H.R. 8479 — Protecting Consumers from Deceptive AI Act

This section finds that most Americans obtain news online, with 86 percent of U.S. adults often or sometimes getting news from a smartphone, computer, or tablet (including 56 percent who do so often). It states that advances in generative artificial intelligence have increased the production of convincing deepfakes, with one study showing participants only 62 percent accurate at distinguishing deepfake from real images despite high self-reported confidence. The section identifies resulting risks of consumer deception through fake endorsements and scams, national security incidents such as the 2023 deepfake image of a Pentagon explosion that briefly affected stock markets, and election interference such as AI-generated robocalls impersonating President Biden in New Hampshire, and concludes that clear labeling of deepfakes is needed to protect consumers, national security, and an informed electorate.

This section directs the Director of the National Institute of Standards and Technology to establish task forces within 90 days of enactment to develop technical standards and guidelines for content provenance metadata, watermarking, digital fingerprinting, and other measures to identify audio, visual, and text-based content created or substantially modified by generative artificial intelligence, with a goal of making metadata cryptographically verifiable and watermarks difficult to remove. The task forces must include representatives from relevant federal agencies, AI developers, standards organizations, social media and online platform providers, privacy advocates, human rights experts, media organizations, creator associations, and technical experts in digital forensics and cryptography. Each task force must submit recommendations to the Director within 270 days of establishment and annual reports to the House Committees on Science, Space, and Technology and Energy and Commerce and the Senate Committee on Commerce, Science, and Transportation for five years thereafter, while also considering privacy-preserving methods for storing and displaying provenance data. This section further requires providers of generative artificial intelligence applications to incorporate machine-readable disclosures acknowledging the AI origin of audio or visual content they generate, implement measures to prevent removal or tampering of such disclosures, and collaborate with online platforms to facilitate identification and access to the disclosures.