U.S. national laboratories tackle some of the nation’s most demanding—though often underappreciated—high-stakes innovation projects. This work demands intense concentration from scientists, researchers, and specialists who are developing breakthroughs and solutions to the most pressing threats we face now and in the future.
In these high-pressure environments, every team member contributes to accelerating progress, no matter their specific role. This critical work requires robust support systems.
Idaho National Lab’s CIO Mark Holterman is fully committed, as are his IT team and their partners.
“The key is building strong partnerships with mission teams, making sure they have all the resources they need to address their challenges—essentially removing the burden of operationalizing their proofs of concept and research,” Holterman explained. “On the operations side, maintaining flawless, secure 24/7 operations is our primary focus, and we consistently operate at that pace.”
Holterman readily acknowledges that the speed of operations is rapidly approaching the limits of human-driven IT management. He is a strong advocate for integrating autonomous AI to process increasing volumes of data more quickly and effectively.
Clarifying autonomy and setting shared standards
He is also careful to define what he means by “autonomous”—and what he doesn’t. Human oversight, reviews, and validation remain essential. Fully autonomous? Not yet.
“Automation has existed for some time. However, the tools, the computing power behind them, and their availability have advanced significantly in recent years,” Holterman noted. “In the federal and lab contexts—and this is similar to the commercial sector, depending on the industry—we must validate results. But there is tremendous potential in leveraging rules-based, deterministic, script-driven automation.”
Agentic AI, which he describes as automations triggering other automations and multiple automations working simultaneously, creates high confidence that an agent could perform human checkpoint roles. While technically feasible, he emphasized, it’s not acceptable.
“Checkpoints are still necessary. In our context, we must ensure everything is auditable, traceable, and defensible. Comprehensive logging is essential,” Holterman said. “But it is definitely accelerating our pace of work.”
A clear real-world example is the lab’s work protecting critical infrastructure—an extension of cybersecurity, but distinct.
“We research safeguarding critical infrastructure, and the data from sensors and process logs, the ability to analyze that data and detect, preempt, and prevent incidents is something we also apply in traditional cybersecurity. So it strengthens our defense, while we maintain human oversight,” he said.
This defense is increasingly vital. Malicious actors are developing agentic tools that systematically seek vulnerabilities in U.S. systems, intensifying the AI race and the lab’s pace.
“We’re trying to stay ahead of [zero days] by identifying behavior with large models analyzing data, tasks that would require multiple analysts. Agents can surface these issues faster and detect, prevent, and act,” Holterman said. “That’s the next step: It’s not just alerting but taking action. It could be as simple as quarantining, or as dramatic as shutting down access to an unknown entity causing a problem.”
Setting boundaries: where autonomy ends and human judgment begins
As Holterman pursues an aggressive path with automated AI, he ensures collaboration with experts, researchers, participants, and the broader AI community.
This includes a Center of Excellence for advanced analytics, machine learning, and LLM models for handling large datasets. There’s also a Center of Excellence for intelligent automation—exploring everything from basic automation to advanced AI, with a cross-lab group prioritizing tasks.
“We’ve established a ‘citizen development’ governance structure. It’s a Community of Practice. We have guardrails defined, written out, to keep people within boundaries. Auditing is challenging,” he said. “We want the power, momentum, efficiency, and productivity from quick wins, but it’s difficult for people to know when they’ve crossed certain lines.”
The citizen-development community encourages discussion, problem-solving, and collaboration—even voting—on guidance. It’s a delicate balance, but Holterman believes it’s essential to progress.
“As much as I’d like to be conservative from a security perspective, we want to advance the mission,” he said. “So we’ve got to keep these things moving.”
Copyright
© 2026 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.
