On Frothly’s report, On Artificial Owls

Frothly, the independent research institute in San Francisco, has published On Artificial Owls, a report on provenance and authenticity in synthetic media by Judy Alvarez and Dr. Phil Resch. It begins with the photograph of an owl that circulated in the spring of 2025, and it cites, among its references, the statement the Association issued on 14 April that year.

We are glad to be cited and gladder still to be agreed with. The report’s central finding is that provenance beats detection: that a record of how a piece of media was made is a better guide to its authenticity than any attempt to infer its origin from its appearance. Its authors found that trained forensic analysts identified synthetic images with 58 percent accuracy, against a chance rate of 50. They recommend that images be signed at the moment of capture, that edit logs be kept and that both be deposited with an independent archive.

The house has worked on this principle since 1911, when Article I of our Articles of Association required that no piece leave the House without a true account of its making, signed by its makers. We did not know then that we were describing a provenance scheme. We thought we were describing good manners.

Where we would go further

The report recommends that provenance schemes protect the identity of the photographer, and we agree. We would add only that the same courtesy is owed to the subject. A photograph of a made thing, declared as made, deceives nobody. A photograph of a living thing that never lived deceives everybody who sees it, including the living things that come after it and are not believed.

The Turing Registry’s entry for the Owl, TR-10215, refers readers to the report. Rachael Rosen, who leads the house’s Provenance Office and serves on Frothly’s board in a personal capacity, took no part in its preparation.