Datasets:
๐ฉ Report: Copyright infringement
We received a DMCA takedown notice from the representatives of Transformative Works (link) alleging that materials in this repository uploaded by you infringe their copyright.
When we receive a DMCA notice, we are required to remove access to the allegedly infringing materials.
@nyuuzyou
, You may submit a counternotice if you believe the takedown request is unfounded. Please send it over to dmca@huggingface.co and we will forward it to the reporter. A sample counter-notice can be found at https://www.copyright.gov/512/sample-counter-notice.pdf.
Donโt hesitate to contact us if you have any questions.
Thank you,
Hugging Face Team
For the community: I have filed a counter-notice in response to the DMCA takedown of my dataset. According to the DMCA process, the dataset should be restored within 10-14 business days unless OTW takes legal action. I believe the takedown request was unfounded, and I'm working to make this dataset available again as soon as possible.
Stop stealing from us.
Stop stealing from us.
Your 'work' is stolen from copyright holders in the first place. Unless you claim to own Naruto, One Piece, and Harry Potter?
None of those. Original fiction, actually. And regardless, the tos on this very website disagrees with you as well as the laws regarding transformative works. Fanfiction is by literal copyright law definition not stealing anymore than the divine comedy is stealing from the source it's based on. AO3 made it very clear that this was illegal and their legal team is currently issuing takedowns, from my understanding. Something doesn't have to be the best art ever to still be legally protected. Not stealing isn't hard. Just be a decent person. These are the works of millions of people. Just don't. It was already free for you to enjoy. But I guess some people just like to spit on the joy of others.
Stop stealing from us.
Your 'work' is stolen from copyright holders in the first place. Unless you claim to own Naruto, One Piece, and Harry Potter?
In that case, this dataset would include work stolen from the copyright holders of Naruto, One Piece, and Harry Potter and would still be against TOS (and the law).
While we can disagree on what fanfiction is or it's value, no matter what this is a stolen dataset.
Then the final point is that under US law while the characters and IP are owned by the original copyright holders, the written work or art created by fans is owned by those fans.
Just like an official artist would hold the copyright of an image of Mickey Mouse they made (even if they don't own the character) until or unless they signed away the copyright to Disney as part of their contract.
Additionally they can give Disney distribution rights (part of copyrights) while keeping the right to make derivative works for themselves (Disney would rarely allow this, this is a metaphor). This is similar to fanfic authors or fanartists giving OTW the right to distribute their work on Archive of Our Own.
This is in contrast to this dataset which has not been given distribution rights by the fanfic authors or fan artists whose work has been stolen.
We could get into derivative vs transformative works as well, but that's for another day.
Stop stealing from us.
Your 'work' is stolen from copyright holders in the first place. Unless you claim to own Naruto, One Piece, and Harry Potter?
In that case, this dataset would include work stolen from the copyright holders of Naruto, One Piece, and Harry Potter and would still be against TOS (and the law).
In which case the actual copyright holders or their registered agents are the only ones who can file a DMCA, not people who are purposely infringing on copyright themselves.
And no, you don't have permission from Disney or any other person to distribute your plagiarised works. You do realise AO3 is a far more viable target for a DMCA than this dataset, right?
My works are among the ones scraped and the majority of that is original works (about 2 million words worth of original work and some 500k fanfic) and I have not consented to either being used.
My works are among the ones scraped and the majority of that is original works (about 2 million words worth of original work and some 500k fanfic) and I have not consented to either being used.
Original works are all you can DMCA, you don't have copyright over fanfictionโgo right ahead and they can remove anything tagged original works.
My works are among the ones scraped and the majority of that is original works (about 2 million words worth of original work and some 500k fanfic) and I have not consented to either being used.
Original works are all you can DMCA, you don't have copyright over fanfictionโgo right ahead and they can remove anything tagged original works.
That is not true, the creator of transformative works (ie most fanfic and such) own the copyright to their works, what they don't own is the characters and events from the original works, any original characters added or storyline created, they own.
My works are also in the dataset, and I did not and do not consent to them being scraped, or their use for machine learning training of any kind.
I've long found machine learning to be incredibly fascinating, this sort of behavior leaves me perpetually disappointed in what the technology has become.
I naively used to believe that image generation models would only be trained on open-source images, that any language models that might be profited off of would be trained on freely provided text and classical works.
As a computer scientist, I am ashamed to be associated with people like you, nyuuzyou.
Fanfiction sits in a legal gray area where it is permissible only if it is not profited from. Since you clearly refuse to remove this dataset from the internet, I hope that some idiot uses it to catch the attention of the likes of Disney, and that they sue your pants off.
This is coming back up whether you like it or not.
You all who condone this upload full of stolen content are fucking thieves and I hope when you're taken to court, you lose painfully, and you get all your property stolen.
Archive of Our Own hosts 360,718 stories tagged as "original work", meaning there are at least 360,718 stolen pieces of original fiction that do not fall under any form of "legal gray area" that may be used to justify stealing from writers of fan fiction.
This dataset violates copyright, which Transformative Works has the right to defend on the behalf of the creators who upload to their site.
My works are among the ones scraped and the majority of that is original works (about 2 million words worth of original work and some 500k fanfic) and I have not consented to either being used.
This is not how copyright works. You've published your work without any license or binding terms, and people who have that version can use your work for anything. That is how copyright works. If you don't want your future work to be used for machine training, publish it with a commercial license.
This dataset violates copyright, which Transformative Works has the right to defend on the behalf of the creators who upload to their site.
The Organization for Transformative Works (OTW) is a nonprofit organization established by fans to serve the interests of fans by providing access to and preserving the history of fanworks and fan culture in its myriad forms. We believe that fanworks are transformative and that transformative works are legitimate.
OTW is failing at its core mission, and apparently they don't even remember it. The Organization for Transformative Works claims to preserve fan works and culture, yet their current position directly contradicts their founding principles. By taking actions against the preservation of fan content, they undermine their own stated mission. This "archive" risks becoming lost media the moment OTW ceases to exist, and ironically, the donations they receive are being used to destroy the very history they claim to want to preserve.
You all who condone this upload full of stolen content are fucking thieves and I hope when you're taken to court, you lose painfully, and you get all your property stolen.
This dataset being up on Huggingface or not changes absolutely nothing - crawler bots from big companies and mere individuals have already scrapped the entirety of AO3 at least a dozen times (and they continue to do so ad nauseam as you read this). You've posted to a site that requires no sign-ups or PoW to access and read its contents. If you actually cared, you would've published a book.
Obviously the problem you and everybody else find with this upload isn't that your AO3 works were mirrored on another site (with attribution in the form of author's name no less), it's that it was mirrored on a website for AI/machine learning. It'd be nice if you were honest about this, even if you make it evident implicitly.
Regardless, no amounts of kvetching and bullying will stop people and companies from training on your publicly accessible content.
More highlights from the peanut gallery...
This is not how copyright works. You've published your work without any license or binding terms, and people who have that version can use your work for anything. That is how copyright works. If you don't want your future work to be used for machine training, publish it with a commercial license.
I know arguing with a stranger online over this is pretty pointless but also, copyright tends to be a special interest of mine! I'm not even a fanfic writer but I do have a lot of writer friends and like I said. Special interest. Also, important part of my job.
And, I mean, honestly, you're just wrong. This is exactly how copyright works.
Copyright goes into effect the minute something tangible is created; a story, written down, a drawing, created, whether digitally or on paper or canvas. It doesn't matter where, it just needs to exist, in a tangible form. So ideas that you have in your head or not protected, because you can't prove that you had the idea first, but as soon as you write it down? Bam. copyrighted. You don't need to buy any licence, register with any company, copyright exists for all creative works that, at one point in time, have been... created. Anything you write down, you own copyright on. But here's the tricky thing with copyright: copyright can't be used on anything intangible. An author can complain about someone using similar tropes all they want, but as long as the characters and writing are different, they can't do anything. Because copyright exists only on the tangible work. It becomes copyright infringement once lines are directly taken from the work in another work, and/or (and here we go into the "gray area" that's fanfic) the same characters are used. Specifically names. 50 Shades of Grey is famously a former fanfic; all the author had to do technically to make it okay again is change the names Edward and Bella. But because the story itself was unique, Meyer couldn't really do anything about it.
Now, of course I've brought up the bigger point: fanfic uses material created by others, namely characters and IP, which can be copyrighted. However, the US has a very specific doctrine called "fair use". Fair use is anyone's ability to use a copyrighted material without having to first get permission from the copyright owner. Fair use is decided on four factors: the purpose (whether it's for commercialization, aka to profit from the work, or for personal use) and character of the work (it has to be transformative, aka it can't be a 1 on 1 recreation but has to bring something new to the work), nature of the copyrighted work (you can't copy the text, but ideas etc. cant be copyrighted as per above), the amount and the effect on the work's value. For fanfic, these are some key factors: as long as fanfic is not profitted of, it's not commercial. Authors don't make anything at all, they're not profitting of of someone else's works. Furthermore, most fanfic is transformative; it plays with the storyline, with the characters and brings something new to the already existing piece of media. And lastly, the effect on the work's value. Fanfic does not diminish the work's value in any meaningful way, and thus I'd like to argue that it falls under fair use. And is not copyright infringement (for fair use it usually goes on a case-to-case basis anyways and is not super clear cut, but I think most fanfic does meet these criteria).
However, just because it's fair use doesn't mean that it can just be uploaded or distributed by anyone. Like I said above, copyright goes into effect the minute something tangible is created. Like, I don't know, uploading a piece of writing to a site. There now rests copyright on these specific works. Actually, that copyright went into effect the minute someone started writing it. And given that uploading millions of stories created by others without their permission isn't exactly transformative, since it's a 1 on 1 copy, I would argue that that's not exactly an example of fair use but very much one of copyright infringement.
And to the person saying texts have been used in datasets for ages and ages: you're right. I know. I have multiple degrees in literature speficically, and creating data sets is in fact something scholars do. However, because that's for academic purposes, it falls within the realm of fair use because they do it for purposes of scholarly criticism. They are transformative and bring something new to the text. A GenAI-bot? Yeah no, that just plainly steals an reuses works and words by others. They're not acting as if they own the texts, and not distributing it for the public (also part of copyright; the protection against the work being copied and distributed without permission. Usually, the datasets they use for research are not made available. Only their findings)
If publishers don't want any of their books scraped and mined by AI because it's an infringement of copyright, why would fanfic be any different?
I wanna say source: just trust me, bro, but I don't think that will help in convincing you that this is downright just bad behaviour and that your arguments about fanfic in and of itself being copyright infringement are wrong, so have some actual sources (aside from me dealing with copyright on a day-to-day basis)(and idk maybe if not you, I can at least convince anyone else reading this):
On what is copyrighted and what isn't: https://www.copyrighted.com/blog/copyright-infringement-act and https://www.copyrighted.com/blog/copyright-infringement#what-is-copyright-infringement (fun fact! did you know unpublished works are also copyrighted? it literally only has to exist)
On fair use: https://www.britannica.com/topic/Copyright-Act-of-1790 and https://websites.umass.edu/copyright/fair-use/fair-use-explained/
And, maybe a fun read, on specifically fanfic as fair use and the legality of it: https://en.wikipedia.org/wiki/Legal_issues_with_fan_fiction
However, just because it's fair use doesn't mean that it can just be uploaded or distributed by anyone. Like I said above, copyright goes into effect the minute something tangible is created. Like, I don't know, uploading a piece of writing to a site. There now rests copyright on these specific works. Actually, that copyright went into effect the minute someone started writing it. And given that uploading millions of stories created by others without their permission isn't exactly transformative, since it's a 1 on 1 copy, I would argue that that's not exactly an example of fair use but very much one of copyright infringement.
You've failed to acknowledge that in the US, legal action in a court of law for copyright infringement technically cannot proceed until the infringed work in question has been registered at the copyright office. Hence these DMCAs will simply be thrown out after about 2 weeks of the counterclaim being filed.
Nothing stops someone from registering their work after the fact, but I guarantee you no one who hasn't done so already will. The bulk of these copyright complaints is moot, they're a knee-jerk reaction to AI training and you'll be hard-pressed to find anyone willing to take this to court.
More importantly AO3 themselves have little right to be filing a copyright claim for the ensemble of works that have been scrapped from their site. Their ToS clarifies that users merely grant them a license to display the content uploaded to their website. I can't find any legal jumbo on said ToS that specifies they are granted permission to pursue copyright takedowns on the behalf of their users. To me this DMCA claim smells dubious, if not a product of ignorance.
We received a DMCA takedown notice from the representatives of Transformative Works (link) alleging that materials in this repository uploaded by you infringe their copyright.
When we receive a DMCA notice, we are required to remove access to the allegedly infringing materials.
@nyuuzyou , You may submit a counternotice if you believe the takedown request is unfounded. Please send it over to dmca@huggingface.co and we will forward it to the reporter. A sample counter-notice can be found at https://www.copyright.gov/512/sample-counter-notice.pdf.
Donโt hesitate to contact us if you have any questions.Thank you,
Hugging Face Team
Asking staff: Is there an update on this? Especially considering how many claims have been filed against this dataset? It shouldn't be available by this point.
However, just because it's fair use doesn't mean that it can just be uploaded or distributed by anyone. Like I said above, copyright goes into effect the minute something tangible is created. Like, I don't know, uploading a piece of writing to a site. There now rests copyright on these specific works. Actually, that copyright went into effect the minute someone started writing it. And given that uploading millions of stories created by others without their permission isn't exactly transformative, since it's a 1 on 1 copy, I would argue that that's not exactly an example of fair use but very much one of copyright infringement.
You've failed to acknowledge that in the US, legal action in a court of law for copyright infringement technically cannot proceed until the infringed work in question has been registered at the copyright office. Hence these DMCAs will simply be thrown out after about 2 weeks of the counterclaim being filed.
That is not true, actually. You own the copyright whether you file or not. And we have proof it's ours and was posted before nyuuzyou stole it because it's on the AO3 website.
Nothing stops someone from registering their work after the fact, but I guarantee you no one who hasn't done so already will. The bulk of these copyright complaints is moot, they're a knee-jerk reaction to AI training and you'll be hard-pressed to find anyone willing to take this to court.
More importantly AO3 themselves have little right to be filing a copyright claim for the ensemble of works that have been scrapped from their site. Their ToS clarifies that users merely grant them a license to display the content uploaded to their website. I can't find any legal jumbo on said ToS that specifies they are granted permission to pursue copyright takedowns on the behalf of their users. To me this DMCA claim smells dubious, if not a product of ignorance.
Why the hell do you think we all filed our own DMCAs?
That is not true, actually. You own the copyright whether you file or not. And we have proof it's ours and was posted before nyuuzyou stole it because it's on the AO3 website.
Why the hell do you think we all filed our own DMCAs?
Again, your work cannot be identified (in court) until it has been registered at the copyright office. Also, you and most people did not "file DMCAs". Spamming the discussions page for the item isn't filing a DMCA. For a DMCA claim to be proper it must be sent to HF's legal email with your name, address and contact information so that it may be forwarded to the relevant parties.
The DMCA is only a legal obligation for Huggingface. You have the right to send one, but nyuuzyou filing a counterclaim renders it moot after 2 weeks - unless you take the matter to court and prove it with the documents. Which legally you cannot do if your work isn't registered. Let's be real, are you going to do so?
Itโs true, the comments left here hardly hold much legal ground, as far as Iโm aware. I highly suggest anyone visiting this page email huggingface your own specific DMCA for your own work.
Iโm very far from a legal expert, but AO3 has plenty of lawyers of their own, so using their takedown notice as a template should likely work well:
Physical or electronic signature: (you could potentially paste in an image of your signature here, or you could simply type out your name or username. Using your real name might give you more legal ground to stand on, but since takedown notices could be public I HAVE seen takedown notices done under usernames be honored before. You can also have someone else act on your behalf, so you could try contacting AO3 or someone else and see if they could act on your behalf.)
Contact information: (include your email and other forms of contact)
Work: (I would include the title as well as work ID for all of your individual works, maybe the author as well to emphasize your ownership. I believe the work ID is included in the URL)
URLs: https://huggingface.co/datasets/nyuuzyou/archiveofourown
The use of the listed content is not authorized by the copyright owner(s), their agent, or the law. I state, under penalty of perjury, that I am authorized to act on behalf of the copyright owner; and I state that the information I have provided is accurate.
Thank you,
(Your name/ your username)
That is not true, actually. You own the copyright whether you file or not. And we have proof it's ours and was posted before nyuuzyou stole it because it's on the AO3 website.
Why the hell do you think we all filed our own DMCAs?
Again, your work cannot be identified (in court) until it has been registered at the copyright office. Also, you and most people did not "file DMCAs". Spamming the discussions page for the item isn't filing a DMCA. For a DMCA claim to be proper it must be sent to HF's legal email with your name, address and contact information so that it may be forwarded to the relevant parties.
The DMCA is only a legal obligation for Huggingface. You have the right to send one, but nyuuzyou filing a counterclaim renders it moot after 2 weeks - unless you take the matter to court and prove it with the documents. Which legally you cannot do if your work isn't registered. Let's be real, are you going to do so?
Your work CAN be identified. You can use other things as proof the work is yours. For example, say this did go to court. I can prove that I posted my fic on my account, because it's there, and I can prove the account is mine. Therefore, I can prove the work is mine, even without it being registered. I also have the original file the draft was written on, and the backup copy. You own a copyright even without it being registered.
And I actually WILL be filing a full DMCA. With Hugging Face's host. Since Hugging Face itself refuses to step up and delete the dataset, and nyuuzyou is clearly too much of a coward to admit they're wrong and do the right thing.
And I actually WILL be filing a full DMCA. With Hugging Face's host. Since Hugging Face itself refuses to step up and delete the dataset, and nyuuzyou is clearly too much of a coward to admit they're wrong and do the right thing.
Cool! I recommend helping others out too on our side so we can take this down! Since this is violating copyright law.
People defending this thief are insane. This user is a thief. I don't fucking care if you people think we don't own our works because they're fan works, think of it this way if your brain can't comprehend it but we own our writing. The language we use, our formats, our diction. They're ours. It's a creative work we put time and effort into. It's infuriating because those are ours. Even fanworks belong to us regardless if what it's based off.
We also don't want people claiming they wrote them. We don't want them used to "help" ai either.
Use your brain if you're defending this person.
Your work CAN be identified. You can use other things as proof the work is yours. For example, say this did go to court. I can prove that I posted my fic on my account, because it's there, and I can prove the account is mine. Therefore, I can prove the work is mine, even without it being registered. I also have the original file the draft was written on, and the backup copy. You own a copyright even without it being registered.
And I actually WILL be filing a full DMCA. With Hugging Face's host. Since Hugging Face itself refuses to step up and delete the dataset, and nyuuzyou is clearly too much of a coward to admit they're wrong and do the right thing.
You do not understand how the law operates lol.
Additionally you would have to be an idiot to not simply register your work and then go to court. Not only would it be a monumental task to prove anything without registration (contrary to what you believe), but on the off-chance you won you wouldn't be paid damages.
Oh, but you haven't even filed a real DMCA yet? How naรฏve are you really? Honestly, I'd remove your work on the basis that your idiocy likely translates to your writings. Seems the amateurs are the only ones making a fuss.
Your work CAN be identified. You can use other things as proof the work is yours. For example, say this did go to court. I can prove that I posted my fic on my account, because it's there, and I can prove the account is mine. Therefore, I can prove the work is mine, even without it being registered. I also have the original file the draft was written on, and the backup copy. You own a copyright even without it being registered.
And I actually WILL be filing a full DMCA. With Hugging Face's host. Since Hugging Face itself refuses to step up and delete the dataset, and nyuuzyou is clearly too much of a coward to admit they're wrong and do the right thing.
You do not understand how the law operates lol.
Additionally you would have to be an idiot to not simply register your work and then go to court. Not only would it be a monumental task to prove anything without registration (contrary to what you believe), but on the off-chance you won you wouldn't be paid damages.
Oh, but you haven't even filed a real DMCA yet? How naรฏve are you really? Honestly, I'd remove your work on the basis that your idiocy likely translates to your writings. Seems the amateurs are the only ones making a fuss.
You clearly don't know how copyright law works. "Do we own our copyright if we don't register it" is typically the first question a new writer asks, and the answer is YES. You have to prove it, and I have the proof for mine. We all do. I'm not looking to be paid damages, I'm looking to have my work removed from this dataset.
Considering you're supporting blatant theft and can't seem to understand that you own your intellectual property, I wouldn't talk about anyone else's intellect, if I were you.
You clearly don't know how copyright law works. "Do we own our copyright if we don't register it" is typically the first question a new writer asks, and the answer is YES. You have to prove it, and I have the proof for mine. We all do. I'm not looking to be paid damages, I'm looking to have my work removed from this dataset.
Considering you're supporting blatant theft and can't seem to understand that you own your intellectual property, I wouldn't talk about anyone else's intellect, if I were you.
You do need to register before suing in court, but not before sending a takedown notice to a web host. From the US Copyright Office website, https://www.copyright.gov/512/, FAQs for: "I found my work used online without my authorization":
Do I need to register my work with the Copyright Office in order to request a takedown?
No. A copyright registration is not required before sending a takedown notice. However, for U.S. works, a copyright registration is required before suing in court for copyright infringement.
From my layman's understanding, registration can be done at any time before filing a lawsuit.
I encourage anyone who is harmed by this dataset's publication to consult with a qualified copyright attorney if they have the time and money for it to discuss options, and at least to register copyrights on your works if you have even a passing motivation to pursue this legally if or when the dataset gets republished on a US host (I understand this is a nonzero amount of effort and money, but it allows you to pursue punitive damages if it gets that far, especially since fannish works on AO3 are explicitly non-commercial and so won't have actual damages associated with them, to my understanding).
This dataset violates copyright, which Transformative Works has the right to defend on the behalf of the creators who upload to their site.
The Organization for Transformative Works (OTW) is a nonprofit organization established by fans to serve the interests of fans by providing access to and preserving the history of fanworks and fan culture in its myriad forms. We believe that fanworks are transformative and that transformative works are legitimate.
OTW is failing at its core mission, and apparently they don't even remember it. The Organization for Transformative Works claims to preserve fan works and culture, yet their current position directly contradicts their founding principles. By taking actions against the preservation of fan content, they undermine their own stated mission. This "archive" risks becoming lost media the moment OTW ceases to exist, and ironically, the donations they receive are being used to destroy the very history they claim to want to preserve.
WE OWN OUR WRITTEN WORKS. We get to choose if we post or remove them. Part of that is so that if there is a reason that author no longer wishes for it to be accessible then thatโs our choice. The part of why the preservation part came about was because of sites like ff.net deleting lgbt works. Donโt quote the guidelines to me as thought I donโt know why they were written. I was there. Delete your illegally copied โarchiveโ?and stop trying to take the rights of artists and writers for your own accolades.
Stop stealing from us.
Your 'work' is stolen from copyright holders in the first place. Unless you claim to own Naruto, One Piece, and Harry Potter?
How is it that people like you are so engaged with open source, data and AI but cannot google how copyright works before embarrassing yourself online like that? If google is too hard, use chat GPT.
Stop stealing from us.
Your 'work' is stolen from copyright holders in the first place. Unless you claim to own Naruto, One Piece, and Harry Potter?
How is it that people like you are so engaged with open source, data and AI but cannot google how copyright works before embarrassing yourself online like that? If google is too hard, use chat GPT.
Itโs probably that they did and forgot chat gpt makes up a hallucination of the world we fed them.
NOTHING THESE DATA PROCESSORS DO IS TRUE UNLESS CONFIRMED ELSEWHERE.
Also the reason why fan work is possible is the transformative nature. You didnโt change these fics you stole, you grabbed them whole sale. Fanfiction of a fanfiction is actually pretty popular, and if they took my story and no chatgbt, no bot, just from the dome made a new one Iโd be flattered actually.
I'm in a teams chat with several other people working on LM models and they've simply downloaded the torrent (which you can't get rid of) and set up a bot to automatically scrape any new published work on the site. You can't do anything about this so I'd recommend moving on with your lives and let adults work.
I did not consent to my work being included in this dataset. One of them is fully original and therefore you are guilty of copyright infringement. Remove work 123051553 from your dataset or be subject to legal action.
It's been almost 20 days, you were saying about this being up again in 2 weeks? Methinks yall are in the fucking wrong, and you're in denial about it. Stop stealing our work, learn how to write your damn selves, and get a LIFE.
It's been almost 20 days, you were saying about this being up again in 2 weeks? Methinks yall are in the fucking wrong, and you're in denial about it. Stop stealing our work, learn how to write your damn selves, and get a LIFE.
But the torrent is up on datafish, most of us wouldn't even have found it if you guys didn't link to it on the subreddit for clout ๐ฅน
I'm pretraining Gemma 27B on it, and while I expect its writing to degrade, we'll see: it may be the ultimate fanfiction generator in a box!!!!!
It's been almost 20 days, you were saying about this being up again in 2 weeks? Methinks yall are in the fucking wrong, and you're in denial about it. Stop stealing our work, learn how to write your damn selves, and get a LIFE.
But the torrent is up on datafish, most of us wouldn't even have found it if you guys didn't link to it on the subreddit for clout ๐ฅน
I'm pretraining Gemma 27B on it, and while I expect its writing to degrade, we'll see: it may be the ultimate fanfiction generator in a box!!!!!
oh look another throw-away account of yours to stir up bs.
But the torrent is up on datafish, most of us wouldn't even have found it if you guys didn't link to it on the subreddit for clout ๐ฅน
oh look another throw-away account of yours to stir up bs.
Bold of you to assume I heard about it on reddit, that place is a shithole I wouldn't touch with a stick. You're just too lazy to learn how to write fic. Also, I'm a completely unrelated party to whoever has been previously posting in here. There are a lot of very real people upset about this, not just the ones who have been making it known in here.
If you see fanfic as this far beneth you, why use it to train your ai model?
Why use something that you value less than dirt for your ai? You seem to see fanfic as copyright infringement as well, so why use works that don't abide by copyright law to train your ai model? Won't that subject you to legal trouble as well? Instead of scraping archive websites, why not reach out to authors personally for their writing to help train ai, which may go over much better than stealing writing without any warning
TL;DR: If you see fanfic in such a light, why steal it?
@nomad-01 No offense to you and your team, but why the hell is this freakshow still allowed to be on the site? Why on earth are they even allowed to be on your platform? This prick and scum of earth is sharing stolen works across torrents using your platform, and all you have to say about it is "Oh yeah, we received this notice, so submit a counter notice when you have the time"??? Seriously??? What more proof do you need? Do you really not see all the degenerates on this site, the actual lowlifes who can't use your platform ethically? Pathetic and disgusting of both you and them.
you guys didnt just steal writing by the way. you stole creativity. motivation. happiness. i dont think im ever going to create again cuz its very very likely all of my works are in this dataset. im even considering taking them down.
Amazing learning from the model during the continued pretraining step!!!!
Wasn't sure it'd work given the variety but this dataset is cleannnnn sheeeeesh!
A layer of SFT and I think we might have ourselves a winner!
Remove my data shit lips. Remove all of our data shit lips.
I'm on the Ao3 side of things and I've been monitoring the situation here for days. They're not going to remove it. The people who downloaded the dataset initially, aren't going to delete it. Even if the dataset gets taken down, the torrent is up by now. And torrents are as good as permanent.
I don't think they care about what we have to say anymore. They never did to begin with. That's why they stole our works in the first place. They do not respect us.
And it's sad because I feel it would've been so easy to just...let people opt their works out of the Ao3 dataset. Copyright laws set aside, just to do it out of respect for people's works. I think that's what's been so upsetting about this whole thing--it's not the question of copyright, but the blatant disrespect for fan-content. But people who don't see fan-works as the beautiful, creative passion projects they are, never will. They're too hard-set in their own learned views and biases. Granted, they probably see us all as petulant children now, and that's sad too, in its own way. They're people who live day-by-day never understanding how little things can bring people so much joy.
This'll probably be my last post up here. Like everyone else, my takedown request was ignored. I'm tired of arguing with people. I don't know what happens next. But please, my fellow Ao3 writers, don't stop writing. Never stop writing. Ai will be around for years to come, this is only the first battle of many. Keep writing and keep fighting, even if it's not here.
@nyuuzyou @grishymishy stop stealing peopleโs copyright. take down the dataset. youโre single-handedly destroying a community of writers who pour their hearts and souls into this. ALL my works are in your set and you donโt have permissions with my data/works. AI is going to destroy us and the arts as it is, so how willingly you bow down to it to train it with the writing of actual people is disgusting. i hope you never experience anything like this, how you just feel gutted because people stole something you worked hard on, poured your soul into. to some of us itโs a livelihood to write and I do it so others can find reprieve in my works as they read. screw AI. And screw AI bros, hopefully one day you see how terrifying and hurtful this is.
I'm in a teams chat with several other people working on LM models and they've simply downloaded the torrent (which you can't get rid of) and set up a bot to automatically scrape any new published work on the site. You can't do anything about this so I'd recommend moving on with your lives and let adults work.
youโre disgusting. like this is fucking insane btw, like jaw dropping. itโs giving โiโm miserable and like pissing people off with nothing better to do.โ
Amazing learning from the model during the continued pretraining step!!!!
Wasn't sure it'd work given the variety but this dataset is cleannnnn sheeeeesh!
A layer of SFT and I think we might have ourselves a winner!
Could you please explain what exactly is going on? Is it a LoRA being trained or a whole new model? Also, please let me know if it produces quality output. I doubt it will compete with Claude 3.7 Sonnet, but it would be fun to mess around with.
Thanks!
@MindlessForMinerva LEAVE our damn works alone. We did not consent to this. Why is that SO hard to understand for you spineless cowards?
@MindlessForMinerva LEAVE our damn works alone. We did not consent to this. Why is that SO hard to understand for you spineless cowards?
Well Iโm an author on AO3, and I consent to this!
@MindlessForMinerva
Then you consent to YOUR works being scraped. Leave the rest of us alone. Surely that can't be hard to understand for someone who dabbles with words everyday.
Rapists use the same logic to do what they do; doesn't mean they're right. Get your priorities right and if you have a moral compass at all. LEAVE. US. ALONE. What is SO hard to understand about leaving people - people who did NOT consent to this - out of your work??? We didn't do anything to you, why do you want to take something that's so personal to us??
Amazing learning from the model during the continued pretraining step!!!!
Wasn't sure it'd work given the variety but this dataset is cleannnnn sheeeeesh!
A layer of SFT and I think we might have ourselves a winner!
Is there any fellow AO3 authors willing to do a joint lawsuit? We have copyright of our works. We could sue nyuuzyou and if grishymishy decides to post this we could sue them as well.
Is there any fellow AO3 authors willing to do a joint lawsuit? We have copyright of our works. We could sue nyuuzyou and if grishymishy decides to post this we could sue them as well.
Lets wait to see what comes of the OTW legal team's negotiations with HuggingFace. PaperDemon is posting regular updates about the situation, and the most recent news post on ao3 has comments by the official ao3 account that says they can't officially speak about the situation yet.
This is why we donate past what they're asking, our money is being put to good use defending us as we speak. So let's leave these incel neanderthals to their mud pies, and keep an eye out for updates on PaperDemon and Ao3's social medias.
Is there any fellow AO3 authors willing to do a joint lawsuit? We have copyright of our works. We could sue nyuuzyou and if grishymishy decides to post this we could sue them as well.
There is no proof of a model trained on this dataset as of now. It is a screenshot of a model training session that has a model trained with a dataset that had that name, which means nothing because you can rename files. Good luck. If you think any kind of law enforcement will pursue an investigation of IP rights because of a screenshot, we're completely fucked because of the power corporate world holds. That day is not today, and I'd rather live in a world that scenario is just absurd.
@grishymishy How much data are you using?
A touch under 1B tokens for the pretrain!
Generating synthetic instructions for the SFT step now and aiming for 10k examples to make it steerable
Is there any fellow AO3 authors willing to do a joint lawsuit? We have copyright of our works. We could sue nyuuzyou and if grishymishy decides to post this we could sue them as well.
It's weird how you can tell how a lot of the comments here are more about having an outlet for general life frustrations and feelings of powerlessness completely outside of this dataset...
A lot of you seem to be straight up singing in deepest parts of your soul at getting a chance to show out! Rallying yourselves to keep opening tickets here feels very "all hands on deck" right?
Let me see if my fic is in there, then I'll really be a part of this!
WE'RE GONNA SUE THEM! REMEMBER THE TIME AO3 WENT TO COURT?!
The truth is some of y'all are getting off on increasingly loud screeching about "just leaving us alone" because it means someone's considered you attacking, and you're getting to band together: it's us being attacked! I'm doing this to protect us guys! It's actually a safety issue guys!
That's why you guys were posting all over the place about it (including Reddit)... and instead of censoring the location and dataset name to limit reach, you linked it every single place you could! Google refuses to list the sketchy russian site where this dataset is currently hosted unless you enter it by name so odds are I wouldn't have found it: but you guys made sure everyone knew where every single copy of it was!
Imma be real with y'all for a second: AO3 was in Common Crawl for years. AO3 didn't leave common crawl until post-ChatGPT when everyone started blocking Common Crawl and ironically AI writing started infecting Common Crawl... so most foundation models are probably using versions of common crawl that have all your fics. Like all your fics.
But truth is
a) your writing is a ridiculously small percentage of the writing there
b) a ton of your writing is probably getting filtered out for moderation and quality reasons
c) it's uh... unique writing, to be kind, by most people's standards (I'll explain why I'm still training on it in a bit)
I'm not training this model because it will be good at writing, after all this dude said it:
I doubt it will compete with Claude 3.7 Sonnet
It won't! It will definitely unlearn some basic grammar, it's not going to have a single non-horny braincell left when writing, and it's going to learn some very unfortunate concepts without careful filtering...
So why make it?
Because there is one specific segment people who will use a model post-trained on this... fellow fanfiction lovers. ๐ฑ
The loudest people in the community are screaming about AI and how harmful and evil it is, but the quiet majority who want to read stories with characters do not care.
- If AI can write about their favorite rarepair with half the competency of a human, they'll take it.
- They're already using Character.ai and clones that were trained on real people's content
- If AI can do it cheaper than a commission, most (not all) will take that
- They don't hate AI art for ideological reasons, they hate AI art because it's been bad and will 100% take if it's good!
The only reason I personally would not train this model, is if I thought no one would want it even if it trains successfully.
But we all know a lot of people want their personalized content no matter how much a vocal minority wants to scream about it, because they just want their fics and don't care about the drama. That is what the people yearn for, the ones who won't show up here, and I am nothing if not a woman of the people ๐ซก
I'm not @nyuuzyou but I imagine their motivation is probably pretty similar: they didn't scrape this because they care enough about AO3 to actively try and harm it... they did it because they thought it'd be neat, and they thought someone could use it.
And they were right! >:3
people who make things because it'd be neat ๐ค people who make things because it'd be neat
I think it's time to clear out folks. This round was lost to the boneheaded, bullheaded, selfish, evil tinkerers like myself who are just gonna tinker.
I personally would focus on trying to ensure that my community is as welcoming and low-stress as possible at this exact moment for the folks still in it, not trying to increase the level of stress that comes from being a part of it.
You guys are really trying to turn up the pressure cooker here and get people feeling super angry and motivated about having their fics stolen, and rally a return volley... but what you're doing might also really badly timed ๐ฌ
- AI is getting better
- People are already under a lot of stress because <gestures vaguely at the state of the world>
- The community was already getting a bit larger so the quality of human interactions was also slipping just a tad: https://www.reddit.com/r/AO3/comments/1amlvb4/comment/kpmsbqo/
The main thing that human generated fanfiction has going for it is the human interactions!
You're adding to the stress to the community around the same time that AI is slowly starting writing less like it's picking similes out of a shallow pond, but if the community starts to falter there won't be much reason left to even interact with human generated fics at all! So what you're doing might end up just being a glancing blow to the community aspect of fanfiction when it's already staggering a bit ๐
I mean think about how many people went through this in the last week:
ugh man I'm so stressed let me go check on AO3
oh no AO3 is on fire, what is this new AI dump thing happening
well let me just go read my favorite comfort fic to get away for a bit
IT'S GONEEEEE THE AUTHOR HID IT BECAUSE OF THE AI STUFF
Now where you think this is headed is: I hate those AI freaks so much, I'll never use AI! ๐ก * commissions a fic of their favorite ship *
Where this is really headed: Welp, I don't even have money for eggs... maybe I'll try using ChatGPT to write my own fic so that never happens again. Wait a minute... this actually... writes pretty well? It's no Shakespeare, but hmm. Oh woah and these images are pretty good!
For now maybe the guardrails on what it writes might be a problem, but alternatives that aren't censored are getting better and more common...
If I were you guys I would work super overtime trying to manage fears, reduce the stress people feel over this, etc. AO3 should put out a statement pointing out there's avenues to deal with it if a model does produce their content later and that people should avoid sharing the dataset links. Make it clear that sharing your fics with the community is still safe and encouraged. Double down on the fact that a lot of the engagement from the other side is bad faith meant to rile you guys up!
You should band together inward not outward and use this as a chance to emphasize why no matter how good AI gets, or what these evil AI freaks do, they can never replace the living breathing hearts that make us a community ๐ฅน
This could be a rallying point for all the newbies who've joined in recent years: there could be that time AI tried to beat AO3 but AO3 came back stronger!
But fearmongering and telling people to spam DCMAs and yell at the AI folks who are mostly going to do their thing anyways is just not how you'll get there. Instead people are going to retreat from it all... and some already are.
Well anyways, uh yeah, thanks for coming to my TED X talk. I've got models to bake.
@grishymishy You know, you fucking moron, all the person would actually need to do is sign up for an account on ao3, it's free, we can't fucking make money off of fanfics so we wouldn't do commissions, asshole. That would get us in fucking court and sued because it breaks fair use laws for the standards set by lawsuits and legal debate.
We wouldn't fucking ask for or offer out commission, moron! This isn't fanart. For us, it's genuinely illegal, just as fucking illegal as this shit.
@grishymishy You know, you *********, all the person would actually need to do is sign up for an account on ao3, it's free, we can't ****** make money off of fanfics so we wouldn't do commissions, *****. That would get us in ***** court and sued because it breaks fair use laws for the standards set by lawsuits and legal debate.
You shouldn't call me a m***n and miss the entire point, you gotta choose ๐
These people are already on AO3, but they're watching their favorite stories disappear and the community they turn to for comfort have a meltdown because a minority is trying to whip up fear and rally the troops... that's why they'll withdraw :(
We wouldn't f*** ask for or offer out commission, moron! This isn't fanart. For us, it's genuinely illegal, just as ***** illegal as this shit.
We really going to pretend this has ever in the history of fanfiction/fanart stopped commissions?
https://adashofstarshine.tumblr.com/post/185070627079/im-opening-up-fanfiction-commissions-i-am-a
https://finally-isaac.tumblr.com/post/627686680409915392/writing-commissions-open-you-will-receive-your
https://everskies.com/club/drawing-commissions/forums/general/cheap-fanfic-commission-prices-45519618
https://fanficsandfluff.tumblr.com/post/652601105711284224
https://www.fiverr.com/goldengukk/do-bts-fanfic-commissions
https://ko-fi.com/conetsu/commissions
https://www.deviantart.com/chittatos/art/Fanfic-Writing-Commission-859867996
https://www.wattpad.com/886085318-fanfiction-commissions-information
https://ko-fi.com/s/74606f98d4
https://aminoapps.com/c/disney/page/blog/closed-claude-frollos-fanfiction-commissions/5BM0_1MWtVuaW8JoM4VgZ2aGKD1DZ2r5aop
https://www.fiverr.com/archonicwriting/do-writing-commissions-original-or-fanfiction
https://www.etsy.com/listing/1661986368/1000-words-personalized-writing
https://www.deviantart.com/nope0101/journal/Writing-Commissions-are-now-open-2024-1005545180
https://bsky.app/profile/ourownsidefund.bsky.social/post/3lh4kypkfnc25
https://aminoapps.com/c/voltron-legendary-defenders/page/blog/starting-writing-commissions/BwZe_Xkswuoqx5o4bZ0powQKM25Pon3x0J
https://www.etsy.com/listing/1747223588/custom-writing-commission-fanfiction
https://ko-fi.com/fateboundauthor/commissions
https://www.facebook.com/NihilistAceWriting/?locale=es_LA
https://www.patreon.com/worldofsouls/about
https://www.patreon.com/Antony444
https://www.patreon.com/user?u=93048562
https://www.patreon.com/hpfanfictioner66/about
https://www.patreon.com/CaptMermain/about
[Thousands and thousands more links...]
Sorry if I gave you the wrong impression but I'm not that removed from fics :)
Despite the law, many many creators make money off fanfiction and just rely on being small enough to not matter. Or they try adding a layer of indirection like taking donations separately of providing access to the art... even though as far as the law cares "donations" are commercial activity.
Plenty of people don't do that, so please don't take this as me producing an excuse to yoink your fics: but let's all also be honest with each other right?
Honestly the whole situation sounds a little bit like the evil AI freaks ๐ค
@DistressedAuthoress maybe don't encourage the AI scrapers to make ao3 accounts. I'm not totally sure how the scraping itself works, but they'll be able to see the locked fics if they make an account. The OTW is currently taking steps to make it harder for scraping to happen, they put some details in the comments section of their latest news post.
Clearly, things are happening that are preventing HuggingFace from reinstating the data here, since it's been over two weeks since the counterclaim was filed and this dataset is still disabled. I recommend you lock your fics, and hide and wait for updates from PaperDemon or Ao3's socials. These people aren't going to stop, they're trying to rile us up and get a reaction, because thats what trolls do.
They don't care about the art form, just consumption. The best we can do is take steps to make it more difficult for them to access our work, and let the OTW legal team handle it.
Delete this data set shit lips
Delete this data set shit lips
Rather aggressive
Delete this data set shit lips
No, hope this helps :3
I will be doing this every day I remember to so o7
But hope you step in dog shit to you too random citizen
But hope you step in dog shit to you too random citizen
JFSJDJGDGJSHSHTFDJGDHKFKHFK
Delete this data set shit lips
No, hope this helps :3
I will be doing this every day I remember to so o7
But hope you step in dog shit to you too random citizen
Reported. Also, itโs โcopyrightโ not โcopywriteโ.
@MindlessForMinerva
You're free to associate with whomever you wish, but I'd be remiss not to point out that people who aren't terminally socially deficient/maladjusted IRL generally don't tend to work this hard to defend being cretinous to people.
Food for thought.
Delete this data set shit lips
No, hope this helps :3
I will be doing this every day I remember to so o7
But hope you step in dog shit to you too random citizen
Reported. Also, itโs โcopyrightโ not โcopywriteโ.
? Okay, report me then. Iโll still keep telling you to remove my data. Iโll be rude about it too.
@Ihateao3copywriteinfringment Keep going. I hear 99.9% of bullies give up their harassment campaigns one insult before their target does what they want.
@DistressedAuthoress maybe don't encourage the AI scrapers to make ao3 accounts. I'm not totally sure how the scraping itself works, but they'll be able to see the locked fics if they make an account. The OTW is currently taking steps to make it harder for scraping to happen, they put some details in the comments section of their latest news post.
Clearly, things are happening that are preventing HuggingFace from reinstating the data here, since it's been over two weeks since the counterclaim was filed and this dataset is still disabled. I recommend you lock your fics, and hide and wait for updates from PaperDemon or Ao3's socials. These people aren't going to stop, they're trying to rile us up and get a reaction, because thats what trolls do.
They don't care about the art form, just consumption. The best we can do is take steps to make it more difficult for them to access our work, and let the OTW legal team handle it.
Sorry, not trying to do that, these assholes keep trying to say that fans will go for ai crap instead of doing cheaper things. (Besides this asshole apparently already got locked works before switching to only public works.)
And I'll try to do what I can. Thanks, I've been just actually doing with stuff. Well, more of actually warning others cause alas my traffic comes mainly from guests most of all alas.
@DistressedAuthoress I'm sure readers love it when their reading experience gets made worse. Making a fic registered users only doesn't affect me, since I have an account, but it does suck that some authors are deleting their works altogether. I wonder if they realize they're just making these datasets more valuable.
@DistressedAuthoress I'm sure readers love it when their reading experience gets made worse. Making a fic registered users only doesn't affect me, since I have an account, but it does suck that some authors are deleting their works altogether. I wonder if they realize they're just making these datasets more valuable.
The ones I have seen are reuploading them after deletion. But maybe authors wouldn't feel the need to take extreme measures to keep their works out of the hands of ai scrapers if ai scrapers stopped training their bots off our works. The actions of the few are affecting the many, and in this case it's the fault of the people who do despicable shit like this who are making it harder for everyone to enjoy the fanwork side of fandom.
It would genuinely be much, much easier for everyone if they just stopped. All of the gatekeeping of works behind blue locks would end if they stopped stealing. It's literally their own faults this is happening.
@NauticNine Yup, it's the faults of these assholes as to why people are either terrified of uploading or deleting their works, and some will upload it again, but others will leave fandom completely. Thank you for understanding. These theives here are nothing but despicable people who hate genuine creativity.
@DistressedAuthoress I'm sure readers love it when their reading experience gets made worse. Making a fic registered users only doesn't affect me, since I have an account, but it does suck that some authors are deleting their works altogether. I wonder if they realize they're just making these datasets more valuable.
Ah yes, this dataset is totally valuable and not the exact reason why people are leaving fandom and fanworks or not joining in. If these assholes didn't do this shit of stealing content for their crappy VIs (cause they're nothing like actual AI), people wouldn't go to delete it. And besides some people have to take down fanfics because those were reworked into original works and the publishing house doesn't want it uploaded anymore, oftentimes for copyright reasons. But good to know you're against creatives trying to protect what they created. These asshole theives are the ones who are making reading experiences worse, not the authors.
@NauticNine Yes, how despicable. How dare someone spend their own money collecting and hosting a dataset, to then freely give it away to whoever wants it. The poor authors (myself included!), who have to deal with the harm of knowing someone else has a copy of their fanfic. A fanfic they wrote and shared with no expectation of reward. Will this misery never end?
@NauticNine Yes, how despicable. How dare someone spend their own money collecting and hosting a dataset, to then freely give it away to whoever wants it. The poor authors (myself included!), who have to deal with the harm of knowing someone else has a copy of their fanfic. A fanfic they wrote and shared with no expectation of reward. Will this misery never end?
The problem is that no permission was given. There was no opportunity to opt out, it was just taken. AI is already putting people out of real jobs from being trained on datasets made on the backs of authors and artists. It's theft. It's as much theft as it is reposting someone's work on another site and claiming it as your own. It is theft, case closed, and you should be ashamed of yourself.
The problem is that no permission was given. There was no opportunity to opt out, it was just taken. AI is already putting people out of real jobs from being trained on datasets made on the backs of authors and artists. It's theft. It's as much theft as it is reposting someone's work on another site and claiming it as your own. It is theft, case closed, and you should be ashamed of yourself.
It's as much theft as downloading a fanfic from AO3 using the "Download" button. I don't have to ask your permission before I do that, do I?
@NauticNine I respect your commitment to talking sense, but I wouldn't waste your time on this lot. They love to pretend that situational context doesn't exist, like there isn't a difference between someone downloading a fic to solely read it, and someone using it for future replication purposes.
People in the right don't have to keep bending over backwards to justify how they took things that aren't their work to use in their little projects that the real creators are explicitly against. So they revert to either playing dumb, or just embrace their deficiency/inability. They can't do anything without resorting to violating other people's consent, and they're aware of it (on some level).
The problem is that no permission was given. There was no opportunity to opt out, it was just taken. AI is already putting people out of real jobs from being trained on datasets made on the backs of authors and artists. It's theft. It's as much theft as it is reposting someone's work on another site and claiming it as your own. It is theft, case closed, and you should be ashamed of yourself.
It's as much theft as downloading a fanfic from AO3 using the "Download" button. I don't have to ask your permission before I do that, do I?
no, but you DO have to ask and get permission if you want to upload it somewhere else.
It's as much theft as downloading a fanfic from AO3 using the "Download" button. I don't have to ask your permission before I do that, do I?
There is a distinct difference between downloading a fic for personal use offline, like for an e reader, and downloading it to train generative ai to replicate it. One is personal use, the other is blatant theft.
Now, one could argue that attempting to replicate the writing style of another author is the same thing, or using their sane idea to make something of your own, but that is still different. No one can perfectly replicate another person's writing style, because humans are naturally fallible. Machines aren't perfect either, but they could be. They will rip text word for word out of the works it's trained on and not add anything new, because it can't. What it's making isn't art, it's forgery.
@DistressedAuthoress I'm sure readers love it when their reading experience gets made worse. Making a fic registered users only doesn't affect me, since I have an account, but it does suck that some authors are deleting their works altogether. I wonder if they realize they're just making these datasets more valuable.
if this AI dataset never existed, i wouldn't have locked my fics. i wouldn't be considering taking them down. any fics being mass-deleted or locked right now is a direct response to this dataset being made. a direct response to someone violating our consent
@DistressedAuthoress Ah yes, this dataset is totally valuable and not the exact reason why people are leaving fandom and fanworks or not joining in.
They're not saying it's good that people are scrapping the words to train models, but that fic deletion make these datasets more valuable since people looking for the content (their favorite fic, now deleted, for example) could be looking to get their hands on the dataset that has a copy of it.
This is nothing new, when authors delete fics, people will ask for anyone that has downloaded a copy or use archive/mirror datasets. I, too, have a copy of AO3 downloaded because I like preserving fics for this reason. A copy I've had for sometime now, that I don't use for anything other than personally reading fics that have been deleted. (Of course fic deletion is the right of any author, I'd be silly to get angry at them for doing what they want with their work).
Anyways as much as it pains me to admit it @grishymishy isn't that far off (sidenote: I've gotten comments in my fics telling me I'm the reason people use AI because I take too long to update). Personally, I will be doubling my efforts of commenting and interacting with fic writers and the community. Because I do not believe AI will ever truly replace what human authors can make you feel, and because it pains me to see so many people distraught and distancing themselves from fandom because of this.
Before AI got it's recent boom with ChatGPT launching, I remember I tried to train the models that were out there to write fanfiction (and write my own shitty ones that didn't produce a sentence lmao) and dream of the day LLMs could produce good content. One of my firsts prompts to ChatGPT was to see how it handled creating fanficfion. Now, it's just so sad to see what it has become. I wouldn't say it's "blatant" theft (afterall the reason we are here is the lack of proper legislation around these issues)...but isn't it terrible how content consumption at all cost has driven us to accept soulless work and disregard how it harms human creators? Maybe I was too naive to see how it would all play out.
Sending a hug to every author that is reading this thread, too โค๏ธ
like there isn't a difference between someone downloading a fic to solely read it, and someone using it for future replication purposes.
Who says this has to be used for "replication purposes"? Maybe I just want an offline version of AO3 to read when the power goes out. Maybe I can put it on a drive and give it to the Pope like MatPat did with Undertale. Maybe I want to create an automatic tagging system for all the writers who can't be bothered to tag their own works. Maybe I want to make a recommendation system to show me fanfics I would have otherwise missed. Maybe I want a safeguard against the current administration banning AO3 for being too queer. Hell, maybe I do want to make an AI that can write smut, just because none of the big dogs have the guts to. The possibilities are endless!
Now, one could argue that attempting to replicate the writing style of another author is the same thing, or using their sane idea to make something of your own, but that is still different. No one can perfectly replicate another person's writing style, because humans are naturally fallible. Machines aren't perfect either, but they could be. They will rip text word for word out of the works it's trained on and not add anything new, because it can't. What it's making isn't art, it's forgery.
I don't know what any of this means.
- Fanfics aren't theft because humans aren't perfect - ok, I can see that.
- Machines aren't perfect, but it's still theft when they do it - you've lost me.
- Machines could be perfect - God I wish. But writing isn't like math - there isn't one "perfect" word to always follow the last. Quality is subjective.
- Machines don't add anything new - then what are you afraid of?
- What it's making isn't art, it's forgery - no, it's derivative. One could even say it's ... transformative. Millions of words go in, a latent space is formed, patterns are learned, and out comes something not quite like anything ever seen before. A brand new creative work.
if this AI dataset never existed, i wouldn't have locked my fics. i wouldn't be considering taking them down. any fics being mass-deleted or locked right now is a direct response to this dataset being made. a direct response to someone violating our consent
Deleting your fics has no impact on anyone who downloaded this dataset. Nor does it affect OpenAI, Google, or any of the other big companies who scraped your work long ago. All it does is make AO3, the site we all love, worse. And as I mentioned, every user sad to see your deleted fic is one more user interested in this dataset.
Personally, I will be doubling my efforts of commenting and interacting with fic writers and the community. Because I do not believe AI will ever truly replace what human authors can make you feel, and because it pains me to see so many people distraught and distancing themselves from fandom because of this.
This x 1000. AO3's strength isn't the quality of the fanworks, or how resilient it is to scraping. It's the community of passionate individuals daring to create. To put their works out for the enjoyment of others, never expecting a reward. Who cares what some AI enthusiasts do with your fics? Your readers want you.
Gooood evening shit lips. Delete this shit.
For the community: I have filed a counter-notice in response to the DMCA takedown of my dataset. According to the DMCA process, the dataset should be restored within 10-14 business days unless OTW takes legal action. I believe the takedown request was unfounded, and I'm working to make this dataset available again as soon as possible.
Hi, would you mind explaining? There's a lot of folks arguing and I want to know what your actual thoughts are.
I will freely admit that I am one of the authors from whom you used content but I am not actually opposed to it. As a compsci major I just want to know the logic and purpose. A lot of the data gathered will be porn, poorly written or other. Have you got a way to sory that out? Have you stipped the writing completely out? I know it says that each entry is one line in the JSON file but after working with data like that myself it's a bit of a nightmare.
For language models would it not have been better to go with a social media site?
To be honest if you plan to use this data for analytics I would love so see what you find, AI useage doesn't just mean generative as so many people now think.
The major issue that I've not actually seen anyone mention is that legally the works belong to AO3/OTW themselves, much the way all of your tweets belong to X/Twitter and posts belong to Facebook. The grey area comes from the non-profit of it. I'm sure you're aware of the laws given the counter claim you've filed so I trust that you know what you're doing but I would like to hear it from you anyway.
Have a nice day ^u^
Who says this has to be used for "replication purposes"? Maybe I just want an offline version of AO3 to read when the power goes out. Maybe I can put it on a drive and give it to the Pope like MatPat did with Undertale. Maybe I want to create an automatic tagging system for all the writers who can't be bothered to tag their own works. Maybe I want to make a recommendation system to show me fanfics I would have otherwise missed. Maybe I want a safeguard against the current administration banning AO3 for being too queer. Hell, maybe I do want to make an AI that can write smut, just because none of the big dogs have the guts to. The possibilities are endless!
Notaspy1234 You hit the nail and issue on the head. The possibilities are endless. Fanfiction works on a grey zone because it's not monatised. As the dataset is currently free if it was used at all the issue isn't major.
But. And this is a very big but.
What happens when someone starts charging for a product made out of it? Made out of something no one had the copyright for. Not you, not me. Do you rememer the beginning of fanfiction? How many sites crashed and burned due to authors and their publishers suing writers (notebly Anne Rice) who couldn't fight back because they didn't have the money? Do you remember the fairly recent goings on with both Games Workshop (Warhammer 40K) and Wizards of the coast (D&D)? Do you remember just how much fan content we lost because of it?
The issue isn't replication.
The issue is the potential monitisation of content that won't just hurt you but will burn the community near to the ground.
I have no issue with being used for sample or training data. As proof of concet I want to see it go as far as it can. But you can't promise me someone won't sell it and take us all out.
Can you promise me that we're both safe?
If you want I can cite sources and laws but I don't think I need to. I'm just hoping you have a better answer than what I think.
Hi,
I've already explained how the data can be used: https://huggingface.co/datasets/nyuuzyou/archiveofourown/discussions/195#680e2ac7a0520e2b7ececb18
And also OTW themselves have not claimed copyright on user-generated content: https://archiveofourown.org/tos
The OTW does not claim any copyright in or ownership of your Content. We repeat: we do not own your Content. Nothing in this agreement changes that in any way. However, running AO3 requires us to make copies, and backup copies, on servers that may be located anywhere around the world.
This was the main reason for the counter-notice, as well as the fact that the dataset does not contain any content that is directly owned by the OTW (design, site code, etc.).
Hi,
I've already explained how the data can be used: https://huggingface.co/datasets/nyuuzyou/archiveofourown/discussions/195#680e2ac7a0520e2b7ececb18
And also OTW themselves have not claimed copyright on user-generated content: https://archiveofourown.org/tos
The OTW does not claim any copyright in or ownership of your Content. We repeat: we do not own your Content. Nothing in this agreement changes that in any way. However, running AO3 requires us to make copies, and backup copies, on servers that may be located anywhere around the world.
This was the main reason for the counter-notice, as well as the fact that the dataset does not contain any content that is directly owned by the OTW (design, site code, etc.).
Oh I am 100% behind this. This is such a good idea! I hate that when people see AI they think of generative now >:/
Hi,
I've already explained how the data can be used: https://huggingface.co/datasets/nyuuzyou/archiveofourown/discussions/195#680e2ac7a0520e2b7ececb18
And also OTW themselves have not claimed copyright on user-generated content: https://archiveofourown.org/tos
The OTW does not claim any copyright in or ownership of your Content. We repeat: we do not own your Content. Nothing in this agreement changes that in any way. However, running AO3 requires us to make copies, and backup copies, on servers that may be located anywhere around the world.
This was the main reason for the counter-notice, as well as the fact that the dataset does not contain any content that is directly owned by the OTW (design, site code, etc.).
Clearly you don't understand how this works. They don't own our work but we do. We've said several times we don't consent to you taking our stuff. However because you scraped all of AO3 the OTW legal team is looking into it to protect the company. You could be considered a Cybersecurity risk, a harassment, and copyright infringement. I want your email as some of us are filing lawsuits for copyright against you.
Hi,
I've already explained how the data can be used: https://huggingface.co/datasets/nyuuzyou/archiveofourown/discussions/195#680e2ac7a0520e2b7ececb18
And also OTW themselves have not claimed copyright on user-generated content: https://archiveofourown.org/tos
The OTW does not claim any copyright in or ownership of your Content. We repeat: we do not own your Content. Nothing in this agreement changes that in any way. However, running AO3 requires us to make copies, and backup copies, on servers that may be located anywhere around the world.
This was the main reason for the counter-notice, as well as the fact that the dataset does not contain any content that is directly owned by the OTW (design, site code, etc.).
Clearly you don't understand how this works. They don't own our work but we do. We've said several times we don't consent to you taking our stuff. However because you scraped all of AO3 the OTW legal team is looking into it to protect the company. You could be considered a Cybersecurity risk, a harassment, and copyright infringement. I want your email as some of us are filing lawsuits for copyright against you.
You done?
Hi,
I've already explained how the data can be used: https://huggingface.co/datasets/nyuuzyou/archiveofourown/discussions/195#680e2ac7a0520e2b7ececb18
And also OTW themselves have not claimed copyright on user-generated content: https://archiveofourown.org/tos
The OTW does not claim any copyright in or ownership of your Content. We repeat: we do not own your Content. Nothing in this agreement changes that in any way. However, running AO3 requires us to make copies, and backup copies, on servers that may be located anywhere around the world.
This was the main reason for the counter-notice, as well as the fact that the dataset does not contain any content that is directly owned by the OTW (design, site code, etc.).
Oh I am 100% behind this. This is such a good idea! I hate that when people see AI they think of generative now >:/
The problem people have isn't that it's ai. It's that they weren't asked permission for their original works or anything at all. I love the idea of ai being trained properly: with consent and not stolen work. Nyu could've chose a completely different way of going about this and have a different result.
Hi,
I've already explained how the data can be used: https://huggingface.co/datasets/nyuuzyou/archiveofourown/discussions/195#680e2ac7a0520e2b7ececb18
And also OTW themselves have not claimed copyright on user-generated content: https://archiveofourown.org/tos
The OTW does not claim any copyright in or ownership of your Content. We repeat: we do not own your Content. Nothing in this agreement changes that in any way. However, running AO3 requires us to make copies, and backup copies, on servers that may be located anywhere around the world.
This was the main reason for the counter-notice, as well as the fact that the dataset does not contain any content that is directly owned by the OTW (design, site code, etc.).
Clearly you don't understand how this works. They don't own our work but we do. We've said several times we don't consent to you taking our stuff. However because you scraped all of AO3 the OTW legal team is looking into it to protect the company. You could be considered a Cybersecurity risk, a harassment, and copyright infringement. I want your email as some of us are filing lawsuits for copyright against you.
You done?
Oh cry harder because you guys stole original works. Leave this to the adults kid.
Do you rememer the beginning of fanfiction? How many sites crashed and burned due to authors and their publishers suing writers (notebly Anne Rice) who couldn't fight back because they didn't have the money? Do you remember the fairly recent goings on with both Games Workshop (Warhammer 40K) and Wizards of the coast (D&D)? Do you remember just how much fan content we lost because of it?
This is why I think fanfic authors and AI enthusiasts shouldn't be fighting. They're on the same side of copyright law. Freedom to use other people's work is the backbone of both communities. And datasets like this are a safeguard against losing fan content. Anne Rice can do whatever she wants, but she can't take down a torrent.
I'm back after reading US copyright law. I'll show screen caps and cite. I am sharing information not condemnation. Do with this as you will.
(Source)
https://www.copyright.gov/title17/title17.pdf
(Source Pg2)
https://www.copyright.gov/legislation/dmca.pdf
Once again for people reading along:
- Writers are not on the same side as AI, because AI/LLMs can't be trained without violating people's autonomy as to what they want done with their creative work. Someone composing an offshoot of someone else's story is a human endeavor that we've been doing since we started telling stories. People pressing a button so slop comes out is not an act of conscious or creative decision making. There's no choice, no deliberation, no selection, no intent. It's just "gee what will we get" and regurgitation to fit a pattern. It's mad libs but lazier.
- The archive is a safeguard against a loss of fannish content because people upload their works willingly and to be read/downloaded in specific contexts. Stealing them is removing that autonomy from the writer. The fact that people like this keep conflating these things shows that they have no consideration for the people actually doing the writing.
- Anne Rice is dead, for the people who weren't paying attention. Which is deeply unsurprising, given the venue.
This is why I think fanfic authors and AI enthusiasts shouldn't be fighting. They're on the same side of copyright law. Freedom to use other people's work is the backbone of both communities. And datasets like this are a safeguard against losing fan content. Anne Rice can do whatever she wants, but she can't take down a torrent.
If we're on the same side, then stop training your bots on our work without permission. Make nice with us instead so we stop being pissed at you for this kind of shit.
@NauticNine See, that's the rub: they can't admit that they can't do anything without stealing from people. Their whole shtick falls apart without violating people's consent, because they can't generate the amount of work they need to make their little projects functional on their own. Even if they just used the works from every wannabe Pick Me who's turned up so far, they'd have squat. So they obfuscate and bully and demean, because they know if they had to actually do this all on their own, they're incapable.
Once again for people reading along:
- Writers are not on the same side as AI, because AI/LLMs can't be trained without violating people's autonomy as to what they want done with their creative work. Someone composing an offshoot of someone else's story is a human endeavor that we've been doing since we started telling stories. People pressing a button so slop comes out is not an act of conscious or creative decision making. There's no choice, no deliberation, no selection, no intent. It's just "gee what will we get" and regurgitation to fit a pattern. It's mad libs but lazier.
- The archive is a safeguard against a loss of fannish content because people upload their works willingly and to be read/downloaded in specific contexts. Stealing them is removing that autonomy from the writer. The fact that people like this keep conflating these things shows that they have no consideration for the people actually doing the writing.
- Anne Rice is dead, for the people who weren't paying attention. Which is deeply unsurprising, given the venue.
- I don't see how those differences are relevant. Anne Rice didn't want you using her work to make your own. You don't want AIs using your work to make their own. OpenAI doesn't want Deepseek using ChatGPT's work to make their own. And so it goes.
- The archive is a centralized organization subject to U.S. law. And given who runs the U.S. at the moment, I'd rather be safe than sorry.
- Anne Rice is a synecdoche. Her body is dead, but her spirit lives on. In fact, there's a part of her in you right now!
- The archive is a safeguard against a loss of fannish content because people upload their works willingly and to be read/downloaded in specific contexts. Stealing them is removing that autonomy from the writer. The fact that people like this keep conflating these things shows that they have no consideration for the people actually doing the writing.
The line is too blurry. If page on the Internet is public, it will be found and processed by robots from Google to Internet Archive. If it were considered โtheftโ, neither search engines nor link previews would be possible. One of the goals of the OTW is โpreserving the history of fanworks and fan culture in its myriad formsโ. At the same time, this same organization is trying to destroy copies of its library. A non-profit library.
Fantastic news guys! This dataset has been permanently disabled! LETS GO ๐๐๐
- The archive is a safeguard against a loss of fannish content because people upload their works willingly and to be read/downloaded in specific contexts. Stealing them is removing that autonomy from the writer. The fact that people like this keep conflating these things shows that they have no consideration for the people actually doing the writing.
The line is too blurry. If page on the Internet is public, it will be found and processed by robots from Google to Internet Archive. If it were considered โtheftโ, neither search engines nor link previews would be possible. One of the goals of the OTW is โpreserving the history of fanworks and fan culture in its myriad formsโ. At the same time, this same organization is trying to destroy copies of its library. A non-profit library.
You were planning to use it for ai and to make ai generated fics. that's completely different than tthe other libraries.
I have an LLC and CCL that my original stories are protected by. As someone as part of the community i suggest you take down every fic that you did not get consent to or I'd gladly help my fellow community in a law suit.
Fantastic news guys! This dataset has been permanently disabled! LETS GO ๐๐๐
Good, however data thieves are like the youngest child. they throw a fit when they get in trouble and point fingers. Then after that they'll try to find loop holes to do what they want anyway. I'll gladly offer protection of others works without any profit so it can keep cyber terror*st away.
@NauticNine Still available on datafish.ru :)
Oh thanks, now I can go find them there and remind them to remove our shit ๐
@NauticNine Still available on datafish.ru :)
Oh thanks, now I can go find them there and remind them to remove our shit ๐
Remind me where the DMCA takedown email on this Russian website is? Or how one is supposed to take down a torrent?
Remind me where the DMCA takedown email on this Russian website is? Or how one is supposed to take down a torrent?
Shh, don't spoil the fun. Imagine their reaction when they learn who runs datafish :)
Remind me where the DMCA takedown email on this Russian website is? Or how one is supposed to take down a torrent?
Shh, don't spoil the fun. Imagine their reaction when they learn who runs datafish :)
A bunch of thievesโฆ but donโt worry Nyuuzyou knows where Iโll keep at it.
Here and elsewhere.
Also: if you want I can add you to the list. Iโm sure youโd love the same amount of attention theyโre getting
@NauticNine Still available on datafish.ru :)
Oh thanks, now I can go find them there and remind them to remove our shit ๐
Remind me where the DMCA takedown email on this Russian website is? Or how one is supposed to take down a torrent?
Everyone go block MindlessForMinerva
Theyโre fine with stealing your work
https://archiveofourown.org/users/MindlessForMinerva/pseuds/MindlessForMinerva
@NauticNine Still available on datafish.ru :)
Oh thanks, now I can go find them there and remind them to remove our shit ๐
Remind me where the DMCA takedown email on this Russian website is? Or how one is supposed to take down a torrent?
Everyone go block MindlessForMinerva
Theyโre fine with stealing your work
https://archiveofourown.org/users/MindlessForMinerva/pseuds/MindlessForMinerva
That checks out considering they write on the holy bible and Harry Potter
by JKtransphobe.
That checks out considering they write on the holy bible and Harry Potter
by JKtransphobe.
I mean I didnโt wanna say it but Iโm sure glad you did!
That checks out considering they write on the holy bible and Harry Potter
by JKtransphobe.I mean I didnโt wanna say it but Iโm sure glad you did!
I like to call out b-llshit. I find it fascinating they write and post on an archive that accepts and loves the lgbtq, supports mental health, and supports a community while they actively hate on it.
@NauticNine Still available on datafish.ru :)
Oh thanks, now I can go find them there and remind them to remove our shit ๐
Remind me where the DMCA takedown email on this Russian website is? Or how one is supposed to take down a torrent?
Everyone go block MindlessForMinerva
Theyโre fine with stealing your work
https://archiveofourown.org/users/MindlessForMinerva/pseuds/MindlessForMinerva
I did let people on Reddit know in case they wanted to protect their work. I muted and blocked them.
@NauticNine Still available on datafish.ru :)
Oh thanks, now I can go find them there and remind them to remove our shit ๐
Remind me where the DMCA takedown email on this Russian website is? Or how one is supposed to take down a torrent?
Everyone go block MindlessForMinerva
Theyโre fine with stealing your work
https://archiveofourown.org/users/MindlessForMinerva/pseuds/MindlessForMinervaI did let people on Reddit know in case they wanted to protect their work. I muted and blocked them.
Reported on Reddit for violating rule 3 of the subreddit - no usernames.
@NauticNine Still available on datafish.ru :)
Oh thanks, now I can go find them there and remind them to remove our shit ๐
That area (any place that uses .Ru) hates majority of what ao3 rights about. We just have to flag it with the tags.
@NauticNine Still available on datafish.ru :)
Oh thanks, now I can go find them there and remind them to remove our shit ๐
Remind me where the DMCA takedown email on this Russian website is? Or how one is supposed to take down a torrent?
Everyone go block MindlessForMinerva
Theyโre fine with stealing your work
https://archiveofourown.org/users/MindlessForMinerva/pseuds/MindlessForMinervaI did let people on Reddit know in case they wanted to protect their work. I muted and blocked them.
Reported on Reddit for violating rule 3 of the subreddit - no usernames.
Go cry harder that you got called out. I'm sure everyone would feel safer that you can't see their stuff.
Go cry harder that you got called out. I'm sure everyone would feel safer that you can't see their stuff.
I blocked them too ๐ i don't think it hides our stuff from them though, i just did it to be petty
For the record, I don't have the same username here as i do there
Go cry harder that you got called out. I'm sure everyone would feel safer that you can't see their stuff.
I blocked them too ๐ i don't think it hides our stuff from them though, i just did it to be petty
If you also mute them it prevents them from seeing your stuff. Both blocking and muting
Go cry harder that you got called out. I'm sure everyone would feel safer that you can't see their stuff.
I blocked them too ๐ i don't think it hides our stuff from them though, i just did it to be petty
For the record, I don't have the same username here as i do there
I did let a bunch of authors know and they are absolutely baffled that someone is like them doing this.
Go cry harder that you got called out. I'm sure everyone would feel safer that you can't see their stuff.
I blocked them too ๐ i don't think it hides our stuff from them though, i just did it to be petty
If you also mute them it prevents them from seeing your stuff. Both blocking and muting
Yeah- no, I'm sorry but it doesn't work like that. It just stops them from interacting from you, it doesn't hide your work
Go cry harder that you got called out. I'm sure everyone would feel safer that you can't see their stuff.
I blocked them too ๐ i don't think it hides our stuff from them though, i just did it to be petty
If you also mute them it prevents them from seeing your stuff. Both blocking and muting
Yeah- no, I'm sorry but it doesn't work like that. It just stops them from interacting from you, it doesn't hide your work
You can also block them and other users from your work via workskin https://archiveofourown.org/works/39038346
Go cry harder that you got called out. I'm sure everyone would feel safer that you can't see their stuff.
I blocked them too ๐ i don't think it hides our stuff from them though, i just did it to be petty
If you also mute them it prevents them from seeing your stuff. Both blocking and muting
Yeah- no, I'm sorry but it doesn't work like that. It just stops them from interacting from you, it doesn't hide your work
You can also block them and other users from your work via workskin https://archiveofourown.org/works/39038346
That still only hides THEIR works from YOU, not YOUR works from THEM
Jeez, you would think a bunch of AO3 authors would know how their own site works ...
Jeez, you would think a bunch of AO3 authors would know how their own site works ...
I naturally don't post often nor had to block people
@NauticNine Still available on datafish.ru :)
Oh thanks, now I can go find them there and remind them to remove our shit ๐
Remind me where the DMCA takedown email on this Russian website is? Or how one is supposed to take down a torrent?
Everyone go block MindlessForMinerva
Theyโre fine with stealing your work
https://archiveofourown.org/users/MindlessForMinerva/pseuds/MindlessForMinerva
Everyone report Delirious and Vanoss-esque comment
OP is looking at 12M counts of copyright infringement against a website that has a legal team on retainer, i hope the OTW presses charges! ๐