"AI Image Disclosure Act"; concerns disclosure of certain AI-generated content.
A5089, the “AI Image Disclosure Act,” would require disclosure of certain AI-generated content created or altered by generative artificial intelligence systems. The bill applies to images, video, and audio content, and it requires two kinds of disclosures: a visible “manifest” disclosure that identifies the content as AI-generated, and a non-visible “latent” disclosure embedded in the content that identifies the provider, the system name and version, and the date and time of creation or alteration. These disclosures must be permanent or very difficult to remove, to the extent technically feasible, and the latent disclosure must be accessible through instructions or a tool provided by the provider.
The bill also regulates licensing relationships for GenAI systems. Covered providers must require licensees to preserve latent disclosure capability, and licensees may not remove or block such disclosures. If a provider learns a licensee has done so, the provider must revoke the license within 96 hours. The bill further allows users to opt in, with express consent and notice, to include personal data in the latent disclosure, but only after being warned that the information will be publicly accessible and permanently embedded in the content’s digital footprint.
A5089 extends beyond providers to large social media platforms, requiring covered social media companies to label content when machine-readable provenance data indicates it was AI-generated. Those platforms may not delete provenance data or digital signatures from uploaded content, and they may not retain personal data obtained through provenance data. The bill also makes it unlawful to offer tools primarily designed to remove or edit latent disclosures or to add such disclosures to non-AI content.
The bill would create new enforcement mechanisms under the Attorney General, including civil penalties of $5,000 per violation, daily violation accrual, injunctive relief, and attorney’s fees for prevailing plaintiffs. It would supplement Title 56 of the Revised Statutes and take effect on February 1 of the year after enactment, with the licensing-agreement requirements applying only to agreements entered into on or after the effective date.
Because there are no committee transcripts or recorded votes provided, the overall sentiment cannot be measured from legislative debate or roll call history. Based on the bill text alone, the measure appears aimed at transparency and consumer protection in response to concerns about deepfakes, synthetic media, and hidden AI manipulation. The main points of potential contention are likely to be the compliance burden on AI developers and social media companies, the technical feasibility of permanent disclosures, the privacy implications of embedding user personal data, and the restrictions on tools that remove or alter provenance information.
The bill would add new disclosure, labeling, and anti-tampering requirements for generative AI content in New Jersey, affecting AI developers, licensees, and large social media platforms. It would also authorize the Attorney General to enforce these requirements through civil penalties and injunctions, while creating new prohibitions on software or services designed to remove or falsify AI provenance markers. In practical terms, the measure would supplement Title 56 and impose a new regulatory framework for synthetic media transparency.
No committee discussion or voting record was provided, so there is no documented legislative sentiment to summarize. On its face, the bill reflects a pro-transparency, pro-consumer-protection approach to AI-generated media, suggesting support for disclosure of synthetic content and skepticism toward undisclosed AI manipulation. At the same time, the structure of the bill suggests likely concern from industry stakeholders about technical implementation, platform obligations, and enforcement exposure.
The likely areas of contention are the scope of the disclosure mandates, the feasibility of making disclosures permanent or difficult to remove, and the burden placed on covered providers and large social media companies. AI developers and licensees may object to the licensing restrictions and the requirement to preserve latent disclosures, while social media companies may resist obligations to detect, label, and preserve provenance data. Privacy concerns may also arise from the optional inclusion of user personal data in latent disclosures, even though the bill requires express consent and notice. Finally, the prohibition on tools that remove or add latent disclosures could raise questions about legitimate editing, interoperability, and security research.