Court docs allege Meta trained its AI models on contentious trove of maybe-pirated content
- Reference: 1736489769
- News link: https://www.theregister.co.uk/2025/01/10/meta_libgen_allegation/
- Source link:
The accusation was made in a [1]document [PDF] filed in the case of Richard Kadrey et al vs Meta Platforms, in which novelist Kadrey (and others including comedian Sarah Silverman) allege stolen versions of their work were used to train AI models. Several similar suits are in motion, targeting different AI players.
The document claims that Meta decided to download documents from Library Genesis – aka “LibGen” to train its models. LibGen is the subject of a [2]lawsuit brought by textbook publishers who believe it happily hosts and distributes stolen works, and even accepts donations to fund its operations.
[3]
The filing from plaintiffs in the Kadrey case claims that documents produced by Meta during the discovery process – the pre-trial activity of gathering relevant documents – describe internal debate about accessing LibGen, a little squeamishness about using BitTorrent in the office to do so, and eventual escalation to “MZ” who approved use of the contentious resource. The filing states that evidence about use of LibGen is new and was made available by Meta late in the discovery process.
[4]
[5]
Another [6]filing [PDF] claims that a Meta document describes how it removed copyright notifications from material downloaded from LibGen, and suggests the company did so because it realized including such text could mean a model’s output would reveal it was trained on copyrighted material.
A [7]third document [PDF], this one filed by Meta, argues that the plaintiffs have unjustifiably claimed that use of LibGen is new material and contends that it was on the record for months.
[8]We did warn you – 2025 may be the year AI bots take over Meta's 'verse
[9]New Orleans attacker used Meta smart glasses to plan New Year's Day massacre
[10]Nick Clegg steps down as Meta's top flack in favor of more Trump-friendly candidate
[11]Fining Big Tech isn't working. Make them give away illegally trained LLMs as public domain
The nub of the matter appears to be an attempt by the plaintiffs to use the info about Meta’s user of LibGen to add an action under the California Comprehensive Computer Data Access and Fraud Act. That law makes it a crime to access a computer or network without permission with the intent to defraud or commit other crimes. Meta doesn’t think the extra action is justified.
Meta’s filing includes a statement that the company “rejects the notion that it has ‘distributed’ LibGen”, seemingly to address plaintiff’s arguments that merely using BitTorrent meant it spread stolen content to others. But if there’s a denial that LibGen was accessed, we can’t find it.
[12]
Meta tried to have the filings we’ve linked to above sealed on grounds of commercial sensitivity. The judge in the case rejected that, arguing that Meta just wants to avoid publicity.
US District Court Judge Vince Chhabria also noted that in one of the documents Meta wants to seal, an employee wrote the following:
“If there is media coverage suggesting we have used a dataset we know to be pirated, such as LibGen, this may undermine our negotiating position with regulators on these issues.”
Sorry if we undermined you, Zuck.
The allegation of using LibGen is very on-brand for Meta, given its business model is built on free content contributed by users. Why should pesky authors be treated any different? ®
Get our [13]Tech Resources
[1] https://regmedia.co.uk/2025/01/10/pacer_kadrey_vs_meta_1.pdf
[2] https://www.theregister.com/2023/09/18/science_publishers_sue_libgen/
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/legal&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2Z4D91heb0I4Tip_FruBc2wAAAAk&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/legal&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Z4D91heb0I4Tip_FruBc2wAAAAk&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/legal&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Z4D91heb0I4Tip_FruBc2wAAAAk&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[6] https://regmedia.co.uk/2025/01/10/pacer_kadrey_vs_meta_2.pdf
[7] https://regmedia.co.uk/2025/01/10/pacer_kadrey_vs_meta_3.pdf
[8] https://www.theregister.com/2025/01/08/meta_ai_bots/
[9] https://www.theregister.com/2025/01/06/new_orleans_attacker_meta_glasses/
[10] https://www.theregister.com/2025/01/03/nick_clegg_meta/
[11] https://www.theregister.com/2024/12/22/ai_poisoned_tree/
[12] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/legal&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Z4D91heb0I4Tip_FruBc2wAAAAk&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[13] https://whitepapers.theregister.com/
I don't know. Training large LLM's requires so much data that I believe it will be impossible to adhere to all of the licensing of the materials that have been used for training. Also because only the relationships between the words are stored in relation to all other texts of all other sources, instead of the text themselves. Therefore I believe copyright notices would show up at random places if they were to be included.
So it's a bit like reading 10 books on a subject and then writing your own books/texts in your own wording, based on the information you have learned.
And then, in contrast to OpenAI, Meta is at least giving back the open models to be benefit of everyone.
Also I think the world should realize that if copyright restrictions were strongly enforced, then only countries like Russia and China would end up having LLM's since they are likely not to enforce the same restrictions.
So I think it's a difficult situation when it comes to copyright.
copyright notices would show up at random places
I think I've mentioned before that if you ask ChatGPT to reproduce the opening paragraphs of a book that's out of copyright, it will and if you ask it do the same of a book that's in copyright it will refuse. So clearly, built into the system (perhaps supplemental to the AI model) is a notion of copyright. Also built into the system (perhaps supplemental to the AI model) is a means of quoting significant parts of texts verbatim. Not all systems will be the same, but I think you have to look at the system overall rather than focus simply on the LLM.
There are also some very powerful lobbies behind copyright - you simply have to look at the extensive efforts by the entertainment industry to stamp out "piracy", so the AI advocates aren't necessarily pushing at an open door.
And don't forget that the purpose of copyright is to enourage creativity. You may argue that it's no longer necessary if AI can produce all the consumer-oriented pap that might be required. But AI with nothing to ingest other than its own output is going nowhere. If you're effectively proposing that the fruits of people's brains and labour should automatically become the property of megacorporations so they can rent it back to the creators then you're advocating for a grim dystopia to which Russia and China are welcome.
You are Jeff Muckerberg and I claim my £5Bn.
Writing stuff down
As ever I'm not shocked by the things people do, but it's still unbelievable how many people choose to put things down in writing which they know are dodgy just waiting for someone to find.
If they didn't know or didn't care about potential wrongness they wouldn't even have discussed it, so it has to either be a CYA scenario to make sure the boss has signed off or just outright stupidity.
Either way it ends the same.
Re: Writing stuff down
Or could be because they were WFH. Why, I guess, there's so much pressure to RTO.
Meta is now so big, like most mega corps, it will write a small cheque on condition they can claim that they did nothing wrong and just write off the cost as doing business and move on to the next bit of piracy or privacy invading.