In 2026, AI will continue to be a major corporate initiative. But, according to the new Top IT Insights for 2026 report from Unisys, most companies aren’t planning huge, transformational AI integration this year. Instead, many are looking at much smaller implementations, which have quicker pathways to ROI and boost efficiency for existing processes.

“We are going to see more functional deployments of AI, a focus on quality rather than cost-cutting, and the emergence of AI applications that will deliver repeatable, ROI-driven results,” Unisys CEO Mike Thomson said in a release.

Unisys sees three major internal AI applications that companies are likely to develop this year: Enterprise knowledge assistants that index company content and can answer questions, code development agents and AI-powered service operations to help to resolve simpler IT, HR and finance issues. And it’s likely that companies will veer away from using basic, general LLMs in favor of more specific ones—using data that is more customized to their enterprise, industry and way of doing business. There’s less LLM wrangling in this kind of deployment, and companies can concentrate on improving the quality of their AI responses. This can help bring greater ROI through eliminating the need for a human to have to redo the work that AI is supposed to perform.

This also will not be the year of large-scale layoffs due to AI, Unisys predicts. Any company that was once anticipating making broad cuts to human staff, the report states, will learn that layoffs will make innovation more difficult. Instead, job roles are likely to shift more this year, replacing responsibilities that are now done with AI with different ones to further the company’s evolution. The only exception, Unisys writes, is entry-level coders, whose skills will likely not be needed as much.

With AI, going small actually helps companies prepare for larger and more ambitious AI projects in the future, the report states. Smaller projects take less of a financial and personnel investment. It’s also an easier lift for staff and resources to do something that improves what the company already does. But the process of developing smaller and more limited AI deployments brings real learning to the team, increasing their knowledge both of how to make AI platforms work, and what is possible given the data, governance and infrastructure at the company. It also addresses what the real value of AI can be to a company. And with this experience, the years ahead can bring much bolder transformations.

The shady side of the internet known as the dark web has been a go-to hub for things that aren’t quite legal: hackers, illegal trades, hate speech and identify fraud. I talked to Recep Ozdag, former vice president and general manager of cloud, SaaS and cybersecurity for hardware and strategy company Keysight Technologies, about the cybersecurity threats for your business that can be found there. An excerpt from our conversation is later in this newsletter.

Nominations are now open for Forbes AI 50. The eighth-annual list, with sponsoring partner Mayfield, will recognize the most promising startups deploying artificial intelligence in financing, scientific discovery, construction and more.

For a dynamic and innovative company, there are many things that we have come to expect from Nvidia these days. It always beats expectations in quarterly earnings reports; when he’s not at a political function, founder and CEO Jensen Huang always wears a black leather jacket; and the company always has a blockbuster announcement at CES. And last week at CES 2026, a black-leather-clad Huang did not disappoint. In Nvidia’s keynote address, Huang announced an entirely new platform of next generation AI chips and infrastructure named for groundbreaking astrophysicist Vera Rubin. The platform includes the Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU and Spectrum-6 Ethernet Switch.

Forbes contributor Karl Freund distills the major reason this announcement was so groundbreaking: Companies usually develop one component of AI infrastructure at a time. Six new products at once is unprecedented. Forbes contributor Ron Schmelzer added that this is a perception shift at Nvidia in terms of AI development. It’s no longer a single chip, product launch or update, he writes, it’s a full-picture overhaul reshaping the way AI systems are built. The platform represents Nvidia’s newest, fastest, most efficient AI infrastructure, bringing together all the parts that are needed for a reliable and capable AI ecosystem. AI is no longer an addition to the system, it is the system.

But Huang talked about more than new hardware, which is expected to be available in the second half of 2026. Nvidia announced a new partnership with manufacturing giant Siemens, building what they describe as an industrial AI operating system, which will integrate AI across product design, electronics, manufacturing execution and supply chain operations. The first fully AI-driven adaptive Siemens manufacturing site under this partnership is expected to deploy later this year in Germany, writes Forbes senior contributor Janakiram MSV.

Huang also announced Alpamayo, a portfolio of AI models for autonomous vehicles. This platform brings more reasoning skills to the table and is intended to help increase understanding of the complex scenarios that go into human decision making when driving.

We’re just two weeks into 2026, and CrowdStrike has already made two huge acquisitions to bolster its security offerings. Last week, the cybersecurity giant announced it would buy identity security startup SGNL for $740 million. SGNL’s technology brings dynamic identity authorization across different layers of systems for human, non-human and AI agentic users, allowing access to be revoked the moment an identity is compromised.

This week, CrowdStrike announced its intent to buy Seraphic Security in a transaction valued at $420 million. Seraphic is a leader in browser-based security, enforcing additional security measures in major browsers as users move about the internet. Both deals are expected to close this spring.

Forbes senior contributor Tony Bradley talked to CrowdStrike CEO George Kurtz about the deals, which provide more data and control to an enterprise cybersecurity team. “You can solve almost any security use case with the right data,” Kurtz said. “If you can connect the dots, you can understand attack chains.”

Last week, the tech world was stunned by announcements that Apple and Google were actually coming together to make it easier for users to switch between the Android and iPhone ecosystems. This week, the almost shocking degree of cooperation went a step further: Apple is building the next generation of its foundation AI models on Google Gemini’s technology. Forbes senior contributor John Koetsier said in the immediate future, this move will make Apple’s voice assistant Siri much more useful. If you ask Siri a question, Siri will use Gemini to get you the answer. But in the joint statement about the new partnership, Apple said its privacy-focused Apple Intelligence isn’t going anywhere, meaning that Google will not gain access to user data.

This move also lets Apple keep building its own AI systems internally, without having to invest in all of the money and resources to match what other companies have already done, Koetsier writes. It isn’t Apple deciding that it’s getting out of AI development, he writes; the company has throughout its history worked to develop and maintain its own tech stack. Forbes senior contributor David Phelan writes that Apple expert Ming-Chi Kuo at TF International Securities posted on X that the deal came about because of “short-term pressure on multiple fronts,” and will give Apple a chance to develop its own AI chips to enter production later this year.

The dark web is where things like stolen data, fraudulent documents, communication that would be censored and illicit transactions can be found. But it’s also a potential and inexpensive market for hacking tools, which can create major problems for enterprise cybersecurity. Late last year, I talked to Recep Ozdag, the former vice president and general manager of cloud, SaaS and cybersecurity for hardware and strategy company Keysight Technologies, about some of the issues there. He left this position at the beginning of 2026 for other opportunities.

This conversation has been edited for length, clarity and continuity.

How does the dark web figure into what hackers do? What are they finding there?

Ozdag: Back in the day, it was kind of like an alternative website. It’s an alternative network. Combined with some of the VPNs, it basically allows you to anonymously browse and do transactions. What made it extra easy is cryptocurrency, because now with cryptocurrency, not only are you able to browse it, but you can also exchange goods and make it anonymous as well.

What is available right now? It’s an [economy worth a] couple of billion dollars. You can have not only peoples’ information—credit cards, credentials— there are also tools out there that aggregate all that information. You have to buy it, of course. You might have to go to one hacker: ‘There was a hack from Chase. I want to get that. There was another hack from Home Depot. I want to get that.’ And they’ll dump you all the files.

It’s not easy, but there are tools now that will just say, ‘I’m looking for [a person’s] information.’ And then they will sort through millions and try to aggregate that so you can get all that information.

No. 2, there is essentially hacking tools as a service as low as $50 a month. Phishing kits where they will automatically send emails to sound like you’re at work: There’s a FedEx for you. Let me click on it immediately. It’s actually not; it was a phishing attack. Denial of service attacks that you can pay $5 per user or $5 per attack. You can rack up your charges to about $50,000 or $100,000. They even have ransomware as a service. There was a ransom hub group that essentially would deploy malware on your behalf, and they only ask for a couple thousand dollars of payments. There’s spyware software that they will sell, and all sorts of similar hacks: voice mimicking software, video mimicking software, and they will combine it with other applications as well.

It’s all ready for you to use, and you don’t have to be a sophisticated hacker. You can essentially select a target and say, ‘I want you to go after it,’ and they’ll try to do the best that they can, depending on the cost.

It is about a $2 billion to $3 billion a year market. That might not seem like a lot, but if you consider that the cheapest service is about $50, this means that is utilized significantly. This doesn’t account for all the hacks, but it shows that people are going there and paying because maybe they don’t have the talent or it’s easier.

How well are cybersecurity providers fighting this? Are they moving as quickly as the bad attacks are coming? And what’s in the future?

It’s not improving as quickly as one would hope, because there’s asymmetry here. If I’m developing a tool that I want to sell on the dark web, it doesn’t have to always work perfectly. Everything’s a bit shady anyway. You’re not going to be like, ‘I’m going to go to the SEC,’ or ‘I’m going to go to the Better Business Bureau and complain about you. I want to talk to your CEO. I’m going to write a horrible review on Yelp or Glassdoor.’ It’s like, ‘I tried. I spent $1,000, $5,000. It didn’t work. Maybe the tool was bad, or maybe the tool was good, but I just couldn’t hack.’

Now apply that same logic to a publicly traded company who’s trying to develop a product, who says, ‘We can defend you against X, Y and Z.’ You better be able to defend it because if you cannot, then people will ask for their money back, they’ll get bad reviews, all sorts of things. So the problem is that the speed at which hackers can evolve is not the same as companies can develop solutions and release them to the public. They have to test for everything: it’s not supposed to crash your systems, it cannot do this, it cannot do that.

The stakes are much higher. And because of that, it doesn’t develop as quickly as possible. So what does this mean? Enterprises are always going to be a bit more cautious when they’re adopting newer technologies. They’re going to wait for others to validate or verify, or make sure that cybersecurity is covered.

With all that’s going on, what advice would you give to CISO or CIO to stay on top of the threats that are facing them, and how to at least mitigate them with a degree of speed?

There’s definitely companies that are monitoring the dark web and including that in their threat intelligence. When you’re getting threat intel from one company or another, add that. What are they doing in that regard? It has to be a company that’s looking at the dark web. There are now AI tools that will do automatic monitoring for you as well, and it’s not too expensive, [but make sure you can ask them,] ‘What are you seeing in the dark web? What are the newer activities?’

Continue with a zero-trust framework. Zero trust has been out there for a while, but it’s a framework. It’s not a single tool that you buy. It’s kind of like the layers of the onion.

The other piece is: Continue to do segmentation. It goes hand-in-hand with zero trust. The idea is that you segment everything in your network and data center, so even if part of it is hacked, which I’m sure it will be, at least it’s a bit more contained. As opposed to when you invite guests to your house, they could practically go everywhere: kitchen, bathroom, this and that. You want to be able to say, ‘You guys just stay at the kitchen, and if you want to go to the bathroom, I’m going to do an introduction again: Who are you?’

The other piece is that they have to invest in monitoring their own networks, whether they outsource it to professionals, third parties, or they invest in monitoring [themselves]. I know that this might sound simple [and it might feel cost prohibitive], but because the amount of data that’s being generated is increasing, it’s important to do: monitor your cloud deployment, endpoints, data center and network.

Frequent backups. It doesn’t prevent the data being from stolen, but if it’s stolen and you can’t get it back, that’s even worse.

It definitely makes sense to get in touch with those other CISOs and try to share ideas as much as possible. Trying to do everything in a vacuum doesn’t really help. You need to be educated as much as possible.

Unfortunately, there’s no, ‘Hey, if you do this, you’re never going to be hacked.’

It’s a new year, and leaders often make resolutions to do their jobs better. But that mindset can lead to quick burnout because doing “better” often is translated to working more hours and intensity. It’s smarter to look at all that you do, and when you add things, take others away.

Cybersecurity has traditionally moved in terms of the calendar: Weekly scans, monthly reports, quarterly updates. But hackers do what they want, when they want, so it’s time to change up that security schedule.

Which tech company recently hired former Trump Administration national security adviser Dina Powell McCormick as its president?

A. Oracle

B. Google

C. Samsung

D. Meta

See if you got the answer right here.

© Forbes