This talk explores the impact of generative ML models on the production of digital commons. Traditionally, digital commons such as open source projects, Wikipedia, or the movies released by the Blender Foundation have relied on copyright licenses to ensure openness and protect against unilateral appropriation. With the advent of generative ML models that rely on large amounts of online content for training, some of these arrangements are breaking down.

How can we ensure that large-scale collaborative projects can continue to thrive under these conditions, and are there ways to channel some of the surplus generated by generative ML systems back into the commons?