The Nuclear Regulatory Commission (NRC) is an independent agency of the U.S. government tasked with protecting public health and safety related to nuclear energy.
Created by the Congress back in 1974, it is meant to ensure the safe use of radioactive materials to benefit civilians, while protecting people and the environment from radioactive hazards.
To do this, it functions include overseeing reactor safety and security, administering reactor licensing and renewal, licensing radioactive materials, radionuclide safety, and managing the storage, security, recycling, and disposal of spent fuel. The NRC also regulates commercial nuclear power plants and other uses of nuclear materials, such as in nuclear medicine, through licensing, inspection and enforcement of its requirements.
In all, the NRC regulates the operation of 93 commercial nuclear power reactors that generate electricity.
In the era where practically everything can be connected to the internet, the NRC wants outside researchers to determine whether artificial intelligence (AI) and machine learning (ML) can be used to detect and respond to cyberattacks at nuclear power plants.
NRC plans to regulate AI and ML tools, but before that, it wants to know if the technologies can identify abnormal activity resulting from cyberattacks on increasingly complex nuclear power plant systems.
In this case, the NRC wants outside researchers to determine how AI and ML can detect and respond to cyberattacks at nuclear power plants.
“The vendor must provide and use an existing experimental infrastructure (e.g. personnel, equipment, facilities) to conduct research and implement a test case,” reads the notice that was issued on February 2.
“The research conducted by the vendor is expected to produce data that evaluates the impacts of AI/ML concepts, technologies and applications on nuclear power cybersecurity outcomes and programs — especially those outcomes and programs that may be relevant to new and advanced reactor designs.”
NRC intends to select one of the vendors, that would later choose a single technology to evaluate for accuracy and reliability, need and availability of ML training data, utility, and risks.
That vendor would then be tasked to create a technical report with the test case results.
According to the NRC, the idea vendor should be able to simulate nuclear plant systems, including how they integrate with operational technology (OT); measure the consequences of cyberattacks on them; develop ML and AI tools for detecting system anomalies from cyberattacks, and do so within 16 months of a contract award.
The selected vendor should also provide all necessary research personnel who understand how nuclear plants work, have expertise in plant cybersecurity.
Realizing that AI and ML show promise in improving nuclear reactor safety while offering economic savings, to get a better understanding of current usage and future trends in the technology in commercial nuclear power industry, the NRC has been seeking comments from the public, the nuclear industry, and other stakeholders, as well as other interested individuals and organizations.
Previously, the NRC has held a series of workshops to provide a forum for the NRC, nuclear industry and stakeholders to discuss the state of knowledge and research activities related to data science and AI and their application in the nuclear industry.
The NRC wanted to learn more about the "state of practice, benefits, and future trends related to the advanced computational tools and techniques in predictive reliability and predictive safety assessments in the commercial nuclear power industry."
Questions include, and not limited to, understanding the status of the nuclear power industry’s development or use of AI/ML tools to improve aspects of nuclear plant design, operations or maintenance, or decommissioning, the cost, which areas of operation and management will benefit the most, the potential of using AI/ML in automation for efficiency, design, maintenance and more, and concerns about data security.
Before AI and ML can be implemented, the NRC must first know "potential pitfalls and challenges associated with their application."
