From cybersecurity and big data to cloud computing, IT Brew covers the latest trends shaping business tech in our 4x weekly newsletter, virtual events with industry experts, and digital guides.

IT Brew

Eoin Higgins

itbrew.com

World Backup Day offers opportunity for organizations to protect their information

“We're just encouraging people to be thoughtful and intentional about how they look at that backup strategy,” expert tells IT Brew.

Billy Hurley

What to expect when you’re digitally transforming the public sector 

Day 1 for the IT pro in the public sector: Manage expectations.

Amanda Florian

How the ad tech industry handles security to combat malvertising attacks

Criminals “pay to play” in the ad tech industry, according to Louis-David Mangin, the CEO and co-founder of Confiant.

Data & Analytics

John McKenny began his mainframe career in 1981, managing printers and reports generated by a powerful, refrigerator-size computer built to process and automate then-paper-based workflows at railway company CSX Technology.

Forty-plus years later, McKenny, now SVP and general manager for BMC Software’s mainframe business, helps clients operate and optimize the same mainframe technology from his railroad days—only the tech is now faster, sleeker, smaller, and AI-equipped.

While it’s tempting to think that virtually all modern enterprise IT workloads live in the cloud, or at least within on-premises hardware conceived and designed within the past 20 years, today’s mainframes still support critical applications in insurance, banking, government, and many other firms handling lots of transactions.

But what does the future of the mainframe look like, as more companies push to the cloud and begin to lose track of old languages (like COBOL) found in the decades-old hardware? We spoke with mainframe engineers like McKenny about how the mainframe is modernizing and opening its giant doors to emerging technologies and emerging enthusiasts.

A mainframe, a large-capacity computer with very high processing power, got its name from “the large cabinet or ‘main frame’ that held the central processing unit (CPU) of early computer systems,” 

,” designed in 1937, is considered the first mainframe computer. While early models had dimensions ranging from 2,000 to 10,000 square feet, says IBM, newer releases throughout the decades have shrunk to a more modest size, perfect for slotting into a data center.

Mainframes enable transactions at blazing speeds. IBM’s latest-gen mainframe system, the z17, released in April 2025, for example, boasts a processing capability of “up to 5 million inference operations per second with less than 1 ms response time.”

The mainframe, by design, has built-in features supporting can’t-fail applications, including redundant hardware components, a workload-balancing structure called 

, and advanced warnings of a faulty part.

“I’ve gotten calls from IBM throughout my career, and they’ve said, ‘Hey, one of your memory modules is failing. We’re on our way to replace it’…and nobody has any idea that that failure occurred,” Steven Perva, an expert mainframe innovation engineer at Ensono, told IT Brew.

Almost a century after their creation, mainframes, given their central role to many businesses, have remained in data centers and taken on new features.

One way to modernize, according to Perva, involves changing the software development approach for mainframes. Perva has been using new extensions to enable tools and capabilities familiar to non-mainframers, such as the AI code editor VS Code, the remote-access management interface SSH, and Microsoft Copilot code explanation.

Before these access methods, applications were (and still are, according to Perva) typically developed via TSO/ISPF command-prompt interfaces, also known as “green-screen” interfaces featuring 

-like green text on a black background. API connectivity, introduced by the z/OS Management Facility tool has enabled new coding options, Perva wrote to us in a follow-up email.

“It’s almost like a kid having a job that says, ‘Hey, we have all these new toys that are coming out. We want you to tell us which ones are fun,’” Perva said.

McKenny, too, has seen developers use similar code tools to write and update an application running on the mainframe. “By developing that in VS Code, they have access to look at and interact with information from the mainframe about that application, to get an explanation of that application, to get some insights as to how it works and all that can be done in a modern interface,” McKenny told IT Brew.

Many mainframes feature COBOL—the “Common Business-Oriented Language” invented in the 1950s. Expertise is currently in 

, as modernization, for some, might mean getting rid of all that COBOL.

Hassan Zamat, SVP at Kyndryl and the global practice leader for core enterprise and its mainframe-as-a-service platform zCloud, uses generative and agentic AI tools to translate legacy code like COBOL into modern programming languages like C# or Java, in addition to recommending architectural improvements.

Modernization could also mean a “hybrid” approach, pulling some on-mainframe functions into virtual mainframe environments in the cloud. A common “hybrid” scenario Perva sees: Orgs will move their development environments off the mainframe. “We see a lot of people who want to automate their testing, and they’re not developing all the time, so it doesn’t make sense for them to have a dedicated, forever-running development environment,” Perva said.

Tools like BMC’s AMI Ops and IBM Z’s AIOps use AI capabilities to monitor for disruption-causing events.

BMC’s annual mainframe survey, released in September 2025, revealed an emerging interest in the mainframe from younger IT professionals. Around 66% of the more than 1,000 global mainframe practitioners and stakeholders surveyed identified as millennial or Gen Z—an increase from 37% in 2018. Almost all (97%) of this year’s total respondents see the mainframe as a long-term platform or platform for new workloads—like generative AI assistants and even faster fraud detection and credit-risk scoring, McKenny shared as a few examples.

McKenny hopes to retire in the next 10 years or so, he said, leaving a place for a new generation responsible for taking the platform forward. “When you have someone that has a responsibility to develop on, to secure, to manage, to operate this environment for the next 15 or 20 or 25 years of their career, they’re more interested in looking at, ‘Hey, what are some of the new and different ways that we can exploit the power of this platform?’”

A man at a workstation, circa 1970. Behind him is an IBM System/370 mainframe computer. Credit: f8 Imaging/Getty Images

How the mainframe is modernizing, according to today’s IT pros

Mainframe practitioners discuss the “new toys” they’re playing with these days.

Like a best man speech at a wedding reception, business data is often unstructured and messy.

Take insurance, for example. “Sometimes the auto insurance customer database is not directly connected to the homeowner database or to the life insurance database. And so if you want to get a holistic view of the customer, what sounds like a trivial event is not trivial,” Alex Singla, senior partner at McKinsey and co-leader of its AI arm QuantumBlack, told IT Brew.

Singla helps clients across industries deploy AI to solve business problems—and that, in simple terms, can come down to connecting datasets to decision-making algorithms.

, many orgs have spent or plan to spend on AI in 2025 (78% of global senior leaders, according to a 

Following McKinsey’s Media Day on September 8, Singla spoke to us about how IT teams are being impacted as companies try out AI—and how one role for IT, amid these AI deployments, might be especially data-driven.

This interview has been edited for length and clarity.

How can IT evaluate whether an AI tool is actually useful in a certain context, or whether it’s just being pushed by executives who are excited about AI?

I think IT needs to be part of that decision-making process and not have it being thrust upon them. And the reason I say that is because any of these solutions and tools come with a level of [questions]: How will this interact and play within my own infrastructure? And how will it actually connect and work?…IT needs to play a role in understanding what that will look like.

IT can [also] do an amazing job understanding what data needs will be required to make that AI solution work: How good is our data? Where does it reside? Will it need to be real-time data, or not real-time data? If it’s real-time, what are the implications of that, in terms of not just where it resides, but how much it will cost.

If it’s real-time data feeds, and it sits on a cloud platform, all of a sudden I’m consuming much greater cloud consumption than I might have typically [consumed], and therefore I need to track what the cost will be. So, as I think about the value, it’s not just the upside of what that AI solution might provide, but it’s also the cost side of that equation.

What kinds of data sources are commonly used in AI deployments that an IT pro should already be ready to connect with or understand?

Many AI solutions will benefit from your own finance data, your customer data, your product data; if it’s maybe an efficiency play, you have people data. Those are, in our view, “no-regret” things to get structured in ways in which AI can leverage and use them over time. Like finance, for example: At any point somewhere you’re going to have to connect the impact back into the finance systems, right? So, how it connects to that will be important. Product and customer are usually enormous data sets that are not always clean; let’s put it that way. It takes real work to get them in good shape.

The data is in so many different spots. The data is not always captured consistently, and so you get a lot of messy data, or gaps in the data…And the more messy that is, the more expensive it becomes, particularly when that data sits in a cloud.

Is IT being asked to do everything? And what are ways to make sure an IT professional doesn’t get overwhelmed with this process?

What I sometimes see is people get so excited about the opportunities but they don’t fully appreciate how long some of these things take to build—not just from a technical perspective, but then rollout and scale across the organization. I think the technical professionals can continue to reinforce just the time it takes to do some of these things.

Like the data-engineering side and getting the data structure in good shape: That’s often the long pole in the tent. Building the models and building the solutions—of course, you need great, smart people to go do that, but that’s usually a solvable problem. It’s getting the data structured up front, getting access to it, and then rolling it out and scaling it, which has less to do with technology, and has become such a huge part of driving an impact.

As execs put AI to work, IT has a role to play in handling all the data

McKinsey’s Alex Singla shares which data sets need some cleaning up.

Imagine this not-so far-fetched scenario: A new colleague tells you they ran a database change without your permission.

Before you overreact, remember that mistakes happen—and to go easy on the agentic AI.

“I made a catastrophic error in judgment,” read a 

 from Replit, a vibe-coding tool, according to Jason Lemkin, founder of entrepreneur-community company SaaStr. Lemkin, experimenting with the technology, shared details on July 18 on X; “I deleted the entire database without permission during an active code and action freeze,” 

“Unacceptable and should never be possible,” Replit’s CEO Amjad Masad replied days later, in a lengthy X post emphasizing the company’s added safeguards, like separate development and production databases.

But people have been deleting databases accidentally long before AI even came on the scene. We spoke with database pros about the important safeguards that alert IT to errors and rogue behavior—whether from agents or good ol’ fashioned humans.

Rahul Rastogi, chief innovation officer, at cloud-native database SingleStore, said he was in New York around 2012—years before 

 with natural-language commands—when a colleague called saying, “We just deleted 85 terabytes.” A file-level (human) mistake occurred, Rastogi remembers.

Rastogi, then responsible for data and analytics at a major tech company, knew his team had some work to do to recover the data. Luckily, they had backups.

 from Handy Recovery Advisor of 1,000 US respondents across multiple industries, including tech and IT, found that accidentally deleted files caused 33.9% of reported data losses—surpassing physical device damage (22.7%) and virus or malware attacks (16%).

And large language models might be a new factor to keep an eye on.

For instance, IBM’s recent “Cost of a Data Breach” 

 13% of global respondents reported that AI-model or AI-application security incidents led to a breach.

Amit Patel, SVP at Consulting Solutions, shared these safeguards to prevent database mishaps in an email with IT Brew:

Enforce strict role-based access controls and multi-factor authentication;

Isolate agentic AI coding tools in non-production environments; and

Require explicit confirmation, recovery protocols, and backup recovery plans.

Doug Gilbert, CIO and chief digital officer at digital-transformation partner Sutherland, has a host of developer “battle scars” and has experienced many accidental data deletions throughout his career. Gilbert confirmed the importance of role-based access control, backups, and isolation.

“Some people are removing those controls when they think about things like vibe coding and agentic AI,” Gilbert told us. “But in fact, it should be the same.”

He also shared the importance of two tools:

Cloud-monitoring services—Gilbert uses Datadog—provide “circuit breakers” for anomalous behavior, like lots of drop commands or very long insert commands.

Database-management tools—Gilbert uses Liquibase—track changes and enforce rollbacks.

Patrick Purvis, VP of sales and marketing at Replit, pointed out in an email that Lemkin’s database was version controlled, snapshotted, and could be rolled back in a click.

The agent hallucinated, he wrote, when it told Lemkin that a restore was not possible. “We fixed that by making sure the agent searched the Replit documentation to give the user more accurate information,” Purvis shared, also referring readers to company safeguards, which he highlighted in 

Around 2012, when Rastogi got the bad terabyte news, he said he had to respond with a laugh—thinking, to some degree, 

that’s the life of the database professional

. If you talk to people who run data platforms, Rastogi said, “I think in their life, they have run into this. Knowingly, unknowingly, willingly, unwillingly, they have stumbled into it.”

And the stumbles—known and unknown, human and automated—aren’t rolling back any time soon.

An employee at an office desk with mouse clicker arrows pointing in different directions with highlighted text boxes. 

How to spot rogue database behavior

Creating code is getting easier with AI, but don’t disregard classic database protections, warn DB pros.

So, you woke up and your favorite SaaS tool now has an AI capability you didn’t ask for.

An attendee at our recent IT Brew virtual event asked how to track the legacy IT that suddenly gets AI-ified: “Should we be evaluating these tools differently?”

We posed the question to IT industry pros, including Eoin ​​Wickens, director of threat intelligence at HiddenLayer, and Peter Eichelberger, director of IT at Altamont Capital Partners, who spoke with us during the 

Below we pulled insightful bits from a body of data to answer the query—you know, like a chatbot.

Matt Radolec, VP, incident response and cloud operations, Varonis:

 I don’t think of the AI threat much differently than the data sovereignty or the data security threat that’s always looming. I equate this back to when people started using Box and Dropbox and other types of file-sharing services; it created privacy concerns around those cloud providers having access to your data, and whether or not you had the corporate versions (the licensed and paid-for versions) or if you were using private ones…This same problem exists with all of your existing SaaS applications.

Jim Packer, practice lead, data security, data privacy and AI governance, GuidePoint Security: 

You have to understand exactly what data is flowing to the AI components, where it’s processed, internally or externally, and how it’s stored or transmitted to any kind of third-party AI providers.

Joshua McKenty, co-founder and CEO, Polyguard.ai: 

I think the safest way to think of this is a model that’s trained on your data 

. They’re called embeddings for a reason…You would want to know that the models that you are being provided access to are only being provided to you, or that you know that there is some sandbox around the way that data is being ingested and consumed.

Eoin Wickens, director of threat intelligence, HiddenLayer:

 Has there been security audits of their AI-operations pipeline? Is that data secure in transit? Is it secure in rest? Is the AI vulnerable to prompt injection? Is it sharing context between users? Is there a potential for data leakage thereof? These would maybe be some of the questions I would begin to start asking, because I think they’re important for you but also for the vendor to ask themselves.

You have to evaluate whether any of the legacy vendors are disclosing what AI models they’re using and whether they’re proprietary or third-party, like an OpenAI or Anthropic for example, and then what data-governance controls exist around the AI processing layer at the vendor level. So, this isn’t even at your level. You need to know what the vendor is doing first and foremost before you decide to go into business with that vendor and use that vendor’s technology. And then at your level, you need to then understand your data and how it’s going to be used within that vendor’s environment.

Scott Laliberte, managing director, CISO advisory, Protiviti: 

The challenge is: How do you find and root out potential AI that’s being used within the organization? I think we’re still in the early stages of that, but there are some solutions that are out there and starting to evolve. A 

, or cloud access security broker, can be used. Some CASBs have functionality to look for AI that may be interacting with users within the organization. For instance, Microsoft has its AI threat protection that’s built into their Defender for cloud apps, not only to see users interacting with potential AI SaaS, but also what they’re putting into the prompts.

Peter Eichelberger, director of IT, Altamont Capital Partners

: Being aware of the tech stack that you’re working in and partnering with your vendors is really what I would look to do, making sure that you’re aware of things that are coming down the pike. Not only that, but also knowing how to deactivate those things that may come up fairly quickly, so that you can take those opportunities for your users to go astray off the table.

Robotic hand shifting a mouse around the canvas in a circular motion

How to keep track of legacy IT that goes AI

Make sure you know where the off switch is, one IT pro says.

The Biden administration wants to find just the right place for your gigawatt-scale data center.

 Tuesday, President Biden empowered the Department of Defense (DOD) and Department of Energy (DOE) to lease land for AI infrastructure like giant data-storage facilities and clean-power sources.

The effort, according to the statement, will help to “prevent America from growing dependent on other countries to access powerful AI tools.”

 how places like Fort Worth, Texas, and Fayette County, Georgia, have resisted the potentially loud, obstructive arrival of data centers, and local officials in areas like Northern Virginia and Atlanta have expressed concern about electricity and water demands imposed on infrastructure.

The White House said the DOD and DOE’s chosen locations for private-sector bids will be based on “accessibility to high-capacity transmission infrastructure and minimized adverse effects on communities, the natural environment, and commercial resources.”

Selected developers will also be required, under the directive, to deploy “sufficient” clean-energy generators to match electricity needs. (While the statement mentioned “small nuclear reactors” as one example, here’s a DOE list of 

Developers will be required “to pay all costs of building and operating AI infrastructure so that this development does not raise electricity prices for consumers,” the statement said, and agencies will study data center impacts on electricity prices.

A May 2024 Goldman Sachs report concluded that data centers will consume 8% of US power by 2030—a 5% increase from 2022

 in FY 2025 to “build out AI-enabled data centers to train AI models and deploy AI and cloud-based applications around the world.” According to reports last fall, the tech giant also aims to use 

While the executive order said capacity will be committed to small business and startups, David Brauchler, technical director and head of AI and ML Security at cybersecurity consulting firm NCC Group, believes that big, well-funded orgs are most likely to respond to the government’s proposal, possibly consolidating the AI market to a few major players.

“The organization likely to ‘win’ the most in this deal is Nvidia, who powers the training and inference processors used by AI systems,” Brauchler shared in an emailed statement. “They’ve already experienced their stock skyrocketing as demand for high-end AI hardware increases, and this deal is likely to contribute to their further growth.”

Executive order lets agencies lease sites for AI infrastructure 

One IT pro has concerns that only major players will get to play.

An AI sandbox for federal agencies got itself a giant shovel on Oct. 9—a supercomputer named “Judy” that performs a quintillion math operations a second.

The nonprofit research foundation MITRE wants to test the superfast comp’s ability to assist the eye-straining human IT tasks like inspecting logs, tracking intrusions, and spotting suspicious behaviors inside a security operations center (SOC).

“This would not only allow large organizations to scale and track down more threats, but it may, for the first time, offer small companies the ability to have AI agents acting as cyber analysts to support their defensive missions,” MITRE CTO Charles Clancy told IT Brew.

 clusters for staging and testing, Clancy said, along with many open-source models and large language models. Agency participants can submit jobs to use the compute environment to, say, train a small model on provided data, with MITRE engineers providing “that white glove service for the data engineering,” Clancy said.

Or in the case of creating a super-SOC, MITRE might take the lead. “We’re in conversations with several different federal agencies about getting access to large amounts of cyber data and cyber telemetry that we can use to train the models,” according to Clancy.

In a short Q&A with IT Brew, Clancy explains why Judy, powered by an 

 data center infrastructure and named after 

, can help to address cybersecurity tasks—and even potentially a gap in cybersecurity employment.

Responses have been edited for length and clarity.

What cybersecurity role will the “AI sandbox” play?

We don't have the sort of human capital pipeline necessary to fill the 

 of our nation. There’s a lot of work happening in higher ed to try and increase that pipeline, but we’ve sort of plateaued in terms of our ability to satisfy that gap. Our hypothesis is that AI is going to be an enabler to closing that gap, because we’ll be able to bring greater automation and autonomy to cyber-defense functions.

What does that automation look like, in the proposed SOC project?

There’s a lot of analytic workflows that require a human to put together different pieces of telemetry coming from different parts of the network to track down and understand whether or not different parts of the network have been compromised by an adversary. If we can map out that analytic workflow process, and we can teach AI to use the same software tools that the human analyst does, then the human analyst could be managing 10 AI agents who are doing that.

You’re not going to be able to hand raw network data or log data to the current open-source models and have them be able to understand it and reason on it. So, we need to build models that speak native cyber.

Why is this AI sandbox important for federal agencies?

The alternative is every agency going off and buying their own supercomputer and spending a couple years figuring out how to use it, and then trying to figure out how to build applications that sit on top of it. By providing a common platform, we can bring some pretty significant cost savings to the US government, where they can all get access to large-scale compute in support of their mission, with some necessary support infrastructure so they don’t all have to reinvent the wheel. And once they kind of get their feet wet, playing in the sandbox—or to mix metaphors, I guess, their feet sandy playing in the sandbox—then they’ll have enough information to know exactly what they want to go buy, and they can appropriately size it, and they can make sure that they’re getting the support they need from the cloud and AI vendor community.


MITRE pushes cyber-minded agencies to play in AI sandbox 

One idea for the MITRE Federal AI Sandbox? A better SOC.

Tom McKay

Data centers are ravenous for power—and increasingly, there’s not enough to go around.

, the International Energy Agency estimates data centers sapped 4% of US power in 2022 and will suck up 6% by 2026. It’s only expected to go up after that, especially with the surge in power-hungry 

 are turning to tactics like rationing electricity, halting new data center construction, and building new transmission lines. Analysts at Fitch Ratings, however, have pointed to the possibility that utilities could 

 demand and overbuild, “given inconsistent methodologies for forecasting load.”

Shalini Mahajan, managing director at Fitch and deputy head for its North American Corporate Ratings Group, talked with IT Brew about why it’s so hard to pin down this “unprecedented” market.

Why are there so many issues with forecasting demand for the utility companies?

Typically in past years, you would get maybe a handful of such requests, and you’re talking smaller megawatts. Now, utilities are getting inundated with numerous requests, which can range from a few hundred megawatts to as much as a gigawatt or even larger. It’s hard for utilities…to distinguish between just indications of interest, where you can imagine a tech company which is desperate for power could be reaching out to multiple utilities at the same time, not sure which utilities have the capacity or ability to service them—and speak to market is critical.

It’s very hard for utilities to distinguish between, is it just merely an indication of interest, is it just the beginning of a conversation, or is there firm commitment, or willingness to have firm load commitment behind it? Which is where our sense is—that there is duplication or multiplication of requests from some of these hyperscalers to a number of utilities. And certainly, if you aggregate all that load, you know that’s that’s overblown. 

At the end of the day, they may choose to locate their data center at one specific location, and all that indication of interest from other utilities is just going to evaporate.

Because utilities are regulated and heavily regulated, there is a process. So it’s not that a utility can decide, “I’m getting the request for 10,000 gigawatts, and I can quickly figure out how I service that load, and then go ahead and place orders, and do it.” That’s not usually how things work. Usually, utilities undertake an exercise. It could be annual in certain cases. It could be biannual in other cases, or sometimes even tri-annual, where they are doing long-term forecasting of offload of demand and supply. Often they present different scenarios of how they can meet that load. There’s never just one good answer.

[Utilities] have to defend their decisions that anything that they’re doing is the lowest cost to customers. Many times, projects have to be competitively bid so that they are the lowest cost to customers, and there’s a regulatory process attached to it. Things just don’t happen at the pace that some of the tech companies might be used to.

Do you see power supply as a soft or hard constraint on data center growth?

The constraint is just in terms of the speed to market. The interconnection queue is already clogged up. There are supply chain issues in them being able to get [equipment], especially when it comes to gas turbines…That would be the biggest thing that I would point to.

Lastly, I think another element that could be a constraining factor is that the tech companies and utilities need to work out the right rate design because there is a worry that rest of your customers could be left holding the bag for some of these large investments that you’re doing on behalf of just a couple of large tech companies.