Alphabet's Jigsaw Develops Tool to Help Journalists Identify Manipulated Imagery

Image for article titled Alphabet's Jigsaw Develops Tool to Help Journalists Identify Manipulated Imagery
Photo: Getty

The proliferation of doctored or misleading images that manage to spread like wildfire has required that newsrooms, fact-checkers, and journalists be especially wary of disinformation and targeted campaigns in various media. Jigsaw, an Alphabet-owned company, has announced a new tool aimed at helping journalists better identify when images and other media have been manipulated.


Jigsaw announced the tool, called Assembler, in a blog post on Tuesday. The company’s founder and chief executive Jared Cohen described the product as an “early stage experimental platform advancing new detection technology” on which Jigsaw working with Google researchers as well as academic partners from the University of Maryland, University Federico II of Naples, and the University of California, Berkeley.

Basically, the tool has been trained to identify several ways that photos or media are doctored. The various detection models contributed by Jigsaw’s academic partners are used by the tool “to show the probability of manipulation on an image,” including one specifically meant to help identify deepfakes.

“Assembler brings together multiple image manipulation detectors from various academics into one tool, each one designed to spot specific types of image manipulations,” Cohen wrote. “Individually, these detectors can identify very specific types of manipulation—such as copy-paste or manipulations to image brightness. Assembled together, they begin to create a comprehensive assessment of whether an image has been manipulated in any way.

Cohen said that the tool will be tested by newsrooms and fact-checkers including Agence France-Presse, Animal Politico, Code for Africa, Les Décodeurs du Monde, and Rappler in order to help improve it.



An increasingly important tool. I see that they are inviting image data sets, but it seems it would be more productive to distribute betas, and let people then submit image pairs with source and altered versions, in which it failed to detect manipulation.