These techniques are how the nuclear sector effectively manages the 10,000 metric tons of used fuel waste created by reactors, which contribute to 10% of global electricity annually. However, the introduction of new nuclear designs might complicate the management of nuclear waste further.

Most currently functioning reactors at nuclear facilities adhere to a similar fundamental design: They utilize low-enriched uranium as fuel and are cooled by water, typically situated in large central power plants. Nevertheless, a wide array of innovative reactor designs anticipated to come online in the near future will likely necessitate adjustments to guarantee that existing systems can accommodate their waste.

“There is no singular solution regarding whether this array of new reactors and fuel varieties will simplify waste management,” states Edwin Lyman, director of nuclear power safety at the Union of Concerned Scientists.

A guide for nuclear waste disposal

Nuclear waste is generally categorized into two main types: low-level waste, which includes contaminated protective gear from medical facilities and research institutions, and high-level waste, which demands more meticulous handling.

The overwhelming majority by volume is low-level waste. This type of waste can be stored on site and often can be disposed of like regular trash once its radioactivity has diminished sufficiently (with some extra precautions). Conversely, high-level waste is significantly more radioactive and tends to be very hot. This category primarily encompasses spent fuel, a mix of materials including uranium-235, which is the fissile element of nuclear fuel—the part necessary to maintain the chain reaction in nuclear power plants. The material also comprises fission products—the occasionally radioactive remnants resulting from atom splitting that release energy.

Many experts concur that the most effective long-term resolution for managing spent fuel and other high-level nuclear waste is a geological repository—a deep, meticulously managed hole in the earth. Finland is leading the way in the construction of such a site, with its facility on the southwest coast expected to be operational this year.

In the 1980s, the US identified a location for a geological repository; however, political disputes have hindered progress. As it stands, in the US, used fuel is stored on-site at both functioning and decommissioned nuclear power plants. After being removed from a reactor, it is usually placed in wet storage, essentially immersed in water pools to cool it down. This waste can subsequently be transferred into protective cement and steel containers known as dry casks, a phase referred to as dry storage.

Experts suggest that the industry will not be required to entirely overhaul this strategy for the newer reactor models. 

“Our management of spent fuel will largely remain the same,” states Erik Cothron, research and strategy manager at the Nuclear Innovation Alliance, a nonprofit think tank dedicated to the nuclear sector. “I don’t lose sleep over how we’ll manage spent fuel.”

However, new designs and materials might call for some engineering adaptations. Moreover, there is a vast diversity in reactor designs, leading to an equally broad range of potential waste types that must be managed.

Uncommon waste

Some emerging nuclear reactors will appear quite akin to existing models, so their spent fuel will be handled similarly to current protocols. Yet, others utilize innovative materials for coolants and fuels. 

“Uncommon materials will result in unusual waste,” mentions Syed Bahauddin Alam, an assistant professor of nuclear, plasma, and radiological engineering at the University of Illinois Urbana-Champaign.

Certain advanced designs may lead to a rise in the quantity of material classified as high-level waste. For instance, reactors implementing TRISO (tri-structural isotropic) fuel contain a uranium core surrounded by multiple protective layers and embedded within graphite shells. The graphite encasing TRISO will likely be combined with the rest of the spent fuel, resulting in waste that is much bulkier than conventional fuel.

Currently, isolating those layers would be complicated and costly, as indicated by a 2024 report from the Nuclear Innovation Alliance. Hence, the entire assembly would be categorized as high-level waste. 

The firm X-energy is developing high-temperature gas-cooled reactors that employ TRISO fuel. It has already submitted plans for spent fuel management to the Nuclear Regulatory Commission, which regulates US reactors. The form of the fuel could aid in waste management: The protective shells incorporated in TRISO eliminate X-energy’s reliance on wet storage, facilitating dry storage from the outset, according to the company.

Liquid-fueled molten-salt reactors, another novel category, may also elevate waste volume. Within these designs, fuel and coolant are not separated as they are in the majority of reactors; instead, the fuel is directly dissolved into a molten salt that serves as the coolant. This implies that the entire batch of molten salt would need to be managed as high-level waste.

Conversely, some other reactor designs might yield a reduced quantity of spent fuel, but this does not automatically translate to a smaller issue. For instance, fast reactors achieve a greater burn-up, consuming a larger proportion of the fissile material and extracting more energy from their fuel. Consequently, spent fuel originating from these reactors generally possesses a higher concentration of fission products and generates more heat. This heat can significantly impact the development of waste solutions.

Spent fuel must remain relatively cool to prevent melting and the release of dangerous by-products. Excessive heat in a repository could also damage the surrounding geological formations. “Heat is a crucial factor in determining how much material can be stored in a repository,” asserts Paul Dickman, a former Department of Energy and NRC official.

Certain spent fuel may necessitate chemical processing prior to disposal, notes Allison MacFarlane, director of the school of public policy and global affairs at the University of British Columbia and a previous chair of the NRC. This could complicate and increase costs.

In fast reactors that utilize sodium metal for cooling, the coolant can infiltrate the fuel and bond to its casing. Separation may present challenges, and sodium is highly reactive with water, requiring specialized treatment for the spent fuel.

TerraPower’s Natrium reactor, a sodium fast reactor that secured a construction permit from the NRC earlier in March, is engineered to effectively address this challenge, according to Jeffrey Miller, senior vice president for business development at TerraPower. The company has devised a strategy to blow nitrogen over the material before it is transferred into wet storage pools, thus eliminating the sodium.

Proximity, Proximity, Proximity

Regardless of the materials employed, even minor changes in reactor size and their placement could complicate waste management processes.

Some new reactors serve as essentially smaller versions of the large reactors presently in use. These small modular reactors and microreactors may produce waste that can be managed similarly to that of today’s conventional reactors. However, in locations like the US, where waste is stored on-site, it would be impractical to have numerous small sites each containing its own waste. 

Some companies are exploring the option of sending their microreactors and the waste they generate back to a centralized location, potentially at the same site where the reactors are manufactured.

Operators should be mandated to meticulously consider waste management concerns and incorporate protocols in their designs, and they should be accountable for the waste they generate, asserts UBC’s MacFarlane.

She also emphasizes that planning for waste has so far relied on research and modeling, with actual conditions clarifying only once the reactors become operational. As she states: “These reactors are not yet in existence, so we lack extensive, detailed knowledge regarding the waste they will produce.”

Training AI models with classified data is anticipated to enhance their precision and efficiency in certain functions, according to a U.S. defense official who spoke on condition of anonymity to MIT Technology Review. This development comes as the demand for more capable models surges: The Pentagon has made arrangements with OpenAI and Elon Musk’s xAI for the operation of their models in classified environments and is executing a new strategy to transform into an “AI-first” combat force as tensions with Iran intensify. (As of publication time, the Pentagon had not commented on its AI training initiatives.)

Training would occur in a secure data center accredited to accommodate classified governmental projects, where a duplicated version of an AI model is linked with classified data, as per two individuals familiar with the operational procedures. While the Department of Defense would maintain ownership of the data, personnel from AI companies might, in rare instances, access the data if they possess the necessary security clearance, the official stated. 

Prior to sanctioning this new training, the official indicated, the Pentagon plans to assess the accuracy and efficiency of models trained on unclassified data, such as commercially available satellite imagery. 

The military has historically employed computer vision models, an earlier type of AI, to detect objects in images and footage collected from drones and aircraft, while federal agencies have awarded contracts for companies to train AI models on such materials. Furthermore, AI firms developing large language models (LLMs) and chatbots have produced iterations of their models optimized for governmental functions, including Anthropic’s Claude Gov, designed to work across multiple languages and in secure settings. The official’s remarks, however, represent the first sign that AI companies developing LLMs, such as OpenAI and xAI, might train government-specific versions of their models directly using classified data.

Aalok Mehta, director of the Wadhwani AI Center at the Center for Strategic and International Studies and former head of AI policy at Google and OpenAI, asserts that training on classified data, rather than merely answering queries about it, could pose new risks. 

The foremost risk, according to him, is that classified information which these models are trained on might be revealed to anyone utilizing the model. This would pose challenges if numerous different military branches, each with varying classification levels and information access requirements, were to utilize the same AI. 

“You can envision, for instance, a model with access to sensitive human intelligence—like an operative’s identity—leaking that information to a segment of the Defense Department not authorized to access it,” Mehta explains. This could create a security threat for the operative, one that is hard to completely mitigate if a specific model is employed by multiple groups within the military.

Nonetheless, Mehta mentions that it’s not overly challenging to confine information from being accessed by the wider public: “If you configure this correctly, there will be minimal risk of that data appearing on the general internet or reverting back to OpenAI.” The government possesses some of the necessary infrastructure already; the security firm Palantir has secured significant contracts for creating a secure environment through which officials can inquire AI models about classified issues without relaying the information to AI companies. However, utilizing these systems for training represents a novel challenge. 

The Pentagon, prompted by a memo from Defense Secretary Pete Hegseth in January, has been hastily working to integrate more AI. It has been employed in combat situations, where generative AI has compiled lists of targets and suggested which ones to prioritize, as well as in more bureaucratic capacities, such as drafting contracts and reports.

There are numerous functions currently conducted by human analysts that the military may wish to train prominent AI models to execute, which would necessitate access to classified information, Mehta notes. This could include learning to identify subtle indications in an image in the same manner an analyst would or linking new data with historical context. The classified data could be sourced from the immense volumes of text, audio, images, and videos, in various languages, that intelligence agencies gather. 

It’s exceedingly challenging to identify which precise military tasks would require AI models to train on such data, Mehta warns, “because undoubtedly the Defense Department has substantial incentives to keep that data confidential, and they don’t want rival countries to be aware of the exact capabilities we possess in that domain.”

If you possess information regarding the military’s application of AI, you can confidentially share it via Signal (username jamesodonnell.22).