AI Labeling Act of 2026 (S.4915): requiring a label on AI-generated text, images, audio, and video

1Output you can no longer tell apart

Generative AI can now produce text, images, audio, and video so naturally that they are hard to tell from human work. Convenient as that is, it also creates deepfakes that make fictitious events or statements look real, and situations where AI-written text is mistaken for human writing, raising concerns about misinformation, fraud, and erosion of trust.

Against this backdrop, the idea of labeling, letting recipients tell whether content was made by AI, has drawn attention in policy debates worldwide.

2Labeling as the approach

S.4915, the AI Labeling Act of 2026, is a Senate bill that seeks to put this idea into law. As its name (Labeling) indicates, its purpose is to require a recognizable marker on AI-generated content. The sponsor is Sen. Brian Schatz (D-HI); it was introduced on June 24, 2026, and referred to the Senate Committee on Commerce, Science, and Transportation.

Referral to a committee is a step in the legislative process meaning consideration has begun in the committee overseeing the area. The specifics of labeling (which media, in what form, and who applies it) go beyond the information presented here, so no firm conclusions are drawn.

3What comes with mandating a label

Labelling AI-generated content is expected to lower the risk of disinformation and impersonation through transparency. Mandating it, however, brings its own issues.

Generative AI now produces text, images, audio and video as naturally as human hands. Convenient as that is, it produces deepfakes presenting events and statements that never happened as real, and situations in which text written by AI is taken for text written by a person. The detail of how a label must be applied is beyond what is shown here.