![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
AO3 vs ComicFury on generative-AI policies
From the AO3, May 2023: “At the moment, there is nothing in our Terms of Service that prohibits fanworks that are fully or partly generated with AI tools from being posted to the AO3, if they otherwise qualify as fanworks. […] Depending on the circumstances, AI-generated works could violate our anti-spam policies (e.g. if a creator posts a significant number in a short time).”
As I said at the time — this is a reasonable call, and I support it.
From ComicFury, November 2023: “I will cut right to the chase, we have decided not to allow AI-art based webcomics on the site. […] It’s obviously not possible to perfectly police AI art if the author lies about it and isn’t super obvious about that, especially more subtle uses, so the rule is to some extent based on the honor system, which we think is probably better than making everyone pee in a cup all the time.”
Also a reasonable call, and I support it!
Thought this was a useful mini-case study in “what specific details lead to different generative-AI policies making sense for different platforms?”
The ones at play here:
1) Can bad-faith users exploit your site to profit off of bot-generated content?
AO3 is an explicitly non-commercial site. They already had “we will host creative remixing of material the poster does not own the copyright to” in their founding principles — and one of the legal pillars upholding that is “we are not a sales platform, users are not allowed to make money off of this.”
ComicFury has never done that — and you wouldn’t want them to. It allows users to run advertising, link their Patreons and crowdfunding campaigns, sell books/merchandise, all the tricks us webcomic creators use to get by.
The core motive for “bot-generate a bunch of low-quality slop you don’t care about” is to run ads next to it. ComicFury is a juicy target for that kind of spam. AO3, which doesn’t let you run ads on any kind of content, is not.
(This is not a moral judgment that all bot users are bad-faith users. I have a genuinely fun time getting the Bing artbot’s interpretation of prompts like “4 panel comic Erin Ptah art style”. But the bad ones exist, and your site policies have to account for that.)
2) How much will it cost you to host the bot-generated content?
ComicFury primarily hosts…comics. To make it a useful platform for that, every user can upload lots of image files, with very large sizes. An influx of bot-generated images could easily overwhelm their server space and bandwidth abilities.
(Don’t know this for sure, but I get the impression this was starting to be a concern already, and that’s one of the reasons for the policy announcement.)
AO3 primarily hosts text. (There are a few limited cases in which they host art, but there’s not a general-use “upload your image files here” option.) Text is small and cheap! An influx of bot-generated text is not at risk of overwhelming the servers. Users who tried wouldn’t even get close before getting slapped down under the “don’t spam” policy.
...and, listen, you can go back and forth all day on philosophical questions like “does the bot-generated stuff have Inherent Artistic Value?” But when platforms are setting policies, it’s okay to let those be guided by practical questions, with findable answers.
(One more generated piece to see us out: the same artbot’s interpretation of “4 panel fantasy comedy romance comic Leif & Thorn style”. Yes, it came up with 5 panels. Hozry 1 slor indeed.)
no subject
...But yeah definitely the monetary aspect would be a huge one. If there's a buck to be made, people will jump on it.
no subject
I don't think either site has a huge number of "people empowered to read and respond to abuse reports," honestly -- but CF deals with it by saying "we're not going to do hard-line enforcement against non-obvious cases," which seems like a reasonable way to handle it. People who like using bot-generated art in a comic they earnestly care about are likely to be upfront about it anyway (e.g. in that one minicomic Ursula Vernon put together), and pure uncaring spammers are likely to be obvious.
At this point, IMO the bot-generated art is easier for a human to spot the problems with on-sight. And it's definitely harder to manually fix the problems afterward. (These comics don't have a clean, uncompressed, multi-layer original file to mess with! Unlike the text, which comes out in the same plain text format you'd be editing anyway.)
But that's hard to quantify! And people have definitely posted bot-generated art while successfully hiding its origins, so I wouldn't make a policy that relies on "we can always reliably tell."
no subject
And yeah, both policies make sense for the platform's needs. I feel like since Ao3 is built on transformative fanwork, it doesn't have much leg to stand on re: AI work.
no subject
(Full disclosure, I did use the bot's most charming outputs for this post. Some of the things it spit out are weird in less-cute ways, and some are just...boring.)
no subject
It really does have trouble with faces and hands, doesn't it.
no subject