AI is making it easier for threat actors to target highly specialized OT systems while also giving defenders new ways to protect systems that have historically been difficult to secure, experts told CRN.
The emergence of ultra-powerful AI cyber models is dramatically lowering the barriers for threat actors to target specialized operational technology (OT) systems, even as it also provides cybersecurity teams with never-before-available capabilities to secure critical infrastructure—overall representing a massive shift in the OT security space, experts told CRN.
For years, the OT systems that are relied upon by critical infrastructure providers—from water treatment plants to manufacturing plants—have presented a major challenge on both the side of defenders and attackers. While the systems are generally too specialized to attract the level of attention that IT security tools receive, OT has also required expertise—and effort—that often discourages hackers from targeting them.
AI is transforming both sides of that equation, according to cybersecurity experts such as Victor Wieczorek, senior vice president for offensive security at Herndon, Va.-based GuidePoint Security.
Attackers are believed to be increasingly using LLMs to understand unfamiliar industrial technology, as well as to look for vulnerabilities in the OT systems that wouldn’t have justified such a level of effort in the past, experts said. Defenders, meanwhile, are able to use the same technologies to develop cyber protections that are better tailored to OT systems—which were often seen as difficult to secure previously due to having such a small installed base, Wieczorek told CRN.
[RELATED: Edge Computing Week 2026]
“There could be just 12 of any given system in the world. And they could be the most critical [components] for a certain municipality,” he said. “And so it’s very hard to develop cybersecurity technology to defend against something that’s such a small footprint and may never exist again.”
Now, however, “with AI, we have the ability to cover those nuanced, niche edge cases,” Wieczorek said. “We can say, ‘Hey, we only have 12 of these [systems]. We need to deploy or customize some protections or solutions to help protect those 12 [systems]. And with AI agents, now we have the scalability to do that.”
The potential is huge for OT systems, which monitor and control physical processes for industrial facilities, utilities and other environments used in critical infrastructure. But realizing that potential will require more than extending IT security tools into industrial environments, according to experts who spoke with CRN.
The reality is that OT professionals must also be given a more central role in deciding how AI is used—in recognition of the fact that there are consequences in interrupting physical processes to address security weaknesses that threat actors can already exploit without such sophisticated new techniques, the experts said.
For GuidePoint Security, No. 32 on CRN’s 2026 Solution Provider 500, addressing those challenges are increasingly a key part of helping customers to navigate the impacts of AI on OT and critical infrastructure security, according to Wieczorek.
“These are human challenges of trust and accountability and responsibility. These are decisions that have to be made, knowing the ramifications,” he said. “There are going to be some human limitations of trust and responsibility that AI is not going to help us with.”
AI-Powered OT Attacks
One of the biggest shifts from AI in the OT security space is that the challenges of understanding and attacking unfamiliar equipment are rapidly decreasing, experts said.
For instance, an attacker interested in a particular organization can use AI to examine firmware or analyze a patch released by a vendor for clues about the vulnerability it addresses, they said.
That changes the calculus for a threat actor, Wieczorek said, since it can now make the attack worthwhile even when the effort has minimal value beyond that one single target.
“Even if it might be throwaway code or effort to go after this organization—and [the attacker] will never be able to use those attack techniques or effort again anywhere else—it doesn’t matter,” he said. “Because now we’re talking [spending] $3 and three minutes with an LLM.”
This shift erodes the “security through obscurity” that many OT systems once possessed, experts said.
Without a doubt, AI is making it possible for hackers to acquire knowledge that, historically, might have required learning directly from an engineer who understood a certain industrial system, according to Harry Thomas, founder and CTO at Charlotte, N.C.-based OT security firm Frenos.
“AI is definitely accelerating the consumption of knowledge [about OT systems],” Thomas said. “Security through obscurity is definitely not something we can hide behind anymore regarding operational technology.”
The implications extend beyond discovering previously unknown vulnerabilities, according to experts.
AI can also help threat actors to assemble documentation and identify existing exploits, as well as to repeat attack techniques that have previously worked elsewhere, said Doc McConnell, head of policy and compliance at Columbus, Ohio-based connected device security vendor Finite State.
In particular, “what AI allows you to do is just scale up and increase your volume and your pace of run-of-the-mill attacks,” said McConnell, a former branch chief for the U.S. Cybersecurity and Infrastructure Security Agency (CISA). “A lot of what we’ve seen in the OT space does not require anything particularly novel.”
Key risks such as internet-exposed equipment, weak credentials and inadequate network segmentation remain central concerns, according to experts. However, AI’s ability to intensify attacks against those weaknesses could put more utilities within reach of cyber adversaries, McConnell said.
There’s no question that AI technology is bringing about a rapid decrease in the expertise required to target industrial processes, according to Samuel Linares, managing director and global OT cybersecurity lead at Dublin, Ireland-based Accenture, No. 1 on CRN’s 2026 Solution Provider 500.
The bottom line is that “with these new models, even someone that doesn’t know anything about OT can still leverage these models to guide them into doing very dangerous things in an OT environment,” Linares said.
A recent incident revealed by industrial cybersecurity vendor Dragos illustrates how that can work in the real world for threat actors.
After compromising the IT environment of a municipal water utility serving Mexico’s Monterrey metropolitan area in January, an attacker then used Anthropic’s Claude platform to investigate the utility’s network, according to Dragos. The model then enabled the threat actor to identify an industrial gateway along with a SCADA/industrial IoT management platform, before then researching how to gain access, Dragos reported.
The attacker used Claude to generate credential lists and attempt an automated password-spraying attack—but those attempts apparently failed, Dragos reported, and no further evidence was found that the attacker had breached the OT environment.
The significance of the incident, according to experts, is that an adversary without demonstrated OT expertise has shown an ability to use AI to go after an industrial target following an IT compromise. Crucially, the activity relied on publicly available attack techniques rather than new types of industrial attack capability, according to Dragos, which is majority-owned by Accenture.
The Mexico incident comes amid a wave of high-profile OT attacks this year, which have demonstrated the potential real-world consequences when critical infrastructure is compromised by threat actors, experts said.
In April and July, U.S. agencies warned that Iran-affiliated threat actors were targeting internet-connected industrial controllers—including across water, energy and local government systems—causing operational disruptions in some cases.
A separate FBI and Environmental Protection Agency alert in July, meanwhile, detailed attacks affecting water and wastewater utilities in at least seven states—with reported effects including pressure loss and flooding.
While the advisories did not establish any involvement of AI, experts said there is every reason to expect the technologies will play a growing role in attacks on OT going forward.
Boosting OT Defense
Nevertheless, the same AI capabilities that are helping attackers to target OT systems also have the potential to help defenders overcome long-standing gaps in protection for industrial systems and critical infrastructure, according to security experts.
For instance, AI agents can make it more practical to customize protections for small numbers of unusual systems, GuidePoint’s Wieczorek said. That can expand the reach of security engineering beyond the types of equipment that are supporting a sizable commercial market, he noted.
Right now, when it comes to OT systems, “people are thinking through how to protect and isolate these systems that before now we just didn’t have the reach to do,” Wieczorek said.
Other opportunities include the potential to analyze industrial security data as well as to perform modeling for attack paths, experts aid.
For Accenture, AI is helping to correlate more data in OT, as well as to accelerate security analysis and response, according to Linares.
In some critical infrastructure deployments, for instance, the company has progressed toward a much greater level of autonomy in security operations for OT, he said.
At Frenos, meanwhile, the company brings together security and network information in a cyber digital twin—which is a digital representation of the environment’s connections and potential attack paths. An AI agent then utilizes that model to emulate how threats might move through the environment, ultimately identifying changes that could block them, Frenos’ Thomas said.
“Our whole goal, aside from trying to identify these attack paths, is to give the risk-reducing recommendations to stop future adversaries from taking advantage of what we found,” he said.
Another opportunity involves using AI to develop protections that are only meant to be temporary while awaiting permanent fixes from vendors, said John Gallagher, vice president of Viakoo Labs at Mountain View, Calif.-based IoT security firm Viakoo.
An environment with equipment from multiple manufacturers, he said, would otherwise often have to wait for each vendor to address a vulnerability. The challenge is providing protection without impacting the stability of the systems that an organization is trying to protect, Gallagher said.
“AI can [help to] create a virtual patch very, very quickly so that the attack could be blunted,” he said.
All in all, “you’ve got concepts driven from AI that actually will lead to better protection of an OT or IoT system as time goes on,” Gallagher said.
Putting OT Experts At The Helm
Ultimately however, the benefits from AI for OT security will greatly depend on who receives responsibility for directing the new technologies—and how well they understand the environment, according to GuidePoint’s Wieczorek.
As is well known, many OT systems were developed during previous eras, and thus were not designed for the cybersecurity challenges of today.
A further issue is that IT professionals inheriting responsibility for these systems may lack a deep understanding of the equipment—crucially including its history, dependencies and operational limitations, Wieczorek said.
Importantly, leveraging an AI tool for guidance does not eliminate that issue, he noted.
The concern, according to Wieczorek, is that IT teams may overestimate how much expertise about OT systems can be provided by LLMs—and may also attempt to deploy familiar IT approaches for protecting OT systems that have very different requirements.
Ultimately, security teams must start with operational continuity and business performance as the key priorities, according to Abel Archundia-Pineda, head of OT/CPS security at Cheseaux, Switzerland-based Kudelski Security.
Taking actions such as deploying firewalls and establishing a joint IT/OT Security Operations Center, for instance, may not deliver their expected benefits without a sufficient level of training and coordination, he said.
“In OT security, you tend to move at the speed of the slowest swim lane,” Archundia-Pineda said.
For solution providers and security vendors, that means that the ideal approach is to design AI capabilities for OT security around legitimate operational technology expertise from the get-go, Wieczorek said.
OT professionals have also not necessarily had the same opportunities as software developers and IT teams to experiment with AI, given that their work continues to be heavily focused around physical processes, he noted.
“The investment hasn’t been there. And certainly the enablement and the training—and really just the focus of the technologies—haven’t panned out to help them,” Wieczorek said. “So now, the time is right. And I’m excited to see that focus change.”
Ultimately, the opportunity is to equip the people who best understand OT systems with AI technology that helps them apply that knowledge more effectively, he said. That could massively improve protection for a class of systems—which powers much of the critical infrastructure in North America—that has long lacked the same level of support for cybersecurity that is available on the IT side.
What is needed, Wieczorek said, is “more OT professionals at the helm, using these cutting-edge tools to help secure us all.”
