In the context of AI, transparency may involve different types of disclosures, such as the use of a machine learning tool to make consequential decisions about consumers or the use of a chatbot to interact with consumers. The U.S. does not currently have a federal law that specifically mandates transparency in AI systems. Some laws of general applicability, like broad consumer protection laws, may require disclosures about AI to avoid consumer deception. On the state and local level, however, a patchwork of laws has developed requiring transparency in different situations. For example:

• California’s Generative AI: Training Data Transparency Act mandates disclosure of high-level details about the training data used in generative AI systems
• California’s TFAIA requires large “frontier” AI developers to publish transparency reports and annually update a public frontier AI safety framework describing how they assess and mitigate “catastrophic risk,” secure unreleased model weights, and respond to critical safety incidents
• Colorado’s AI Act requires developers and deployers of high-risk AI systems to maintain documentation that demonstrates reasonable care in preventing algorithmic discrimination
• Utah’s AI Policy Act mandates verbal or written disclosure when consumers interact with generative AI in regulated service contexts
• New York’s RAISE Act requires large developers to implement and publicly disclose a “safety and security protocol” and report any “safety incident” to mitigate risk
• New York City’s Local Law 144 requires employers to notify candidates when automated employment decision tools are used, and to publish the results of bias audits

These efforts may reflect a growing consensus that transparency is key to responsible AI deployment, particularly in applications such as employment, healthcare, and consumer services. However, the scope and enforcement of transparency obligations vary significantly across jurisdictions, contributing to a fragmented compliance landscape.