to save around $100 million annually. It plans to hire new sales and engineering staff in other areas in 2025, it said.

June 4, 2024: Microsoft lays off staff in Azure division

Microsoft laid off staff in several teams supporting its cloud services, including Azure for Operations and Mission Engineering. The company didn’t say exactly how many staff were leaving.

April 4, 2024: Amazon downsizes AWS in a fresh cost-cutting round

Amazon announced hundreds of layoffs in the sales and marketing teams of its AWS cloud services division — and also in the technology development teams for its physical retail stores, as it stepped back from efforts to generalize the “Just Walk Out” technology built for its Amazon Fresh grocery stores.

April 1, 2024: Dell acknowledges 13,000 job cuts

Dell Technologies’ latest 10K filing with the US Securities and Exchange Commission disclosed that the company had laid off 13,000 employees over the course of the 2023 fiscal year; it characterized the layoffs and other reorganizational moves as cost-cutting measures. “These actions resulted in a reduction in our overall headcount,” the company said. A comparison to the previous year’s 10K filing, performed by The Register, found that Dell employed 133,000 people at that point, compared to 120,000 as of February 2024. Dell announced layoffs of 6,650 staffers on Feb. 6, but it is unclear whether those cuts were reflected in the numbers from this year’s 10K statement.

See news of earlier layoffs.