[Congressional Bills 118th Congress] [From the U.S. Government Publishing Office] [S. 4664 Introduced in Senate (IS)] <DOC> 118th CONGRESS 2d Session S. 4664 To require the Secretary of Energy to establish a program to promote the use of artificial intelligence to support the missions of the Department of Energy, and for other purposes. _______________________________________________________________________ IN THE SENATE OF THE UNITED STATES July 10, 2024 Mr. Manchin (for himself and Ms. Murkowski) introduced the following bill; which was read twice and referred to the Committee on Energy and Natural Resources _______________________________________________________________________ A BILL To require the Secretary of Energy to establish a program to promote the use of artificial intelligence to support the missions of the Department of Energy, and for other purposes. Be it enacted by the Senate and House of Representatives of the United States of America in Congress assembled, SECTION 1. SHORT TITLE. This Act may be cited as the ``Department of Energy AI Act''. SEC. 2. FINDINGS. Congress finds that-- (1) the Department has a leading role to play in making the most of the potential of artificial intelligence to advance the missions of the Department relating to national security, science, and energy (including critical materials); (2) the 17 National Laboratories employ over 40,000 scientists, engineers, and researchers with decades of experience developing world-leading advanced computational algorithms, computer science research, experimentation, and applications in machine learning that underlie artificial intelligence; (3) the NNSA manages the Stockpile Stewardship Program established under section 4201 of the Atomic Energy Defense Act (50 U.S.C. 2521), which includes the Advanced Simulation and Computing program, that provides critical classified and unclassified computing capabilities to sustain the nuclear stockpile of the United States; (4) for decades, the Department has led the world in the design, construction, and operation of the preeminent high- performance computing systems of the United States, which benefit the scientific and economic competitiveness of the United States across many sectors, including energy, critical materials, biotechnology, and national security; (5) across the network of 34 user facilities of the Department, scientists generate tremendous volumes of high- quality open data across diverse research areas, while the NNSA has always generated the foremost datasets in the world on nuclear deterrence and strategic weapons; (6) the unrivaled quantity and quality of open and classified scientific datasets of the Department is a unique asset to rapidly develop frontier AI models; (7) the Department already develops cutting-edge AI models to execute the broad mission of the Department, including AI models of the Department that are used to forecast disease transmission for COVID-19, and address critical material issues and emerging nuclear security missions; (8) the AI capabilities of the Department will underpin and jumpstart a dedicated, focused, and centralized AI program; and (9) under section 4.1(b) of Executive Order 14110 (88 Fed. Reg. 75191 (November 1, 2023)) (relating to the safe, secure, and trustworthy development and use of artificial intelligence), the Secretary is tasked to lead development in testbeds, national security protections, and assessment of artificial intelligence applications. SEC. 3. DEFINITIONS. In this Act: (1) AI; artificial intelligence.--The terms ``AI'' and ``artificial intelligence'' have the meaning given the term ``artificial intelligence'' in section 5002 of the National Artificial Intelligence Initiative Act of 2020 (15 U.S.C. 9401). (2) Alignment.--The term ``alignment'' means a field of AI safety research that aims to make AI systems behave in line with human intentions. (3) Department.--The term ``Department'' means the Department of Energy, including the NNSA. (4) Foundation model.--The term ``foundation model'' means an AI model that-- (A) is trained on broad data; (B) generally uses self-supervision; (C) contains at least tens of billions of parameters; and (D) is applicable across a wide range of contexts; and (E) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that pose a serious risk to the security, national economic security, or national public health or safety of the United States. (5) Frontier ai.-- (A) In general.--The term ``frontier AI'' means the leading edge of AI research that remains unexplored and is considered to be the most challenging, including models-- (i) that exceed the capabilities currently present in the most advanced existing models; and (ii) many of which perform a wide variety of tasks. (B) Inclusion.--The term ``frontier AI'' includes AI models with more than 1,000,000,000,000 parameters. (6) National laboratory.--The term ``National Laboratory'' has the meaning given the term in section 2 of the Energy Policy Act of 2005 (42 U.S.C. 15801). (7) NNSA.--The term ``NNSA'' means the National Nuclear Security Administration. (8) Secretary.--The term ``Secretary'' means the Secretary of Energy. (9) Testbed.--The term ``testbed'' means any platform, facility, or environment that enables the testing and evaluation of scientific theories and new technologies, including hardware, software, or field environments in which structured frameworks can be implemented to conduct tests to assess the performance, reliability, safety, and security of a wide range of items, including prototypes, systems, applications, AI models, instruments, computational tools, devices, and other technological innovations. SEC. 4. ARTIFICIAL INTELLIGENCE RESEARCH TO DEPLOYMENT. (a) Program To Develop and Deploy Frontiers in Artificial Intelligence for Science, Security, and Technology (FASST).-- (1) Establishment.--Not later than 180 days after the date of enactment of this Act, the Secretary shall establish a centralized AI program to carry out research on the development and deployment of advanced artificial intelligence capabilities for the missions of the Department (referred to in this subsection as the ``program''), consistent with the program established under section 5501 of the William M. (Mac) Thornberry National Defense Authorization Act for Fiscal Year 2021 (15 U.S.C. 9461). (2) Program components.-- (A) In general.--The program shall advance and support diverse activities that include the following components: (i) Aggregation, curation, and distribution of AI training datasets. (ii) Development and deployment of next- generation computing platforms and infrastructure. (iii) Development and deployment of safe and trustworthy AI models and systems. (iv) Tuning and adaptation of AI models and systems for pressing scientific, energy, and national security applications. (B) Aggregation, curation, and distribution of ai training datasets.--In carrying out the component of the program described in subparagraph (A)(i), the Secretary shall develop methods, platforms, protocols, and other tools required for efficient, safe, and effective aggregation, generation, curation, and distribution of AI training datasets, including-- (i) assembling, aggregating, and curating large-scale training data for advanced AI, including outputs from research programs of the Department and other open science data, with the goal of developing comprehensive scientific AI training databases and testing and validation data; (ii) developing and executing appropriate data management plan for the ethical, responsible, and secure use of classified and unclassified scientific data; (iii) identifying, curating, and safely distributing, as appropriate based on the application-- (I) scientific and experimental Departmental datasets; and (II) sponsored research activities that are needed for the training of foundation and adapted downstream AI models; and (iv) partnering with stakeholders to curate critical datasets that reside outside the Department but are determined to be critical to optimizing the capabilities of open-science AI foundation models, national security AI foundation models, and other AI technologies developed under the program. (C) Development and deployment of next-generation computing platforms and infrastructure.--In carrying out the component of the program described in subparagraph (A)(ii), the Secretary shall-- (i) develop early-stage AI testbeds to test and evaluate new software, hardware, algorithms, and other AI-based technologies and applications; (ii) develop and deploy new energy- efficient AI computing hardware and software infrastructure necessary for developing and deploying trustworthy frontier AI systems that leverage the high-performance computing capabilities of the Department and the National Laboratories; (iii) facilitate the development and deployment of unclassified and classified high- performance computing systems and AI platforms through Department-owned infrastructure data and computing facilities; (iv) procure high-performance computing and other resources necessary for developing, training, evaluating, and deploying AI foundation models and AI technologies; and (v) use appropriate supplier screening tools available through the Department to ensure that procurements under clause (iv) are from trusted suppliers. (D) Development and deployment of safe and trustworthy ai models and systems.--In carrying out the component of the program described in subparagraph (A)(iii), not later than 3 years after the date of enactment of this Act, the Secretary shall-- (i) develop innovative concepts and applied mathematics, computer science, engineering, and other science disciplines needed for frontier AI; (ii) develop best-in-class AI foundation models and other AI technologies for open- science and national security applications; (iii) research and deploy counter- adversarial artificial intelligence solutions to predict, prevent, mitigate, and respond to threats to critical infrastructure, energy security, and nuclear nonproliferation, and biological and chemical threats; (iv) establish crosscutting research efforts on AI risks, reliability, safety, trustworthiness, and alignment, including the creation of unclassified and classified data platforms across the Department; and (v) develop capabilities needed to ensure the safe and responsible implementation of AI in the private and public sectors that-- (I) may be readily applied across Federal agencies and private entities to ensure that open-science models are released responsibly, securely, and in the national interest; and (II) ensure that classified national security models are secure, responsibly managed, and safely implemented in the national interest. (E) Tuning and adaptation of ai models and systems for pressing scientific and national security applications.--In carrying out the component of the program described in subparagraph (A)(iv), the Secretary shall-- (i) use AI foundation models and other AI technologies to develop a multitude of tuned and adapted downstream models to solve pressing scientific, energy, and national security challenges; (ii) carry out joint work, including public-private partnerships, and cooperative research projects with industry, including end user companies, hardware systems vendors, and AI software companies, to advance AI technologies relevant to the missions of the Department; (iii) form partnerships with other Federal agencies, institutions of higher education, and international organizations aligned with the interests of the United States to advance frontier AI systems development and deployment; and (iv) increase research experiences and workforce development, including training for undergraduate and graduate students in frontier AI for science, energy, and national security. (3) Strategic plan.--In carrying out the program, the Secretary shall develop a strategic plan with specific short- term and long-term goals and resource needs to advance applications in AI for science, energy, and national security to support the missions of the Department, consistent with-- (A) the 2023 National Laboratory workshop report entitled ``Advanced Research Directions on AI for Science, Energy, and Security''; and (B) the 2024 National Laboratory workshop report entitled ``AI for Energy''. (b) AI Research and Development Centers.-- (1) In general.--As part of the program established under subsection (a), the Secretary shall select, on a competitive, merit-reviewed basis, National Laboratories to establish and operate not fewer than 8 multidisciplinary AI Research and Development Centers (referred to in this subsection as ``Centers'')-- (A) to accelerate the safe and trustworthy deployment of AI for science, energy, and national security missions; (B) to demonstrate the use of AI in addressing key challenge problems of national interest in science, energy, and national security; and (C) to maintain the competitive advantage of the United States in AI. (2) Focus.--Each Center shall bring together diverse teams from National Laboratories, academia, and industry to collaboratively and concurrently deploy hardware, software, numerical methods, data, algorithms, and applications for AI and ensure that the frontier AI research of the Department is well-suited for key Department missions, including by using existing and emerging computing systems to the maximum extent practicable. (3) Administration.-- (A) National laboratory.--Each Center shall be established as part of a National Laboratory. (B) Application.--To be eligible for selection to establish and operate a Center under paragraph (1), a National Laboratory shall submit to the Secretary an application at such time, in such manner, and containing such information as the Secretary may require. (C) Director.--Each Center shall be headed by a Director, who shall be the Chief Executive Officer of the Center and an employee of the National Laboratory described in subparagraph (A), and responsible for-- (i) successful execution of the goals of the Center; and (ii) coordinating with other Centers. (D) Technical roadmap.--In support of the strategic plan developed under subsection (a)(3), each Center