This commentary discusses the rise of generative machine learning tools (so-called artificial intelligence [AI]) and their joint informational and environmental harms. It takes the argumentative stance that most AI today can be usefully described as a kind of trash, both in terms of the quality of its outputs and its proliferating social and ecological costs. Extending the analogy further, it explores how the AI trash problem is best approached like other discard problems, namely by pursuing policy solutions that address the preconditions for waste rather than relying on individual awareness or moral judgements about consumer behaviours to fix injustices. It concludes with a few directions to this end as well as a call for coalition-making on the part of environmentalists and tech critics.