The AI Image Disclosure Act mandates that any content generated or altered by a generative artificial intelligence system (GenAI system) must include both a manifest and a latent disclosure. The manifest disclosure must clearly identify the content as AI-generated, be easily understandable, and be difficult to remove. The latent disclosure, which is not visible or audible, must provide specific information about the content's origin, including the name of the provider and the GenAI system used, and must also be permanent and compliant with industry standards. Additionally, covered providers are required to ensure that any licensing agreements with third parties (licensees) maintain the capability for these disclosures and prohibit the removal of such disclosures by the licensee.

Furthermore, the bill imposes obligations on covered social media companies to label any detected machine-readable provenance data in AI-generated content and prohibits them from deleting such data or retaining personal data acquired through it. It also establishes penalties for violations, including civil penalties of $5,000 for each infraction. The act is set to take effect on February 1 of the year following its enactment, with specific provisions applying to licensing agreements entered into after this date.