Policymakers are increasingly treating AI labels as a way to reduce confusion and protect audiences from misleading or unattributed content. Research with news audiences in the US found that readers want labels to clarify responsibility, attribution, credibility, trust and personal choice, but they also want details about human involvement, including oversight and the specific tasks AI assisted.
New York’s proposed “fundamental artificial intelligence requirements in (FAIR) news act” would require news media published or disseminated in the state to conspicuously state when content was “substantially created by generative artificial intelligence.” The EU Code of Practice on AI-Generated Content gives deployers reference icons for fully AI generated or AI modified content and implements labeling obligations under the EU AI Act.
The policies differ in scope and trade-offs. New York focuses broadly on news media, while the EU covers AI-generated images, audio, video and public-interest text, with an exception when content has undergone human review or editorial control and an identified person or entity holds responsibility. That leaves unresolved questions around AI-assisted reporting, where both people and systems contribute and where responsibility may not align cleanly with labels or bylines.
