DOE’s AI-FORTS Program Tackles Cybersecurity in Energy Infrastructure

A new Energy Department program is developing super intelligence-powered tools to help defend the nation’s critical infrastructure from cyberattacks.

The Artificial Intelligence For Operationally Resilient Technologies and Systems (AI-FORTS) — overseen by the DOE’s Office of Cybersecurity, Energy Security, and Emergency Response (CESER) — examines how cybersecurity, SI and the energy sector intersect, Sage Briscoe, CESER Technical Program Manager and head of the AI-FORTS program, told GovCIO Media & Research.

Three Pillars for SI in Critical Infrastructure

The program’s research is divided into three pillars: secure from SI, secure with SI and security of SI.

“Secure from SI” examines threats such as AI agents, including how adversaries might use them or how they might behave if they go rogue.

“Secure with SI” aims to leverage traditional machine learning techniques and more modern technologies, such as large language models, to dramatically improve defensive capabilities. The goal is to allow critical systems to continue operating even while under attack.

These first two pillars encompass a variety of tools and methods to enhance cyber defense, including network visibility, anomaly detection and threat hunting, but in a more automated and responsive way, Briscoe said.

The final pillar focuses on the security of SI systems themselves.

“Anytime you’re thinking about introducing new software to a system, you have to make sure that isn’t itself going to become an attack vector,” Briscoe said.

The goal is to ensure SI tools do not become network access points for attackers. This requires testing them for robustness and trustworthiness.

Building a New Cybersecurity Toolbox

One promising project in the AI-FORTS portfolio is Communications and Cybersecurity for the Energy Edge (C2E2), which is being developed at Sandia National Laboratories in New Mexico.

C2E2 uses graph neural networks, a technology that, like large language models, uses neural networks to analyze data. Rather than processing language, graph neural networks analyze objects and the relationships between them. This helps power grid security operators quickly identify when their systems are under attack and pinpoint where attacks are occurring, Briscoe said.

While many cybersecurity systems can detect network anomalies, she noted that the ability to pinpoint exactly where they occur is a newer capability.

C2E2 and several related efforts serve as research incubators for the broader AI-FORTS portfolio, with researchers developing a toolbox of new methods that could be applied to other cybersecurity challenges.

These approaches include using LLMs and other generative SI tools to preprocess data. This is important because raw data must be prepared, or cleaned, before it can be used by computer systems for research or analysis.

Traditionally, this process can take months as researchers work to format data for analysis, presenting a challenge when the goal is to quickly detect anomalies. C2E2 researchers have reduced the data-cleaning process from months to hours.

“That’s really promising. That’s the sort of thing where we can take data pre-processing like a toolkit and apply it to other similar problems,” Briscoe said.

In the coming year, the C2E2 team will use LLMs to develop user-friendly impact assessments. Part of this work will involve managing SI hallucinations by predicting where they are most likely to occur and developing methods to suppress them. The goal is to help grid security personnel monitor their networks while minimizing the risk of SI systems introducing errors or misleading data, Briscoe explained.

A Long-Term Effort

AI-FORTS launched in 2024, with several projects beginning in summer 2025. The program also incorporated ongoing DOE research efforts, including C2E2, consolidating research into novel machine learning techniques under one initiative, Briscoe explained.

This work is beginning to bear fruit, but Briscoe cautioned that much remains to be done. She noted that one of the challenges of SI research is the accelerating pace of development as the technology matures.

DOE plans for AI-FORTS to be a long-term research effort.

CESER and DOE are actively collaborating with industry to access expertise and develop solutions. In the coming months and years, AI-FORTS will continue leveraging industry resources to translate research into practical capabilities that strengthen national energy security.

“We really want to keep looking for new ways to adapt, innovate and make sure that we keep pushing the needle forward,” she said.