The Department of Energy’s new Genesis Mission is being presented as a major national effort to use artificial intelligence to speed up “national security work”. The White House launched Genesis in November 2025, describing it as a coordinated federal effort to build an integrated AI platform using federal scientific datasets, national laboratory resources, supercomputers, universities, industry partners, production plants, and national security sites. The executive order explicitly frames the effort as a race for technological dominance and compares its urgency and ambition to the Manhattan Project.
Lawrence Livermore National Laboratory (“LLNL”) is expected to play a significant role in this new AI mission. The Lab’s work is almost entirely for the DOE’s subautonomous National Nuclear Security Administration, which maintains and develops nuclear weapons. For fiscal year 2027, roughly 89% of the LLNL’s requested budget is for nuclear weapons activities, so it naturally follows that the Lab’s AI work is mostly related to nuclear weapons activities as well.
LLNL’s work includes operating major high-performance ”super” computing systems (some of the most powerful in the world) that are used to compute data for developing new nuclear weapons in the absence of explosive nuclear testing. They conduct “stockpile stewardship” work i.e. maintaining the existing nuclear stockpile and ensuring its safety, security and reliability. The Lab supports materials and high-explosives research including supporting new plutonium pit production at Los Alamos and Savannah River and developing new and novel high explosives used in nuclear weapons detonation. The lab houses the National Ignition Facility, which is mostly used to create temperatures and densities that mimic an exploding nuclear weapon to conduct experiments in service of nuclear weapons activities. Because of this combination of classified datasets, experimental facilities, and computing capacity used to both maintain and further nuclear weapons activities for the US stockpile, LLNL’s role in Genesis deserves close public attention.
The scale of LLNL’s involvement is now becoming clearer. In July 2026, LLNL announced that it had been selected to lead 10 Phase I projects under the Genesis Mission and participate in 19 additional projects led by other institutions. The LLNL-led projects include using AI for nuclear and particle physics, laser-plasma experiments, precision manufacturing of fusion targets, materials development, high-performance computing, autonomous experimental systems, and quantum technologies. Several of these areas overlap with the facilities and technical capabilities that support LLNL’s nuclear weapons mission, including the National Ignition Facility, advanced manufacturing, high-energy-density experiments, and weapons-related computing. Even projects described as basic science or energy research may strengthen the same personnel, facilities, models, and computing infrastructure used for nuclear weapons activities.
The announcement also leaves major questions unanswered. LLNL does not disclose how much funding each project may receive, which facilities will be used, whether the work will require new equipment or construction, or how it could affect energy consumption, water use, hazardous-material handling, experimental activity, or waste generation. Before these projects move forward, DOE and LLNL should disclose their scope, funding, locations, participating contractors and companies, infrastructure requirements, and connections to classified or nuclear weapons programs.
DOE describes Genesis as a project that would bring together huge amounts of scientific information, powerful computers, AI tools, and lab experiments in one connected system. Some of the work DOE has described is directly connected to the nuclear weapons complex. This includes using AI to find and test new materials for nuclear weapons programs, helping nuclear facilities run more experiments, turning old nuclear weapons records into searchable digital files, and using AI to help with safety paperwork and production planning.
These areas raise important oversight questions. AI tools may be used to speed up research, sort through large datasets, guide experiments, generate simulations, assist with safety documentation, or support decisions about materials and production processes. In the nuclear weapons complex, these are not ordinary administrative tasks. They can affect worker safety, facility operations, environmental risk, emergency planning, and the pace of weapons modernization.
For worker safety, the risk is that AI could be used to speed up safety reviews or work plans without fully understanding the real conditions inside a facility. At a place like LLNL, safety decisions can involve hazardous materials, radiation controls, high explosives, lasers, contaminated buildings, ventilation systems, fire risks, seismic risks, and aging infrastructure. If an AI tool misses a hazard, relies on outdated records, summarizes a requirement incorrectly, or produces an answer that sounds more certain than it is, workers could be given incomplete protections. In ordinary office work, a bad AI summary may be annoying. In a high-hazard nuclear facility, a bad summary could affect what protective equipment is used, what controls are required, whether a job is allowed to proceed, or whether an accident scenario is taken seriously.
For facility operations, the concern is that AI could be used to make work move faster than the facility can safely support. DOE has described AI tools that could help plan and schedule experiments, steer experiments in real time, combine live data with simulations, and reduce turnaround time between tests. That could create pressure to run more experiments, move more materials, process more work orders, or reduce downtime. If the model is wrong, incomplete, or optimized mainly for speed, it may overlook practical limits that facility workers understand: equipment maintenance, staffing, training, waste handling, emergency access, aging systems, or conflicting activities happening in the same area.
For environmental risk, AI could accelerate decisions about new materials, production methods, experiments, and facility changes before their environmental consequences are fully understood. If Genesis helps LLNL or NNSA qualify weapons-related materials faster, plan more experimental campaigns, or change production workflows, that could also affect waste streams, emissions, energy use, water use, hazardous material handling, and cleanup priorities. The risk is that many small technical decisions could be sped up and treated as internal program choices, even when they have environmental consequences for surrounding communities.
There is also the environmental burden of AI itself. Large AI systems require major computing infrastructure, including supercomputers, data centers, cooling systems, electricity, backup power, and specialized hardware. At a site like LLNL, expanded AI work could increase energy demand, water use for cooling, construction needs, electronic waste, and the overall environmental footprint of weapons-related research. DOE and LLNL should be clear about whether Genesis will require new or expanded computing infrastructure, how much additional energy and water it may use, and whether those impacts will be reviewed publicly.
For emergency planning, A might underestimate a release, misread changing wind conditions, delay a warning, or make it hard for responders to explain why a decision was made. Emergency planning needs clear lines of authority and understandable assumptions. If AI makes emergency decisions less transparent, that is a public safety concern.
For weapons modernization, the concern is that AI could help the nuclear weapons complex move faster while public oversight stays the same or becomes even harder. LLNL and NNSA have already linked AI and Genesis to major parts of nuclear weapons work, including maintaining the nuclear stockpile, making advanced materials and parts, and developing future weapons capabilities. Environmental review, cleanup oversight, public meetings, and access to information are already slow and limited. If AI helps the weapons complex move faster, but communities still have the same limited ability to understand or challenge what is happening, then more nuclear weapons work could happen with less meaningful public scrutiny.
We are concerned whether AI will be used in ways that make nuclear weapons work less transparent, less understandable to the public, or more difficult to independently review. Traditional public oversight depends on access to environmental documents, safety analyses, regulatory records, cleanup reports, meeting minutes, and agency explanations. AI-assisted work may increasingly rely on classified datasets, proprietary software, internal model outputs, contractor systems, and cross-site digital platforms that are not visible through normal public review channels.
DOE and LLNL have acknowledged that AI systems require human oversight and careful validation. Those commitments need to be made specific. In high-consequence nuclear work, human oversight should mean more than a scientist or manager reviewing a model output. It should include clear standards for when AI can and cannot be used, independent verification and validation, documentation of uncertainty, and accountability if an AI-assisted recommendation is wrong.
For communities near LLNL, Genesis also raises practical site-specific questions. Will the initiative require new computing infrastructure at LLNL? Will it increase electricity demand, cooling needs, water use, construction, or hazardous materials activities? Will any LLNL facilities require safety-basis changes because of Genesis-related work? Will Genesis affect cleanup priorities, environmental monitoring, or emergency preparedness? These are basic questions for a community that has lived for decades with the consequences of contamination and federal secrecy.
TVC will continue tracking Genesis as an emerging oversight issue. The public needs clear information about how AI will be used at LLNL. AI should not be allowed to move high-consequence nuclear weapons decisions further out of public view.