OTW Signal, May 2023

Every month in OTW Signal we’ll take a look at stories that connect to the OTW’s mission and projects, including legal, technology, academic, fannish history, and preservation issues that are important for fandom, fan culture or transformative works.

In the News

The OTW was one of 38 groups and organizations which contacted the U.S. Senate Judiciary Subcommittee on Privacy, Technology, and the Law to express their support for the protections that Section 230 offers to hosts of user-generated content online. Chair Betsy Rosenblatt explained:

Section 230 of the U.S. Communications Decency Act is the law that makes it possible for websites like the AO3 that host user-generated content to exist. Section 230 protects websites from being liable for user-posted material, and allows websites to make and enforce their own rules about what kind of content they host (as the AO3 does with its Terms of Service) without having to monitor everything that happens on the website. The OTW continues to watch U.S. lawmakers and courts closely and will continue to oppose efforts to make it harder for people to post content online or otherwise threaten the survival of small websites.


[An element of this newsletter has been removed; please see our update post of 12 May 2023 for more information.]

OTW Tips

Contributing to Fanlore is now easier than ever. Last year Fanlore gained a new visual editor and anyone can give it a try! As their help page says: “The number one rule of wiki editing is to be bold. Dive in and make changes. Other people can correct mistakes later, so have confidence, and give it a try! There are all kinds of editing conventions and philosophies governing the editing of wiki pages, but the “be bold” rule is the most important!
In general try to write clearly and concisely and be sure you always aim to improve the wiki with your edits. A valuable edit might contribute whole pages of information, or it might simply fix a typo.”


We want your suggestions! If you know of an essay, video, article, podcast, or news story you think we should know about, send us a link. We are looking for content in all languages! Submitting a link doesn’t guarantee that it will be included in an OTW post, and inclusion of a link doesn’t mean that it is endorsed by the OTW.

OTW Signal
  1. Ada commented:

    AI fanfics are a Trojan Horse.

  2. shadowmaat commented:

    If “training” computer algorithms to use modernized thinking/language is so important, why not use modern traditionally-published works instead? If the answer is “because copyright law protects them” then maybe you should stop and think about that.

    While AI as a /concept/ is interesting, it’s the /execution/ that matters. Until true artificial intelligence emerges I don’t care about what’s “good” for datamining programs. I do care very much about what’s good for fic authors and other human artists.

    • SLWalker commented:

      I’m absolutely in agreement with shadowmaat. I don’t want my works used to make an AI that can just turn around and write something competent but soulless in two minutes when some of my works have taken me years of effort and skill-building.

      Humans matter way more than some imaginary AI — and I say this as a fan of a number of AI characters! — because at the end of the day, these AIs aren’t here for our benefit. They’re here as a capitalistic power grab over some of the last creative expression we humans, many of us poor, still have control of.

      • Nope commented:

        Hard agree. I don’t think AI generated material has any place on a platform the purpose of which is to support creativity.

      • Choco commented:

        Absolutely!!! With all the ways that algorithmically generated content (it’s not “intelligent”) have already been used to attempt to squeeze intellectual property and profit by excluding actual creators, I absolutely DO NOT consent to training AI on my work and have NO INTEREST in “taking up the challenge” of allowing the people behind these AI scrapers to commodify others’ work.

        • justheretoread commented:

          Same! I’m outraged that a rep of OTW (and thus AO3) would even support this!

        • Emily commented:

          I agree that I do NOT want my writing used in AI training anymore than I would want that to happen to my artwork! It’s our creative property and should not be used like that without the creator’s explicit consent. Art belongs to people, not to AI training databases! Deviantart made the huge mistake of allowing AI scrapers to go through everyone’s work without our consent, only later adding an opt-out mode that a creator had to turn on FOR EVERY INDIVIDUAL WORK. I just ended up deleting my account. AO3 would be wise to learn from that and not allow AI scraping at all unless a creator very specifically requests that their work be used for AI training. Otherwise,
          everyone’s work deserves to be automatically protected and untouched from AI.

    • omgtabby commented:

      This this this

    • Someone commented:

      The AIs are being trained on BOTH fan fiction and traditionally published work. For example, ChatGPT was evidently trained on harry Potter.

    • 27fanficlilies commented:

      This^^^^^

    • hexoteric commented:

      >>> If “training” computer algorithms to use modernized thinking/language is so important, why not use modern traditionally-published works instead? If the answer is “because copyright law protects them” then maybe you should stop and think about that.

      I genuinely don’t understand what you mean by this – that you’d want fic to be copyrighted for further protection? That the fact that it’s based on copyrighted work means it’s illegal to use it for training material? This is a genuine question. Also traditionally-published material *is* being used for training.

      I doubt a professor of Law volunteering at AO3 hasn’t put a lot of thought into copyright laws, also. The legal situation

  3. Mina commented:

    Team, you can’t just drop this kind of AI commentary from your legal chair and walk away without addressing the valid concerns that users and donors have been raising for months.

    Questions of whether scraping is ethical or not is understandably something OTW may not want to comment on, but that does not mean you can stand idle and refrain from letting users know the OTW’s next steps on the topic.

    Some topics you NEED to address and explain for the typical user:
    – Whether the OTW’s legal team believe that AI generated works are legally transformative – and therefore officially acceptable on AO3.
    – Whether AI generated works are considered plagiarism under AO3’s TOS.
    – Whether the OTW are enquiring with OpenAI or other companies suspected to be scraping the archive to confirm whether this happened.
    – Some brief points on what this might mean (in practical terms as well as legal) for AI tools, for OTW/AO3, for IP holders who may have transformative versions of their works now incorporated into a tool which has a profit attached to it, and for fans who were scraped.

    Then and only then can you set off on a flight of fancy about how scraping fan content might lead to more diverse datasets (which certainly is NOT guaranteed or even likely, especially when it comes to fandom issues of racial discrimination or even simply the perpetuation of all sorts of stereotypes).

    Looking forward to more on this topic soon, I hope.

    • Hope commented:

      I absolutely think AI-generated fanfic is plagiarism under AO3’s TOS and any reasonable interpretation of that word.

      • hexoteric commented:

        Why?

      • b3nsn0w commented:

        I’m curious how you would differentiate AI-generated plagarism (which you seem to posit as inevitable in the output of any AI) from human-written plagarism, which arguably encompasses all fanworks. This is a very dangerous argument to make.

        • Cel commented:

          I could write a longer answer, but I’m on my phone so I’ll go straight to the point: one contains actual creativity. The other does not.

    • Impy commented:

      I agree with this comment entirely.

      I will add only that I would sooner remove my works from the archive than allow them to train a dataset. If my consent doesn’t matter and there are no plans to protect it, then I need to know that sooner rather than later.

      And if everyone feels the same way about training datasets, and no longer feels safe to include their work in the archive, then AO3 will have failed in its first and most important mission.

      • rar commented:

        agreed. I love AO3 and would hate to move from it, but I absolutely do not want my works to be used to train AI. I hope the OTW reconsiders its stance on the matter, or at least makes it clear there’s an option to opt out. I’d hate to be forced to remove my work from the only archive that hosts it becuase they are without my consent using it to train AI. Ugh.

      • Shrewreadings commented:

        What they said. Ms Rosenblatt, you have badly misjudged the opinion of your clients’ membership.

      • Nicole commented:

        AO3 has already been used by every machine learning program under the sun. The same tech that scrapes all images on the internet for facial recognition scraped text on the internet for chatgpt and others. No one can “let” them do it or stop it.

  4. Lizonka commented:

    Your stance on AI is very disappointing. I DO NOT CONSENT to AI scraping my fanfics.

    • l commented:

      THIS!!! there is practically nowhere else that’s safe for writers such as myself with the content i create and if yall start letting these folks use our works as training then i will not hesitate to remove everything i have up off the archive. i do not consent to my work being used like this and its honestly disheartening to hear this. its awful.

      • Corvin commented:

        Machine created stories are a regurgitation of real creativity. The fact that the chair of one of the biggest bastions fan creations and transformative works is celebrating and uwu-ing it is terrifying.

    • Marie commented:

      Seconded. I joined AO3 just recently and it’s a great place for publishing fanworks with a wonderful community. However, let it be clear that us authors put a lot of time and effort into our creative works and I believe this speaks for almost everyone here when I say, We would not consent to our content being scrapped. Fanfic or not, they are still our stories, ones we put a lot in, and allowing them to just be used by ai would be plagiarism and a violation of creative rights.
      If this stance turns out to be confirmed and applied, rest assured that numerous authors here would either remove their works or leave this site altogether.

      • Calvin commented:

        Training AI on AO3 works is in no way beneficial to artists, it’s just a way to dress up pushing artists out of their own field as a suitable substitute. It’s not. AI is a piss-poor ripoff and no replacement for the hard work and care put in by real people. Using AO3 fics for training is an insult to author’s ability to revoke consent to be used as training data as well as an insult to the effort it took for them to become the writers they are.

  5. D commented:

    If this is anywhere close to the OTW’s official stance on AIs scraping your site for material, you’re going to see a lot fewer donations next time the drive rolls around. What an absurd head-in-the-clouds response to an issue AO3 users have found horrifying to the point of archive-locking their works and losing both readership and community. I’m actually insulted as a longtime writer that the OTW’s legal chair thinks because I’m a minority, I should be excited that my “broad, inclusive, and diverse” thoughts are going to be eaten, mashed up and extruded by the soulless machine that people will happily use to replace me.

    People have been waiting for the OTW to formulate a position on this, and most of them think the site is going to take action to protect them and their creative works. I sincerely hope more thought will be put into this matter.

    • shadowmaat commented:

      Seconded. If the Legal Chair’s stance reflects how the Board as a whole feels- that content scraping is a good thing (for the scrapers, anyway)- then I’ll be reconsidering any future donations. I’ll also have to look into more ways to protect my content from being harvested, since I’m rapidly losing faith that the OTW will do anything to stop it from happening. This is a bad, bad road to take, team.

      • Paying Member commented:

        I firmly agree. And I will also be speaking with everyone I know in fandom and urging them to do the same.

      • no regrets commented:

        I’ve already decided withhold the donation I was considering to see what comes out of this. I gave concerns about the AI, I’ve seen one of the bots coming out with a paragraph that was almost a copy of my own response to a similar question and kind it AI can’t he copyrighted. (Imagine implications: AI can plagiarize you (and there are already cases where AI put through the plagiarism checking apps fail the test spectacularly, but then as it can’t be plagiarized because created by the AI). I think we need better legislation before anything. And yep. Let AI learn from published works.

        • conspiringangel commented:

          On top of these issues, there is no way to protect from mass misinformation being spread through scraping because AI can’t differentiate fact from fiction. If there is scraping from fanfiction, especially RPF, this could lead to major legal issues for real people and authors. I expected the OTW to protect its writers, and I’m very disappointed in the stance they have taken. I am also withholding a donation because I will not support this path of action.

    • Sparky commented:

      I agree fully. I am not excited about the possibility of my work being stolen from me only to be blended and regurgitated by a program. I’d sooner nuke my creative works entirely than allow that to happen and I believe *strongly* in the preservation that the archive supposedly stands for. But if their “preservation” is going to entail letting a machine violate artist’s works to make others money then to hell with it.

  6. Kazutoes commented:

    Are you serious? Like, genuinely, are you serious? Do you care about your userbase whatsoever? It’s absolutely horrifying that I now have to worry about my own works getting scraped by AI— which, fyi, I don’t consent to, and I’m going to bet you thousands of other users do not either— when I thought this was supposed to be a safe haven. AI is a plague on this world, and you’re going to let it kill people AND creativity if you keep going on this way in terms of your support of AI.

  7. Elena commented:

    What an ideologically troubling and intellectually vacuous survey of issues with LLMs using fanfiction! Let’s run down the list, shall we?

    1. “…machine learning is turning to broader sources from across the internet, including fan works. That means that machines will learn how to describe and express a much more contemporary, broad, inclusive, and diverse set of ideas.” LLM training on data scraped from the internet has to be manually reviewed by human beings who are paid very low wages to expose themselves to content which may harm them. This is true even of fanfic. Coverage of recent LLM training, in fact, has specifically called out fanfic as a vector for trauma for humans training the model. Additionally, the assumption that a LLM’s dataset would be inherently less bigoted simply because fan writers are producing work while female, POC, gay, etc., borders on the absurd. There are plenty of fics on AO3 where gay doctors get slapped around sexily and called slutty bitches. And long may they diagnose! But a LLM could easily be taught that “omega slut” is the most probable clause after “do you like that”, using AO3 data, because a LLM cannot think or reason.
    2. Does the OTW Legal chair read fanfiction? People already write about robots fucking and falling in love. Please be serious.
    3. “We call it machine “learning” for a reason. A well-trained machine won’t generate an infringing work, but it needs as large a pool of data to work from as possible to do that.” This is factually inaccurate and a simple Google produces people describing the process to get a LLM to reproduce training material, e.g. https://medium.com/@neonforge/chatgpt-copyright-concerns-and-potential-legal-consequences-for-openai-56feb6974c27 . If the OTW’s argument is that a best-in-class modern LLM is insufficiently trained, I’d suggest that be made clear! Otherwise, one would be forgiven for assuming OTW Legal is in the habit of making easily disproven statements about technology they lack the domain knowledge to appropriately evaluate.
    4. “Consider the classic Sony v. Betamax case: The VCR can be used to infringe, but it has noninfringing (fair) uses, and therefore the VCR does not inherently infringe. I recognize that the analogy isn’t perfect, but I find it persuasive.” The key decision in the Sony v Betamax case was that the act of recording a video does not constitute infringement. In this case, the correct analogy might be the existence of web scraping technology or even the act of a nonprofit scraping AO3 data for analysis purposes. Upon using that data to train a LLM, however, many are arguing infringement occurs, because a LLM is not a thinking being; it cannot create new phrases absent a pool of data points to draw upon. Because of this, LLMs being used in production have been documented duplicating passages of training data verbatim. I find it profoundly alarming that the OTW thinks a fan’s work is fair use to be chopped up, remixed, and resold to a third party. What precisely is the difference between that position and Fanlib’s attempts to commercialize fic? Why has OTW Legal decided that a pro-commercialization argument is “persuasive”?
    5. “[Google v Oracle] case highlighted that we shouldn’t be locked into one definition of “transformative” work, and that copying for the purpose of engaging in a different technological use can be transformative copying”. I would personally be alarmed if judges took seriously an argument that Google v Oracle and infringement cases against LLMs are remotely comparable. Google v Oracle was focused on whether or not reimplementation of an application programming interface constituted copyright infringement. It had nothing to do with the data used for advanced statistical analysis nor was it related to the output of that analysis (language generation). The inputs of a system are extremely different from the code used to manipulate those inputs.

    Shameful, embarrassing, condescending, and flat-out wrong. The OTW owes its stakeholders an apology.

    • NN commented:

      I appreciate this excellent breakdown of both the faulty logic here and some actual fact checking into cases carelessly mentioned. Thank you Elena. I think sentiment is extremely clear that writers want to be able to opt out of letting AI scrape fic for “content”. How can OTW make it happen?

    • Emily commented:

      brilliant, thorough response. they owe us answers for all of the questions you raised.

    • aaaaaa commented:

      open-source and non-commercial models exist, you know

      • Anti-AI commented:

        Yeah, and? Consent is irrelevant because transformative works are free-to-chum? I’ll pull my works if AO3 permits scrapers.

  8. Deep Disappointment commented:

    Jesus Christ, is the AI article linked here the actual expressed position of the OTW? Has this org really fallen so completely from the days of fanlib, that its leaders have gone from opposing corporate commercialization of fanfic to “Actually, it’s great if the billionaire corp scrapes fanfic to improve its for-profit product because… uh, diversity! Inclusion!” (And, just for the record, if these ultra-capitalist companies really do need to rely on *nonprofit fanfiction labor* to make their product non-bigoted, then that is a good argument for why these companies absolutely should not exist, not a good argument for why the OTW should just let them sit back and do their thing and profit off of fanfic scraping.)

    Other creative works sites are already working with creators to help keep their works out of the hands of corporations as much as possible. It’s tremendously disappointing that this organization, which has long prided itself on being the most pro-fan, pro-creative org., has seemingly chosen to abandon those principles and go the corporation-worshiping route instead. Hopefully there will be a course-correction coming soon, because I can guarantee that the AO3’s userbase at large does not want leadership who would approve of and publish an article like this.

  9. Clarity Jones commented:

    This is a pretty tone-deaf response from the legal committee. I don’t believe anyone on AO3 is concerned with anthropomorfic (a genre that long predates ChatGPT and will be around long after we’re all dust). The Legal Chair’s personal taste in fic is neither here nor there. The major concerns I and other AO3 users/OTW members have are:

    1. The proliferation of AI-written content on AO3, and the limited tools AO3 provides to at least let us filter it out.

    2. Companies scraping our noncommercial fanworks without our consent and using them to train large language models that will be used for commercial purposes, such as Hollywood studios attempting to union-bust by using AI “scabs” to replace striking writers.

    The OTW has been remarkably quiet on the first part. On the second part, we now have a statement from the OTW Legal Chair, advertised on the official OTW site, that she agrees with this use of our fanworks! (It would be useful for the OTW to consider the difference between “can” and “should,” as well as between “legal” and “moral.’) Absent any other statements to the contrary, that looks like an official and appalling OTW endorsement.

  10. SkarlettSkwrl commented:

    AI has been of great controversy as of late. Writing certainly isn’t the only creative industry to be impacted. We are in the age of data, control over information every and is quite lucrative. If this is the stance that’s being taken. I can’t say I’m all that suprised, just disappointed. It goes against why I respected this site.

    However, one middle ground method that should be taken is having a opt in setting on the writer’s profile and or works blacklisting that user’s data from being mined. This way users that don’t wish to endorse AI data mining with works can choose not to. While those that do can.

  11. I DO NOT CONSENT TO AI commented:

    I hadn’t had time to make my annual OTW donation yet, and after seeing this horrible stance on AI, I’m rethinking ever donating again. This goes against all of AO3’s policies. I do not consent to my work being scraped for AI. I seriously suggest you take a step back, rethink this article, and retract and apologize for treating your userbase – people who trusted OTW and AO3 – this way.

  12. S commented:

    I don’t think many users would be in agreement with your stance. Of course you wait to announce your position until AFTER you surpassed your recent donation goal. It’s clear you don’t have our best interests at heart. Overall very disappointed. I will never donate to your organization.

  13. mikkimouse commented:

    The use of AI is literally one of the reasons the WGA is striking now–in fact, a Twitter thread on the subject says “plagiarism is a feature of the AI process.” The way AI is trained and used right now is astoundingly unethical, and I don’t want my fic being used to train programs like this, for many of the reasons stated above. I really hope you all reconsider your stance on this.

  14. L commented:

    Absolutely not. There needs to at least be a tool for us to decline AI scraping our fics. This is beyond tone deaf and doesn’t account for the very real dangers artists and writers are facing with AI.

  15. L commented:

    Absolutely not. There needs to at least be a tool for us to decline AI scraping our fics. This is beyond tone deaf and doesn’t account for the very real dangers artists and writers are facing with AI. This is absurd

  16. An AO3 writer commented:

    Fuck you. What a shitty take.

  17. Former Donor commented:

    Really great way to get a significant chunk of work removed from your site, thereby defeating the entire purpose for which the archive was created.

    This explains why I’ve also recently seen one of my fandoms inundated with several ChatGPT “written” works, but when I reported them I was told they don’t violate the TOS. AO3 coming down in favor of this trash over the real expression of fans’ creativity

  18. AO3 writer commented:

    Supporting AI is an insult to real creators and the effort put in to populate a site lauded as being “for fans”. There is no happy or just outcome to allowing AI to scrape data from the hard work of fans.

    • l commented:

      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ai “writers” have no place on this site. their “work” is derived from actual writers without those writers’ consent and its absolutely abhorrent that you guys (otw) would be fine with this.

  19. disappointed AO3 user commented:

    What? Unbelievably disappointing. Please at least let us opt out. This seems completely at odds with the OTW. AI is the antithesis of transformative and completely out of step with the spirit of fandom. I will not be donating in the future if this isn’t changed or addressed.

  20. drosera commented:

    This is shockingly horrible and is going to be a huge blow to fan communities when people take their works down. Please, please reconsider. At the very least it MUST be set to “opt out” by default. Please don’t be part of the evil default here.

    I cannot believe this is being taken as a stance. This will destroy the archive and all it stands for.

    I’ve been such a strong Ao3 supporter for so long, so this is such a horrible blow. Please, please reconsider.