Nexus Mods began requiring AI disclosure on every new upload this week. Modders now have to declare whether a mod is mostly AI-generated, mostly human with some model assistance, or free of AI altogether, and the tag appears near the top of the mod page, just behind the adult-content tags. Owner Victor Folmann announced the rule on Monday, and PC Gamer reported the details, with GameSpot adding the filtering side.
The middle category carries the most work. A mod tagged AI-assisted needs a written explanation of where the model was used and how much of the result it produced, which is a harder question than it sounds for anything that went through several passes of text, texture or voice generation. Existing mods have 60 days, from November 3 to January 2, 2027, to add a declaration, and Nexus says the rule covers everything on the site rather than only new files.
Three boxes and a paragraph
A disclosure system is only as good as its categories, and three is a small number for a practice that runs from a single generated texture to a mod that is largely synthetic. Folmann's argument for keeping it simple is that a finer taxonomy would be argued over instead of used, and that the visible tag plus the written note gives readers enough to judge a file before they install it.
The site also promised not to train models on hosted mods or sell them to anyone who would, and added a permission that lets authors state whether outside parties may use their work for training, defaulting to asking first. That default is the part with teeth for the long run: it turns an assumption of availability into a permission that has to be granted.
Why a ban would hide more than it stops
The stated reason for allowing AI mods is that a ban would push the practice underground rather than end it. Folmann's line is that prohibition stops people from mentioning AI use, not from using AI, and a platform that cannot see a practice cannot label it or moderate it.
That reasoning puts Nexus in the same position Valve reached with Steam, where disclosure replaced prohibition. It is also the position most large platforms are converging on, because detection is unreliable and appeals are expensive. The alternative, visible in the open-source world, is worse than either: Linux dropped its AI disclosure tag after the label started fights over patches nobody could verify.
The cost of the approach is that enforcement depends on honesty. Nexus says it is putting moderation standards and consequences in place, and also says plainly that not everyone will comply. Both statements can be true at once, and the second one is the honest description of how the system will work in practice.
What moderation cannot check
Nobody at Nexus can tell whether a normal map was generated or painted, and the written explanation is a self-report. That makes the tag a claim by the author rather than a finding by the platform, which is a meaningful difference from how copyright disputes are handled. A 1998 ruling that Duke Nukem maps were derivative works still sets the frame for who owns a mod, and no disclosure box answers that question.
The filter is the part that gives users real power. Once AI mods are tagged, a player can exclude them from search the way they would exclude adult content, and the market decides what the label is worth. If most players ignore the tag, the disclosure has cost the site a little friction and changed nothing. If a visible share of players filters, mod authors will have a reason to keep models out of the parts of a mod that matter.
The 60-day window closes on January 2, and the interesting number after that is not how many mods are tagged but how many are tagged AI-assisted with an explanation that says something specific. That is the figure that will show whether the system produces information or a checkbox, and Nexus has said it will enforce the rule either way.