This bill would create standards and rules about the origins and authenticity of digital media, especially media made or changed by algorithms and artificial intelligence. The Department of Commerce (through the Under Secretary for Standards and Technology) would run a public-private partnership to develop voluntary, consensus-based standards for content provenance information, watermarking, and tools that detect synthetic or synthetically-modified content. The Under Secretary would also fund research and a public education campaign and coordinate prizes or challenges to improve detection and security.
Two years after the bill becomes law, companies that commercially offer tools whose main purpose is to create synthetic content or to create or significantly modify copyrighted works would need to give users the option to include machine-readable content provenance information and, where technically feasible, protect that information so it is not easily removed. The bill makes it unlawful to knowingly remove, alter, tamper with, or disable content provenance information in furtherance of an unfair or deceptive act or practice. Covered online platforms (meeting revenue or user thresholds) generally may not remove or separate provenance information, though a platform may do so for limited security research.
The bill also bans, for commercial purposes, using covered content that has provenance information attached (or that the user knows had provenance removed in violation of the law) to train AI systems or to generate synthetic content unless the person using the content gets express, informed consent from the copyright owner and follows any terms set by that owner. Enforcement would be handled by the Federal Trade Commission as unfair or deceptive acts or practices; state attorneys general may sue on behalf of residents; and owners of covered content may bring private civil suits for removal, tampering, or improper use. Courts may award injunctions, damages, and litigation costs. The bill preserves existing copyright rights and includes a severability clause.
No publicly available information.
The bill’s findings say there is a lack of visibility into how AI systems work, little transparency about the data used to train them, and no widely accepted standards. It says it is getting harder to know the origin and authenticity of algorithmically generated or modified digital content. The bill states these problems harm the public and especially journalists, publishers, broadcasters, and artists whose works are used to train systems or are turned into synthetic content that competes in the digital marketplace. The bill’s sponsors say creating consensus standards, detection tools, watermarking practices, and public education will reduce those harms, spur innovation, and help the United States lead in AI development.
No publicly available information.