Regulators open consultation on real-time model disclosure
The proposed rules would require newsrooms and platforms to label machine-generated output at the point of publication.
- Draft rules would require point-of-publication labelling for machine-generated text, audio and imagery.
- The consultation runs for ninety days and covers platforms above a five-million-user threshold.
- Industry bodies have asked for a phased implementation over two years.
Machine-generated from the reporting on this page and reviewed by an AMN editor before publication.
Regulators opened a ninety-day consultation Thursday on rules that would require newsrooms and platforms to label machine-generated text, audio and imagery at the point of publication rather than in a general disclosure elsewhere on the page.
The draft applies to platforms above a five-million-user threshold and to any organisation holding a broadcast licence, a scope that would capture most national newsrooms.
Industry bodies have already asked for phased implementation over two years, arguing that retrofitting provenance metadata into existing production pipelines is the substantive cost, not the labels themselves.
Several newsrooms, including this one, already publish machine-assisted summaries with explicit attribution and a confidence indicator.
The consultation closes in late October, with final rules expected in the first quarter.