The Image Content Filtration model Wikimedia Commons has been in the works for approximately 2 years (albeit, with a different name previously)
The aim of this model is to ensure that Image Vandalism which comes in the form of editing Wikipedia pages to contain traditionally graphic images (which are not required in the context of that particular page, to shock the user) from Wikimedia commons. This model aims to resolve this issue by scoring images on their probability of being graphic with a custom built dataset tailored to Wikimedia community needs (as opposed to an outsider definition of what graphic means)
Technical details about the model will be added soon
See T264045 for context.
See also T214201