Someone keeps accusing fanfiction authors of writing their fic with AI, and nobody knows why

Synthetic intelligence has… let’s say, a fraught relationship with the humanities neighborhood, and transformative fandom isn’t any exception. Many followers have embraced some instruments like Character.ai bots, however a lot have additionally bristled on the prospect of textual content mills getting educated on their fanfiction; in my fandom circles, I’ve seen a number of individuals say the prospect of an AI-created content material flood discourages them from writing. And previously week, the backlash has created an odd second-order complication: somebody is rampantly accusing non-AI fan writers of being “AI-using cheaters,” and so far as I can inform, no one fairly is aware of why.

In line with posts and responses on Twitter, Reddit, and Tumblr, a slew of authors have been receiving obvious spam feedback on Archive of Our Personal (or AO3), one of the crucial well-liked fanworks web sites. In its earliest iteration, the remark seemed like this:

This work has been generated by HoloAI based mostly on an algorithm.

The authors in query have denied utilizing AI instruments, and a few feedback have been seemingly left on fics that predate the instruments they’re accused of utilizing. The consensus is that it’s a kind of spam, and AO3’s operators acknowledged the issue yesterday, coming to the identical conclusion.

It’s additionally, nonetheless, irritating to authors. The feedback began round April 1st, main many individuals to invest they have been an April Fools’ joke. Some raised the chance they have been an advert for HoloAI or a handful of different writing providers that have been generally talked about in lieu of it, like NovelAI or Sudowrite. However the feedback have continued a number of days into April, they usually’ve acquired a extra judgmental tone. One current iteration warned “all readers” to scan the work with the Winston AI detection system, exhorting them to “name out all AI-using cheaters.” (This has, in flip, led to hypothesis that it’s promoting AI detectors, a expertise that’s fairly actual however nonetheless experimental.)

These don’t appear to be efficient adverts, to place it mildly; AO3 is a noncommercial writing platform for a neighborhood the place many members detest generative AI. They don’t appear aimed toward any explicit fandom or character pairing, so that they don’t seem like a focused harassment try. It is perhaps a real (if overzealous) try to purge AI-written fic, but it surely’s most plausibly a sort of general-purpose insult trolling — fueled by the inventive neighborhood’s antagonistic relationship with a much-hyped new software.

One purpose this bothers authors a lot is that — in comparison with most well-known social platforms — AO3 is a really intimate house. Feedback on fanfiction are a community-building software that many website members take very severely and (particularly in small fandoms) have a good time receiving in any respect. On Reddit, one writer lamented getting their first-ever response to a posted work… and realizing that it was “that fucking AI remark.”

I contacted AO3’s operator, the Group for Transformative Works (OTW), concerning the spam feedback. “We’ve made some adjustments to the positioning code that may forestall responsibly coded bots from accessing it. Nonetheless, stopping all automated site visitors wouldn’t solely block AI bots, however it will break engines like google and a few display readers, amongst different instruments,” communications director Claudia Rebaza wrote in an announcement from the OTW assist workforce. Proper now, OTW recommends individuals merely mark the feedback as spam.

Rebaza famous that the neighborhood has requested quite a lot of questions on AI instruments, together with whether or not AI fic needs to be allowed on AO3 in any respect. “In the intervening time, we don’t prohibit AI-generated fanworks as such, in the event that they in any other case qualify as fanworks,” Rebaza stated. “Nonetheless, we’re monitoring the scenario. Relying on the circumstances, AI-generated works might violate our anti-spam insurance policies, however it is a growing scenario and we now have not made a blanket rule.”

Generative AI has confirmed a nightmare for paid fiction websites, which have confronted a flood of spam-like submissions from AI hustlers. AO3 (which bans any sort of monetization) doesn’t provide the identical incentives. Authors have been posting AI-generated fic because the early days of textual content journey generator AI Dungeon, however typically in a method that highlights the poster’s inventive prompting or the instruments’ capability for absurdity. However as a tradition struggle has grown round generative AI, the backlash for supporting or opposing it has gotten stronger. Now, writers don’t simply have to fret about competing with an AI — they’ve to fret about getting accused of being one, too.