ANS Annual Conference session focuses on AI

The final executive session of the American Nuclear Society's 2026 Annual Conference was titled, "With Great Processing Power Comes Great Responsibility: Ethics and Opportunities for AI and Tomorrow's Workforce."

That session was hosted by ANS Cares, a committee that works to support expanded participation and access to resources across the Society. It focused on how the nuclear industry should work to meet the unprecedented moment that AI is bringing to everything from plant operations to workforce development.

Starting off:The session was moderated by Kathryn Huff, the incoming chair of the Nuclear Engineering and Engineering Physics Department at the University of Wisconsin-Madison. After each panelist gave a brief introduction, Huff began the Q&A by asking how each speaker's respective organizations are specifically rolling out or already using AI.

Jordan Tyman, the director of risk and compliance at PG&E, explained that utilities like PG&E, "the end-user of AI applications," are focused on a knowledge-based approach to implementation and "getting it right the first time." As such, the company "started off a little bit boring by creating a search tool" in partnership with AtomicCanyon that was based on the Nuclear Regulatory Commission's database of documents. Looking to the future, Tyman said PG&E has several plans for AI development, highlighting the company's aim to use it as a tool that helps get new entrants into the nuclear workforce up to speed quickly-potentially filling the training gap left by industry veterans retiring before they have the opportunity to train their replacements.

Matthew Dennis, a senior data scientist at the NRC's Office of Nuclear Regulatory Research who leads the regulator's AI regulatory preparedness and decision-making initiatives, said, "For us, this probably shouldn't come as a surprise, but [the] regulator was very cautious in adopting AI." He added that he has "been surprised at how judicious industry has been about adopting AI in nuclear safety applications," as well. While the NRC has yet to "mine" the ADAMS database, it has gained a license for its staff to use Copilot, which has been widely adopted and broadly used for internal NRC work.

Other than Copilot, Dennis also highlighted an NRC-specific tool called SimplifAI. This tool, he explained, "is regulator specific," meaning "the temperature is turned way down." For example, a user "can't put a request in to get poetry"; it is grounded in NRC documentation and is tailored to specific roles within the NRC staff. Ultimately, Dennis said, the NRC plans to continue to be "cautious and slow to adopt the technology and use it in a way that we feel is hopefully safe and effective for our regulatory mandate."

Shifting focus: Huff then turned to Peter Suyderhoud, the scientific computing & AI deputy division director at Idaho National Laboratory, to ask a bigger-picture question: "Where are we going in the longer term? . . . What are the aspirational uses of AI in nuclear?" Suyderhoud responded that the ultimate goal is "to accelerate the laborious, manually intensive engineering processes and workflows that we see today." He emphasized that "we're not trying to replace the nuclear engineer; we're trying to empower them to move a little bit faster," which he acknowledged as a goal that was both ambitious and nebulous. While an endgame of autonomous plant operation is in the minds of some developers, Suyderhoud said he did not know how soon that scale of AI implementation could realistically be achieved.

Huff then asked Katlyn Turner, an incoming professor in the Department of Information Science at the University of Colorado-Boulder, what it means to develop a responsible workforce in the context of generative AI. Turner responded that the key skill that students must develop before entering the workforce is critical thinking-which is not a skill that she believes AI helps develop.

Huff, adding her own perspective as a fellow educator, pointed out that unobserved assessments have become, essentially, "trash," due to the inability to discern whether that assessment was completed by a student or by AI. Taken together, their thoughts pointed to a general air of concern over students developing the requisite skills necessary to be equipped for the nuclear industry.