Sourcehut is banning LLM-generated content outright on September 10, on the honor system.

The August 27 ToS change prohibits "original content written with or which facilitates the use of LLMs or other generative AI technologies" across source code, assets, tickets and emails, effective for new projects with no retroactive penalty (Sourcehut). The stated reasons are electricity draw, labor conditions in training data pipelines, license-ignoring ingestion, AI-generated projects burning build minutes and storage to the point of platform outages, and maintainer exhaustion from low-quality contributions. There's no detection tooling. Enforcement is that a discovered violation forfeits the assumption of good faith. I don't think that's enforceable at scale, and I also think it's a defensible position for a platform whose costs are being externalized onto it.