Diablo Canyon Nuclear Power Plant: Embracing AI While Navigating Safeguards
Diablo Canyon, the last active nuclear power plant in California, has faced uncertainty several times over the past decade and is now set to commence a prolonged decommissioning process in 2029. Despite its precarious situation, the plant in San Luis Obispo recently received significant computing hardware—a deployment of eight NVIDIA H100s, some of the world’s most powerful graphical processors. These are intended to support a new artificial intelligence (AI) tool aimed at revolutionizing operations within the nuclear energy sector.
Partnership Between PG&E and Atomic Canyon: A New Era of AI in Nuclear Energy
Pacific Gas and Electric (PG&E), the operator of Diablo Canyon, has forged a partnership with the San Luis Obispo-based startup, Atomic Canyon. This collaboration marks a milestone as it is hailed as “the first on-site generative AI deployment at a U.S. nuclear power plant.”
Currently, the AI tool, known as Neutron Enterprise, assists employees in navigating extensive technical reports and regulations—millions of pages of intricate documents from the Nuclear Regulatory Commission spanning decades. The introduction of Neutron Enterprise signifies the potential for enhanced utilization of AI in Diablo Canyon and similar facilities, prompting discussions around necessary regulations and safeguards.
The Phased Deployment of Neutron Enterprise
PG&E is rolling out the AI-powered document retrieval service in phases. The installation of NVIDIA chips was part of the initial stages of this collaboration, with a “full deployment” anticipated by Q3 of this year, according to Maureen Zawalick, PG&E’s vice president of business and technical services. At this point, Neutron Enterprise will function as a data-mining assistant, or “copilot,” to streamline access to both external and internal reports specific to the plant.
“We currently spend around 15,000 hours annually searching through various databases and records,” Zawalick stated. “This tool is expected to significantly reduce that time.”
Atomic Canyon’s Vision: Enhancing Nuclear Operations with AI
Trey Lauderdale, CEO and co-founder of Atomic Canyon, explained that the main goal for Neutron Enterprise is to enable Diablo Canyon employees to efficiently access relevant information. “I can assure you there’s no way in hell I want AI running my nuclear power plant right now,” Lauderdale stated, highlighting a cautious approach regarding AI implementation in sensitive environments.
While both PG&E and Atomic Canyon are committed to limited AI applications for the present, plans to expand AI’s role in the future are not entirely off the table. The startup is in discussions with additional nuclear facilities and entities interested in developing small modular reactor plants, exploring various ways to adopt AI technology.
AI Safety Concerns and Regulatory Oversight
The integration of AI into the nuclear sector raises pertinent questions about the adequacy of regulatory measures to govern these powerful technologies. Though the Nuclear Regulatory Commission (NRC) has begun investigating AI in nuclear operations, its ongoing focus under the current administration remains uncertain. After former President Trump rescinded a prior executive order establishing AI regulatory goals, the Atomic Canyon team is proactively engaging with the NRC to keep them informed.
Tamara Kneese, director of the Climate, Technology, and Justice program at the nonprofit Data & Society, commented on the possible benefits of AI for specific tasks while calling for caution. “Using generative AI for one specific task in a nuclear plant and expecting it to remain limited raises concerns. The oversight of PG&E’s AI applications deserves more scrutiny,” she warned.
San Luis Obispo Assemblymember Dawn Addis echoed similar sentiments regarding safety and job implications stemming from AI deployment at Diablo Canyon. “I have numerous questions about the safety, oversight, and job impacts related to AI integration,” she stated, stressing the necessity for robust regulations in critical environments.
The Evolution of Atomic Canyon and AI in Nuclear Energy
Before founding Atomic Canyon, Lauderdale had experience in healthcare technology with Voalte, a company he created to improve hospital communication. After relocating to San Luis Obispo in 2021, he recognized the synergy between his expertise and the nuclear energy field.
Atomic Canyon launched in 2023 with the task of downloading approximately 53 million pages of publicly available Nuclear Regulatory Commission data, a process facilitated by prior communication with the commission for transparency and collaboration. Following the data download, Atomic Canyon partnered with the Department of Energy’s Oak Ridge National Laboratory, using its high-performance supercomputer, Frontier, to enhance AI capabilities for processing nuclear terminology.
Understanding Neutron Enterprise
PG&E installed NVIDIA hardware at Diablo Canyon to operate the advanced AI software, Neutron Enterprise. Zawalick and Lauderdale emphasized that this software is deployed without cloud access to ensure sensitive documents remain secured on-site. Their data policies comply with all Nuclear Regulatory Commission and Department of Energy standards, subject to continuous testing and evaluation.
As of now, Neutron Enterprise serves only to sift through publicly available regulatory documents. PG&E and Atomic Canyon anticipate extending its capabilities further in late 2025, enabling it to search and summarize internal documents using advanced technologies like optical character recognition and retrieval-augmented generation.
Insight from Legislative Bodies and Local Leaders
Responses from state lawmakers and local officials regarding the Neutron Enterprise initiative have generally been favorable, highlighting the current constraints on AI operations. Senator Henry Stern voiced cautious optimism about such technologies, agreeing that their safety protocols must remain paramount. Senator John Laird emphasized the importance of balancing technological advancements with safety and regulatory oversight.
San Luis Obispo County supervisors echoed these sentiments—Supervisor Bruce Gibson endorsed the merits of AI in organizing essential information, while Supervisor Heather Moreno praised the usage of an efficient search tool for information management at the plant.
Looking Ahead: AI’s Future in Nuclear Energy
As Atomic Canyon engages with various nuclear organizations, Lauderdale embraced the opportunity to showcase the potential of AI applications in the field. His vision includes extending the operational lifecycle of Diablo Canyon beyond its planned decommissioning or facilitating efficient data management during that process.
Lauderdale affirmed that their AI deployment maintains minimal operational risks, indicating that if the AI fails to retrieve required information, employees can revert to traditional search methods. Moving forward, the focus will remain on expanding capabilities cautiously while emphasizing safety and transparency throughout the process.
Frequently Asked Questions (FAQs) About AI in Nuclear Power Plants
What is Neutron Enterprise and how does it work?
Neutron Enterprise is an AI tool designed to assist nuclear plant employees by effectively searching and summarizing extensive technical documents and regulations. It currently focuses on publicly available data, with plans for future enhancements to access internal documents.
How does AI technology enhance safety in nuclear facilities?
AI can improve efficiency in information retrieval, reducing time spent on manual searches, which can enhance operational safety by allowing staff to focus on critical tasks more effectively.
What are potential risks associated with using AI in nuclear energy?
Concerns include unintended errors in AI outputs and the need for thorough regulatory oversight to ensure the safe application of AI technology in sensitive environments like nuclear power plants.
Is there a regulatory framework for AI in nuclear energy?
The Nuclear Regulatory Commission is exploring AI integration, but existing regulations still lack specific guidelines for AI use in the nuclear industry, highlighting the necessity for developed standards.
What future roles could AI play in the nuclear sector?
While immediate uses are focused on document retrieval, there are ongoing discussions around the potential for AI to take on more complex tasks, provided that stringent safeguards are implemented.