S-4915-119
Read twice and referred to the Committee on Commerce, Science, and Transportation.
Sponsored by Brian Schatz (D-HI)
What it does
This bill would require companies that produce AI-generated or AI-modified digital content (images, video, and audio) to embed visible and machine-readable labels identifying that content as AI-generated. It would also require online platforms with 10 million or more monthly U.S. users or over $1.5 billion in annual revenue to display those labels and prohibit their removal. Chatbots would be required to disclose that users are interacting with an AI system. The FTC would enforce disclosure requirements, while the Attorney General and state attorneys general could bring civil actions against those who fraudulently remove, falsify, or circumvent the labels.
Who benefits
Consumers who would be better informed about whether images, videos, or audio they encounter online were AI-generated. Journalists and fact-checkers who would have technical tools to verify content authenticity. Victims of AI-generated non-consensual intimate imagery or defamatory deepfakes. News publishers and photographers whose authentic content competes with AI-generated material. Legitimate AI companies that follow the rules and would benefit from a level playing field. Researchers and academics studying AI-generated misinformation. Disability advocates, as the bill requires accessibility accommodations for disclosures. Smaller platforms not meeting the size thresholds, which are exempt from platform-level obligations.
Who is hurt
AI content generation companies that would face compliance costs for embedding and maintaining labels across all outputs. Large online platforms that would bear technical and operational costs to detect and display labels. Advertisers and marketers who use AI-generated visuals and would need to update production workflows. Open-source AI developers who may lack resources to implement interoperable labeling standards. Users in creative industries (filmmakers, graphic designers, musicians) who use AI tools and would face new disclosure obligations. Smaller AI startups that may find compliance costs disproportionately burdensome relative to larger incumbents. Platforms operating internationally that would need to reconcile U.S. labeling requirements with differing foreign standards.
Supporters argue
Supporters argue that AI-generated deepfakes and synthetic media are already being used to spread disinformation, impersonate real people, and manipulate public opinion, with documented cases affecting elections, financial markets, and individuals' reputations. They contend that mandatory labeling is a narrowly tailored, content-neutral disclosure requirement — not a ban on AI content — that gives consumers the information they need to make informed judgments, similar to how nutrition labels inform food choices. They further argue the bill addresses the "liar's dividend" problem, where bad actors falsely claim real footage is AI-generated to escape accountability, by requiring provenance data that can verify authentic content as well.
Opponents argue
Opponents argue that compelled disclosure of AI-generated content raises First Amendment concerns, as courts have increasingly scrutinized government-mandated speech requirements, and that the bill's broad definition of "covered AI-generated content" — content that "materially alters meaning" — is vague enough to chill legitimate creative and editorial expression. They contend that technical labeling standards may be easily circumvented by bad actors while imposing significant compliance burdens on law-abiding developers and platforms, meaning the policy's costs fall on good-faith users while providing limited protection against the malicious actors it targets. They also argue that delegating standard-setting to a NIST-led working group and the FTC raises post-Loper Bright questions about whether agency-defined standards will survive independent judicial review.
Constitutional context
The bill's disclosure mandates implicate the First Amendment's compelled speech doctrine — the Supreme Court has held in Zauderer v. Office of Disciplinary Counsel (1985) that factual, non-controversial commercial disclosures are permissible, but broader mandates face heightened scrutiny. The bill also delegates standard-setting authority to NIST and the FTC; under Loper Bright v. Raimondo (2024), courts will independently review whether those agencies' interpretive choices are authorized by the statute's text, increasing litigation risk for any rules the agencies promulgate.
Checks and balances
The FTC gains new enforcement authority over AI content providers and platforms, checked by notice-and-comment rulemaking requirements, the major questions doctrine, post-Loper Bright independent judicial review, and concurrent enforcement authority shared with the Attorney General and state attorneys general.
Historical precedent
The Digital Millennium Copyright Act (1998) established a structurally similar framework of mandatory technical standards, safe harbors, and anti-circumvention prohibitions for digital content, enforced through a combination of federal agency action and private civil suits.