AI Deployed For The First Time In Nuclear Plant
For the first time, artificial intelligence is being introduced at a U.S. nuclear facility Diablo Canyon, California’s last operating nuclear power plant. Despite its scheduled decommissioning in 2029, the site has recently adopted cutting-edge computing technology: eight NVIDIA H100 GPUs. These Nuclear powerful processors now run Neutron Enterprise, a generative AI system developed by San Luis Obispo–based startup Atomic Canyon.
The AI’s primary role, for now, is to assist workers in navigating millions of regulatory documents from the Nuclear Regulatory Commission (NRC). By helping employees locate and summarize technical instructions more efficiently, Neutron Enterprise aims to save thousands of hours previously spent combing through databases and physical files.
PG&E, the utility operating Diablo Canyon, emphasized that the AI tool will act only as a “copilot,” not a decision-maker. Its deployment is being carried out in stages, with full rollout expected by the third quarter of 2025. The AI will remain on-site without cloud access, ensuring sensitive internal data stays secure. This careful approach has reassured some officials, although others remain cautious about long-term implications.
Balancing Innovation and Safety
While the AI’s current application is limited to document searches, experts and lawmakers have voiced concerns about the potential for expanded roles. Some fear a gradual shift toward deeper AI integration, possibly without adequate oversight. Trey Lauderdale, Atomic Canyon’s founder, agrees that AI should not control critical nuclear functions, at least not yet. However, he also acknowledges that the technology’s future in the nuclear sector is still unfolding.
To build trust, Atomic Canyon has kept the NRC informed and collaborated with Oak Ridge National Laboratory during development. Their open-source AI models were trained using over 50 million public NRC documents, enabling them to interpret complex nuclear terminology. According to Lauderdale, this context-driven approach reduces the risk of AI “hallucinations” or misinformation.
So far, Diablo Canyon’s use of Neutron Enterprise has been met with cautious optimism. Local officials appreciate the efficiency benefits but emphasize the need for transparency and strict regulations. Notably, the state-appointed Diablo Canyon Independent Safety Committee has found no safety issues with the current implementation.
Future Prospects and Oversight
Although PG&E and Atomic Canyon have committed to a limited scope for now, wider adoption could be on the horizon. Lauderdale is already in talks with other nuclear facilities and modular reactor projects. He views Diablo Canyon as a testing ground for broader applications, including helping decommission or extend plant lifespans.
Meanwhile, regulatory bodies are grappling with how to keep pace. A recent NRC report highlighted numerous gaps in AI governance and recommended the development of new guidelines. Yet with recent changes in federal policy, the future of AI regulation remains uncertain.
In this rapidly evolving landscape, one thing is clear: while AI’s role in nuclear energy may be modest today, the conversation around its future is just beginning.
