Artificial Intelligence Research, Innovation, and Accountability Act of 2024

This bill establishes a framework for artificial intelligence (AI) innovation and accountability in the United States. It amends open data policies, directs research on content authenticity and provenance standards, and requires the development of standards for detecting AI-generated media and anomalous AI behavior. It also mandates a study on barriers and best practices for using AI in government. The bill creates accountability measures for high-impact and critical-impact AI systems, including transparency reports, risk management assessments, certification, and enforcement through civil penalties. It also establishes a working group for AI consumer education.

Title I, Artificial Intelligence Research and Innovation: This title amends the definition of 'data asset' to include 'data model' and defines 'artificial intelligence system' in federal law. It directs the Under Secretary of Commerce for Standards and Technology to research and promote standards for digital content authenticity and provenance, including methods for watermarking and labeling AI-generated content. It also amends the National Institute of Standards and Technology Act to include standards for detecting AI-generated outputs and anomalous AI behavior. Finally, it requires the Comptroller General to study statutory and policy barriers to AI use in government and identify best practices for its adoption.

Title II, Artificial Intelligence Accountability: This title defines key terms like 'artificial intelligence system' and 'high-impact artificial intelligence system'. It requires online platforms using generative AI to notify users. Deployers of high-impact AI systems must submit annual transparency reports to the Secretary. The National Institute of Standards and Technology must develop guidelines for federal agencies to oversee high-impact AI systems. The Office of Management and Budget and covered agencies must create and report on oversight plans. Critical-impact AI organizations must perform risk management assessments before deployment and biennially. The Secretary must establish an advisory committee for certification standards of critical-impact AI systems. The Secretary can enforce the act with civil penalties up to $300,000 or twice the value of the AI system, and can prohibit deployment for intentional violations. Entities that are both developers and deployers must follow both sets of rules. A working group on AI consumer education must be established. A severability clause ensures that if any provision is found unconstitutional, the rest of the title remains in effect.