Sec. 101. Artificial Intelligence Safety Institute
1,525 words·~7 min read·
/bill/118/s/4178/rs/section-101A research copy — for the controlling text, always check the official state or federal source. Not legal advice.
Not later than 1 year after the date of the enactment of this Act, the Under Secretary of Commerce for Standards and Technology (in this section referred to as the Under Secretary ) shall establish an institute on artificial intelligence. The institute established pursuant to paragraph
(1)shall be known as the Artificial Intelligence Safety Institute (in this section referred to as the Institute ). The mission of the Institute is as follows: To assist the private sector and agencies in developing voluntary best practices for the robust assessment of artificial intelligence systems. To provide technical assistance for the adoption and use of artificial intelligence across the Federal Government to improve the quality of government services. To develop guidelines, methodologies, and best practices to promote— development and adoption of voluntary, consensus-based technical standards or industry standards; long-term advancements in artificial intelligence technologies; and innovation in the artificial intelligence industry by ensuring that companies of all sizes can succeed and thrive. The Under Secretary shall appoint a director of the Institute, who shall be known as the Director of the Artificial Intelligence Safety Institute (in this section referred to as the Director ) and report directly to the Under Secretary. The Director may hire such full-time employees as the Director considers appropriate to assist the Director in carrying out the functions of the Institute. In addition to making appointments under paragraph
(1)of this subsection, the Director, in coordination with the Secretary of Commerce, may make appointments of scientific, engineering, and professional personnel, and fix their basic pay, under subsection
(b)of section 6 of the National Institute of Standards and Technology Act ( 15 U.S.C. 275 ) to hire critical technical experts. Such subsection is amended, in the second sentence, by striking 15 and inserting 30 . Subsection
(c)of such section is amended by striking the date that is 5 years after the date of the enactment of this section and inserting December 30, 2035 . The Director may enter into such agreements, including contracts, grants, cooperative agreements, and other transactions, as the Director considers necessary to carry out the functions of the Institute and on such terms as the Under Secretary considers appropriate. In establishing the Institute, the Under Secretary shall— coordinate with— the Secretary of Energy; the Secretary of Homeland Security; the Secretary of Defense; the Director of the National Science Foundation; and the Director of the Office of Science and Technology Policy; and consult with the heads of such other Federal agencies as the Under Secretary considers appropriate. The functions of the Institute, which the Institute shall carry out in coordination with the laboratories of the National Institute of Standards and Technology, are as follows: The following functions relating to research, evaluation, testing, and standards: Conducting measurement research into system and model safety, validity and reliability, security, capabilities and limitations, explainability, interpretability, and privacy. Working with the Department of Energy, the National Science Foundation, public-private partnerships, including the Artificial Intelligence Safety Institute Consortium established under subsection (f), and other private sector organizations to develop testing environments and perform regular benchmarking and capability evaluations, including artificial intelligence red-teaming as the Director considers appropriate. Working with consensus-based, open, and transparent standards development organizations
(SDOs)and relevant industry, Federal laboratories, civil society, and academic institutions to advance development and adoption of clear, implementable, technically sound, and technology-neutral voluntary standards and guidelines that incorporate appropriate variations in approach depending on the size of the entity, the potential risks and potential benefits of the artificial intelligence system, and the role of the entity (such as developer, deployer, or user) relating to artificial intelligence systems. Building upon the Artificial Intelligence Risk Management Framework to incorporate guidelines on generative artificial intelligence systems. Developing a companion resource to the Secure Software Development Framework to incorporate secure development practices for generative artificial intelligence and for foundation models. Developing and publishing cybersecurity tools, methodologies, best practices, voluntary guidelines, and other supporting information to assist persons who maintain systems used to create or train artificial intelligence models to discover and mitigate vulnerabilities and attacks. Coordinating or developing guidelines, metrics, benchmarks, and methodologies for evaluating artificial intelligence systems, including the following: Cataloging existing artificial intelligence metrics, benchmarks, and evaluation methodologies used in industry and academia. Testing and validating the efficacy of existing metrics, benchmarks, and evaluations, as well as TEVV tools and products. Funding and facilitating research and other activities in a transparent manner, including at institutions of higher education and other nonprofit and private sector partners, to evaluate, develop, or improve TEVV capabilities, with rigorous scientific merit, for artificial intelligence systems. Evaluating foundation models for their potential effect in downstream systems, such as when retrained or fine-tuned. Coordinating with counterpart institutions of international partners and allies to promote global interoperability in the development of research, evaluation, testing, and standards relating to artificial intelligence. Developing tools, methodologies, best practices, and voluntary guidelines for identifying vulnerabilities in foundation models. Developing tools, methodologies, best practices, and voluntary guidelines for relevant agencies to track incidents resulting in harm caused by artificial intelligence systems. The following functions relating to implementation: Using publicly available and voluntarily provided information, conducting evaluations to assess the impacts of artificial intelligence systems, and developing guidelines and practices for safe development, deployment, and use of artificial intelligence technology. Aligning capability evaluation and red-teaming guidelines and benchmarks, sharing best practices, and coordinating on building testbeds and test environments with allies of the United States and international partners and allies. Coordinating vulnerability and incident data sharing with international partners and allies. Integrating appropriate testing capabilities and infrastructure for testing of models and systems. Establishing blue-teaming capabilities to develop mitigation approaches and partner with industry to address risks and negative impacts. Developing voluntary guidelines on— detecting synthetic content, authenticating content and tracking of the provenance of content, labeling original and synthetic content, such as by watermarking, and evaluating software and systems relating to detection and labeling of synthetic content; ensuring artificial intelligence systems do not violate privacy rights or other rights; and transparency documentation of artificial intelligence datasets and artificial intelligence models. Coordinating with relevant agencies to develop or support, as the heads of the agencies determine appropriate, sector- and application-specific profiles of the Artificial Intelligence Risk Management Framework for different use cases, integrating end-user experience and on-going development work into a continuously evolving toolkit. The following functions relating to operations and engagement: Managing the work of the Institute, developing internal processes, and ensuring that the Institute meets applicable goals and targets. Engaging with the private sector to promote innovation and competitiveness. Engaging with international standards organizations, multilateral organizations, and similar institutes among allies and partners. Not later than 180 days after the date of the enactment of this Act, the Under Secretary shall establish a consortium of stakeholders from academic or research communities, Federal laboratories, private industry, including companies of all sizes with different roles in the use of artificial intelligence systems, including developers, deployers, and users, and civil society with expertise in matters relating to artificial intelligence to support the Institute in carrying out the functions set forth under subsection (e). The consortium established pursuant to subparagraph
(A)shall be known as the Artificial Intelligence Safety Institute Consortium . The Under Secretary, acting through the Director, shall consult with the consortium established under this subsection not less frequently than quarterly. Not later than 2 years after the date of the enactment of this Act, the Director of the National Institute of Standards and Technology shall submit to the Committee on Commerce, Science, and Transportation of the Senate and the Committee on Science, Space, and Technology of the House of Representatives a report summarizing the contributions of the members of the consortium established under this subsection in support the efforts of the Institute. In carrying out the Institute functions required by subsection (a), the Under Secretary shall support and contribute to the development of voluntary, consensus-based technical standards for testing artificial intelligence system components, including, as the Under Secretary considers appropriate, the following: Physical infrastructure for training or developing artificial intelligence models and systems, including cloud infrastructure. Physical infrastructure for operating artificial intelligence systems, including cloud infrastructure. Data for training artificial intelligence models. Data for evaluating the functionality and trustworthiness of trained artificial intelligence models and systems. Trained or partially trained artificial intelligence models and any resulting software systems or products. The Director may seek, accept, hold, administer, and use gifts from public and private sources whenever the Director determines it would be in the interest of the United States to do so. The Director, in consultation with the Director of the Office of Government Ethics, shall ensure that authority under this subsection is exercised consistent with all relevant ethical constraints and principles, including— the avoidance of any prohibited conflict of interest or appearance of impropriety; and a prohibition against the acceptance of a gift from a foreign government or an agent of a foreign government. Nothing in this section shall be construed to provide the Director of the National Institute of Standards and Technology any enforcement authority that was not in effect on the day before the date of the enactment of this Act.
Connectionstraces to 1
Traces to 1 document
Citation graph
cites case law
Sec. 101
Artificial Intelligence Safety Institute
Cites 1Cited by 0 across 0 sources