Nexus Mods makes AI disclosure mandatory for every new mod

What Nexus Mods actually changed

Nexus Mods, the PC gaming mod repository that most modding communities orbit, announced on 3 November that disclosure of AI use is now mandatory for new mod uploads. Authors have to declare "whether AI has been used in their work," the site said in a long post about AI-generated content that The Verge reported on the same day.

Site owner Victor Folmann, who goes by Foledinho, framed it as a response to user complaints rather than a moral position. "One of the most consistent concerns we've heard is that it's not clearly disclosed when AI has been used to create a mod," he wrote. "We already have tags for AI-generated and AI-assisted mods, but we recognise that there's a lot more we can do to make that information more accessible and clear for you all."

In practice that means three things: mandatory declarations on new files, a public explanation of how AI was used on anything labelled AI-assisted, and more prominent AI tags on mod pages. The site also added controls so users can hide AI-generated content on game pages and on the homepage, with search filters described as next up.

Three labels, three different burdens

The policy page breaks disclosure into categories that carry different evidence requirements. AI Generated Content covers work where the core output came from AI tools: generated code, synthetic voice lines, generated textures, generated text or generated scripts. AI Media covers promotional material outside the game files, such as thumbnails, page banners and videos.

AI Assisted Content is the interesting one, because it is where most real mods will land. If AI played a supporting role, such as texture upscaling, audio cleanup or initial concept art, the author must list the specific tools, software or models used, state exactly which outputs in the final mod came from AI, and provide evidence of human-led work. The policy names development history, project file history, version control commit history, and a genuine understanding of the code that an author "may have to explain in moderation disputes."

That last clause is the enforcement teeth. A claim of AI-assisted development without evidence can be replaced by the moderation team with a different classification, which means the burden of proof sits with the author rather than the platform. Authors also have to update their public AI statement whenever an update adds new AI components, so the declaration cannot go stale after the first upload.

Why there is no ban

The decision that will annoy the loudest part of the community is the one Nexus Mods made deliberately: AI-generated mods stay. "We are pro-human creativity," Folmann wrote, per GamesRadar+, but the site "won't introduce an outright ban on AI-generated mods."

His reasoning is practical. "AI is currently used throughout the whole process of creation, from script autocompletion to generation of textures, voices, and full assets, and it is difficult to draw a clear boundary between what is acceptable and unacceptable," he wrote. "An outright ban will not eliminate AI from the modding process but will prevent people from mentioning its use. We'd prefer the opposite; it should be out in the open, clearly labelled, and regulated according to the rules that we can implement on our platform."

That is the same argument platforms have made about AI slop for two years, and it is hard to refute on its own terms. A ban is cheap to announce and expensive to enforce, because the thing being banned is invisible in a finished file. Labelling at least creates a record that users and moderators can act on.

Nexus Mods also committed not to train AI on hosted mods and not to sell or license them to anyone who does. Given how much of the last year's AI copyright fight has been about exactly that, the pledge is worth more than the wording suggests, even if it is a promise rather than a technical control.

Permissions now default to Ask Me

The change with the longest tail is not the label, it is the permission default. Nexus Mods is introducing explicit settings for AI Training and Tool Usage and for Forks and Parallel Versions, and those settings default to "Ask Me" across all hosted files, including mods published years before the options existed.

That inverts the old arrangement. Under a permissive default, a modder who never thought about AI training was implicitly granting it. Under Ask Me, silence is not consent, and anyone who wants to train on a mod has to get an answer. Nexus Mods also said it will strengthen attribution when a mod uses another author's work with permission, and reserves the right to request source code when it needs to assess whether a file is safe, declining to host files where concerns remain unresolved.

For a repository with decades of archived uploads, retrofitting that default is a bigger operational job than writing the policy. It also shifts the site's relationship with the AI training industry from passive host to gatekeeper, without ever having to argue the legal question in public.

The January deadline and the upload problem

Authors have 60 days to bring existing mods into line. The grace period ends on 2 January 2027, and enforcement of the itemised disclosure rules begins after that. Until then, the site says basic disclosure of AI usage is already required under existing policy, so the new regime is a tightening rather than a switch from nothing.

Nexus Mods also admitted a problem that is only partly about AI. Its Recently Added section was being flooded by authors uploading many mods at once, pushing everyone else off the page. Folmann said this is "less of an issue related to AI, but it's becoming much more prevalent as mod production is speeding up." The fix includes limits on how many mods from one author appear at a time and tighter restrictions on uploads from new accounts.

That combination, faster production plus a flood of low-effort uploads, is the same dynamic every marketplace and app store has hit since generative tools got good. The platforms that survive it tend to do what Nexus Mods just did: keep the door open, make the label mandatory, and give the reader a filter.

Why this matters outside modding

Modding is where the AI disclosure argument gets tested with the least patience for corporate language. Mod users are technically literate, they inspect files, and they notice when a texture pack is upscaled or a voice line is synthetic. If labelling can work here, it can work in less technical communities.

The contrast with Steam is useful. Valve imposed an AI disclosure rule on store listings, and Epic Games chief executive Tim Sweeney called it "abhorrent" and a political statement, as AI Lookout covered in [Sweeney's criticism of Steam's AI disclosure rule](/blog/tim-sweeney-calls-steams-ai-disclosure-rule-abhorrent-and-a-political-statement/). Epic took the opposite approach and left disclosure to developers.

Nexus Mods lands closer to Valve, with one difference that matters: it is asking for evidence of human work rather than a one-word checkbox, and it is defaulting training rights to no. Both are harder to game than a tag. AI Lookout has followed the same problem in [Anthropic's changes to stop AI video game clones](/blog/anthropic-adds-two-sentences-to-claude-code-to-stop-ai-video-game-clones/) and in [Pennsylvania's law requiring labels on AI-generated images, video and audio in ads](/blog/pennsylvania-makes-advertisers-label-ai-generated-images-video-and-audio-in-ads/).

The modding case is the cleanest test of the whole idea. Nobody is arguing about jobs or copyright here. The question is simply whether a community will accept clearly labelled synthetic content, or whether labelling just moves the fight to a filter checkbox. Nexus Mods is betting on the checkbox, with a policy that starts biting on 2 January.