The Great American AI Act, explained

Establishes the short title of the bill as the Great American Artificial Intelligence Act and provides the table of contents.

Defines key terms used throughout the Act, including acceptable levels of catastrophic risk mitigation, artificial intelligence, artificial intelligence models, catastrophic risk, critical safety incident, foundation model, frontier AI framework, frontier developer, frontier model, independent verification organization (IVO), large frontier developer, material modification, model weight, and substantial modification.

Formally establishes CAISI within the Department of Commerce, directed by a Secretary appointed Director. CAISI will develop voluntary guidelines, best practices, and standards for AI security (adversarial robustness, interpretability, supply chain threats, model tampering, etc.), evaluate AI systems and monitor AI progress, support synthetic content detection tools, and administer the IVO licensing regime. $100 million per year authorized for 2027-2029, with additional fee authority from IVOs and large frontier developers. Grants special authorities for CAISI to hire critical technical experts and fix their pay at levels above the GS scale.

Large frontier developers (>$500M in revenue) must write, implement, comply with, and publicly post a frontier AI framework covering risk thresholds and assessment procedures for catastrophic risk, model weight cybersecurity, and both internal and external deployment decisions. Before, or concurrently with, deploying any new frontier model, frontier developers must publish a report disclosing the release date, supported languages, output modalities, intended use, restrictions, risk assessments, and mitigation steps. Redactions are permitted to protect trade secrets, cybersecurity, public safety, or national security. Additionally, developers must file a report with CAISI within 15 days of a critical safety incident or within 24 hours if the incident poses an imminent risk of death or serious injury. State attorneys general may also opt to receive such reports. Violations can result in fines of up to $1M per day, and the federal and state AGs can seek injunctions.

CAISI licenses IVOs to audit large frontier developers' compliance with Secs. 111 and to assess whether developers' AI frameworks achieve acceptable levels of catastrophic risk mitigation. Licenses may be revoked for material failures of independence, obsolete methods, or non- compliance. Large frontier developers must retain a licensed IVO to verify compliance with their framework and to ensure the adequacy of the framework and procedures. These IVOs must be granted sufficient access to company materials, and report their audits to CAISI. IVOs are immune from claims of loss from an AI model they have audited. State AGs may opt into receiving all audit and assessment reports from IVOs. IVOs may refer violations of this section to the AG, or to state AGs who have opted into receiving such referrals.

Prohibits AI companies from discriminating or retaliating in any way against employees or independent contractors for lawfully reporting a violation of federal AI law. If violated, employees are entitled to reinstatement, two times back pay with interest, compensatory damages, litigation costs, and attorney's fees.

Preempts any state or local law or regulation specifically targeting the development of AI models. Expressly does not preempt laws of general applicability, common law remedies, or laws regulating AI use or deployment. Sunsets three years post-enactment.

Requires the Comptroller General to submit a report identifying federal statutes and regulations that directly affect AI innovation or unduly burden AI infrastructure (including energy), evaluating federal AI adoption progress, and providing legislative and administrative recommendations.

NIST must establish a pilot program to create a template for documenting AI models and associated data (covering model name, developer identity, release date, training data cutoff, supported languages, terms of service, etc.), with technical guidelines and objective performance metrics.

Amends the federal mail fraud, wire fraud, bank fraud, and money laundering statutes to increase maximum fines from $1 million to $2 million and increase penalties when AI is used.

Adds penalties when AI is used to impersonate government officials.

Directs the Secretary of Commerce to conduct a study examining how the government encourages or pressures AI companies with respect to content moderation or limiting of expression. The report must include recommendations for legislation providing individuals redress against unlawful government censorship of AI platforms, including transparency requirements and oversight mechanisms.

Directs the NSF Director to support the National STEM Teacher Corps pilot and Computer Science for All programs, consistent with recommendations of the House AI Task Force.

Authorizes NSF to make competitive grants to universities and nonprofits for research on AI literacy curriculum, educator professional development, and evaluation tools for K-12 AI literacy.

Directs NSF to establish a grant program for institutions not among the top 100 in federal R&D expenditures to broaden participation in AI research, education, and workforce development.

Authorizes the NSF to award scholarships and fellowships for students in AI-related programs.

Excellence. Directs NSF to designate up to 8 regionally diverse "Centers of AI Excellence" at community colleges and career technical education schools. Centers must integrate AI into teaching and workforce development, develop best practices, identify career pathways, and facilitate private sector partnerships.

Authorizes NSF to award grants to research AI teaching tools, materials, and integration into K-12 classrooms, with emphasis on low-income, rural, and Tribal student populations. Also establishes a Rural and Low-Income Areas AI Collaborative to create regional peer-support networks for educators.

Adds AI skills development to the National STEM Teacher Corps pilot program and directs consideration of developing AI best practices for high school teachers.

The Secretary of Labor must publish a request for public comment on implementing the Act's data collection, forecasting, and workforce tools. Additionally, the Secretary must convene an initial expert workshop; subsequent annual workshops are required. The workshop must include economists, AI technical experts, industry representatives, labor organizations, and government officials with diverse viewpoints, and must critically evaluate BLS assumptions about AI's workforce impact.

Authorizes the Secretary of Labor to appoint up to 20 AI experts to excepted service positions with GS-15 pay.

Requires the Secretary of Labor to establish an AI Workforce Research Hub, in collaboration with the Census Bureau, Bureau of Economic Analysis, and Bureau of Labor Statistics. The Hub conducts recurring analyses of AI's workforce impact, conducts scenario planning, and generates actionable policy insights.

Establishes a pilot program producing statistics on job-to-job worker flows for at least 15 AI-sensitive occupations designated by the Secretary of Labor.

NIST must launch at least one prize competition to develop reproducible benchmarks for measuring AI's ability to automate or augment tasks/occupations. NIST may also award companion grants for benchmark design and validation.

Requires BLS to establish a voluntary program for AI developers and deployers to share anonymized data on AI adoption in the workforce.

Within one year, the Census Bureau and BLS must revise specified federal surveys (including the Annual Business Survey, Current Population Survey, Occupational Requirements Survey, and American Time Use Survey) to incorporate or improve questions on AI adoption and use, focusing on types of AI, occupational impacts, skill changes, and work outcomes.

Directs the Secretary of Labor to identify standardized data elements for workforce reporting under the Act, report to Congress within 12 months, and lead a voluntary consensus effort to develop federal (and state/local) standards for producing trusted AI-related data.

Amends the Worker Adjustment and Retraining Notification (WARN) Act to require that whenever AI is a "substantial factor" in a qualifying mass layoff, employers' 60-day advance notices must include a statement specifying that AI contributed to the layoff, describing the type and usage of AI involved, estimating the percentage of job losses attributable to AI, and detailing any pre-layoff upskilling or retraining efforts.

Requires the Secretary of Labor to designate at least 15 AI-sensitive occupations every two years and publish annual prediction-interval employment forecasts (2-, 4-, and 8-year horizons) for those occupations, covering the 20th-80th percentile range of projected employment. Forecasts must be accompanied by benchmark comparisons, transparency on methods and data gaps, and periodic accuracy evaluations. A public, machine-readable archive of all forecasts and scores must be maintained.

Directs NSF to establish a recurring prize competition (every 6 months) for accurate short-horizon forecasts and informative rationales on AI labor-market questions (e.g., model benchmark performance, AI adoption rates, occupation- level employment changes).

Within 2 years, the Secretary of Labor must report to Congress on how AI-related data, benchmarks, and forecasts developed under the Act will be incorporated into selection and performance measurement criteria for WIOA grants, apprenticeship programs, and other Labor Department grant programs.

Within 12 months, the Secretary of Labor must complete a study (directly or by grant) examining design options for a Rapid AI Adjustment Assistance Program for workers displaced by AI, drawing on trade adjustment assistance precedents. The study must address eligibility determination, types of assistance, program costs, evidence collection, and interoperability with existing law.

For five years after enactment, states and local workforce boards receiving WIOA Title I-B funds must consider AI-related forecasts and data produced under Subtitle B when updating their in-demand industry and occupation lists. The Secretary must provide technical assistance.

The AI Workforce Research Hub, in consultation with BLS, must publish a public report inventorying existing federal authorities for technology-driven workforce disruption, identifying gaps relative to AI disruption's pace, assessing economic and tax policy mechanisms (unemployment insurance, portable benefits, tax code provisions), and presenting a range of policy options drawn from domestic and international precedents.

Extends the Cybersecurity Act of 2015 from 2025 to 2035, allowing cybersecurity information to be shared between companies without anti-trust concerns.

Authorizes CISA (in consultation with CAISI) to award grants to eligible maintainers of widely used, critical open-source software for security improvements (patching, maintenance, security audits). Large frontier developers must also provide AI model access to eligible maintainers for cybersecurity purposes.

Requires GAO to report on: security protocols protecting AI model weights, whether existing protections are sufficient, and the security situation of the open-source software ecosystem (resource adequacy of maintainers, infrastructure stability, and supply chain vulnerability).

The Secretary of Energy and the NIST Under Secretary, in coordination with NSF, must establish a testbed program to facilitate collaboration between National Labs, federal labs, NIST, NAIRR, and public and private sector entities for testing, evaluating, and assessing AI systems. Activities include developing automated and reproducible evaluations, prioritizing security vulnerability assessments, and organizing hackathons.

Requires the Secretary of Commerce to prevent duplication between Subtitle A activities and existing DOE and private sector work. National Lab resources used under the program must be provided on a reimbursable basis, unless waived by the Secretary of Commerce.

Requires the Secretary of Energy and NIST Under Secretary to jointly lead federal engagement in international AI technical standards development and form coalitions with like-minded governments to promote private sector-led standards adoption, advocate for U.S.-developed AI standards internationally, counter foreign adversary influence, facilitate cross-border AI R&D collaboration, and share cybersecurity best practices. Criteria for coalition membership include high scientific and technological advancement and adherence to WTO-compatible standards principles. China is expressly excluded from coalition eligibility unless USTR certifies WTO compliance.

Amends the NAII Act to direct OSTP to develop a prioritized list of federal data and datasets for public release to support AI training and evaluation, with a focus on publicly funded research data, data advancing novel AI systems in the public interest, and datasets unlikely to be created without federal investment.

Amends the NAII Act to establish a prize competition program administered by OSTP for AI research and development challenges across priority areas, including chip design, AI interpretability and explainability, advanced manufacturing, border security, AI for science, cybersecurity, energy efficiency, and modernizing federal legacy code..

Formally establishes the NAIRR in statute. NAIRR would procure and provide resources such as datasets, training programs, and computational resources for AI development. These would be available to researchers, institutions, government entities, and small private sector entities.

Requires GAO to conduct a comprehensive review of liquid cooling for AI data centers with input from an advisory committee of industry, national labs, and academic experts.

All activities under the Act must comply with the research security requirements of the Research and Development, Competition, and Innovation Act and the NDAA.

Before any non-federal temporary fellow (contractor, consultant, or fellow) performs AI-related work for an agency, both the fellow and the agency head must sign a certification that the fellow will not perform inherently governmental functions.