It's certainly nice to see someone accused of bittorrenting with the bankroll to come up with a decent legal defense team.
Sounds like Meta are banking on the entertainment industry looking at it and deciding that the risk of losing this case is too high given Meta’s almost infinitely deep pockets to mount a legal defence.
If this does break the stranglehold that copyright has over creative acts, especially in the US, this feels like a net good.
But Meta's contention is 'you don't have any proof of that'.
I think there is enough existing case law and ambiguity in the law as it's written that Meta stand a reasonable (although not a good) chance of being able to argue that they did not commit any crime because a.) they did not create the infringing copy (or that the infringing copy that they received was a technical copy, and they did not create an infringing copy themselves) b.) they did not infringe for private or financial gain (the models they trained on this material were released to the public for free). There's an argument that copyright infringement occurs only upon distribution, and as far as I'm aware, there's no case law that just downloading a copy is illegal.
Meta may also be able to argue that their use of the material could be considered 'fair', as it is non-commercial, transformative, and that the use of the material does not harm the market for the original work.
I'm not a lawyer, and I'm not arguing about the merits of these arguments, just that they seem to me to be plausible.
Copyright protects against making copies of the work, which they definitely did.
> There's an argument that copyright infringement occurs only upon distribution
Not in most countries. Certainly not in America.
> b.) they did not infringe for private or financial gain (the models they trained on this material were released to the public for free).
They definitely gained from it. If their argument rests on that then they're screwed.
> Meta may also be able to argue that their use of the material could be considered 'fair', as it is non-commercial, transformative, and that the use of the material does not harm the market for the original work.
Probably their best bet but it's hard to see how that would fly given that it is commercial even if they released it for free, and fair use normally depends on how much of the work you use; they used all of everything.
I agree. But a good lawyer might be able to argue that they only received a copy, they didn’t make one themselves.
> Not in most countries. Certainly not in America.
I think the law itself is clear that reproduction is its own right, but I couldn’t find any case law where someone was prosecuted only for reproduction. There are certainly some concerns with the law as it’s written (such as the first sale doctrine, or home ripping of CDs, etc.).
> They definitely gained from it. If their argument rests on that then they're screwed.
Yes but were the gains private and financial? Again, a good lawyer might be able to argue that actually, Facebook invested a lot (financially) into training the models, and then released for free, so are net negative financially.
> fair use normally depends on how much of the work you use
Fair use is a very difficult one to put an exact definition on, and whatever definitions exist do not determine based purely on the amount of the work used. There is case law that a full work can be considered fair use, and that even minimal parts of the work are not. Again, a good lawyer could perhaps make this argument successfully.
I don’t think anyone without access to a legal team that costs millions would stand much of a chance here, but Meta might.
I’m all for them getting a dose of reality in this case, and nothing consistently whips tech in the pocketbook like whining their interpretation of Fair Use is legal when it clearly is not.
¹ Not necessarily a formal legal precedent, but at least a floor on the "market value" of access to the data
Precedent that LLMs get to keep & use copyrighted data
LLMs get to keep & use copyrighted data without legal precedent
I bet the industry will file amicus briefs to try to support the plaintiffs
If you’re going to have this fight, wait until you have it with a worse-prepared and worse-resourced opponent where you’re more confident of the win.
Facebook could simply buy most of the companies involved if they give them too much shit. We've consolidated way too much power into a few large tech companies. I don't see it very likely that Hollywood could win this.
[1] https://companiesmarketcap.com/walt-disney/marketcap/ [2] https://companiesmarketcap.com/comcast/marketcap/ [3] https://stockanalysis.com/stocks/meta/market-cap/
Is that a problem for them ?
Doesn’t meta make more money than the entire industry of Hollywood including all home entertainment revenue ?
I am certain they do.
EDIT: 2024 full year revenue for meta is ~160B as compared to (roughly) 140B for the entirety of the film industry .
Google paid about $1b to Viacom in the YouTube piracy dispute. That's a lot of money, but do you recall anything seriously changing when that happened?
To me, the funniest product is Beat Saber. The best VR game by far. 99% of the value is tied up in violating musician's rights. Meta saved that game. Did people stop making music? No.
This book torrenting thing is complex. The main thing plaintiffs want is discovery of the training data. It's not complicated. There's no justification for the court to block that, it's a fishing expedition yes, but one that will turn up a lot of fish. Then all AI companies will have to acquiesce to it. That is the "win" for the industry.
The U.S. Media and Entertainment (M&E) industry is the largest in the world at $649 billion (of the $2.8 trillion global market) and is projected to grow to $808 billion by 2028 at an average yearly rate of 4.3% (PwC 2024).
https://www.trade.gov/media-entertainment
Meta Platforms, formerly known as Facebook Inc., continues to dominate the digital landscape with impressive financial growth. In 2024, the company's annual revenue reached a staggering 164.5 billion U.S. dollars, marking a significant increase from 134.9 billion U.S. dollars in the previous year. This upward trajectory reflects Meta's ability to monetize its vast user base across multiple platforms, solidifying its position as a tech giant.
https://www.statista.com/statistics/268604/annual-revenue-of...
If you look at UMG's revenue, one of the largest labels, their revenue was 11B.
Meta’s market cap is higher than that of the US entertainment industry
Meta’s revenue is also higher than that of the US entertainment industry
The parent of the comment you’re replying to conflated the two
If anything, the law should require that they seed their training data so that the competitive landscape converges on actual technological innovation and not moat building through data destruction.
The books copied by Meta, explicitly disallow it, and require payment for distribution.
I'm not unhappy about it; but was never consulted.
https://en.wikipedia.org/wiki/Ticketmaster_Corp._v._Tickets.....
Things should be free, as in speech, not as in beer. Especially in this case. The giants of Silicon Valley could in fact purchase these rights.
Few authors care about people personally enjoying a product through otherwise means. They do care about mass distribution without attribution, without royalty, and without regard.
Used to, but more recently it's probably LLM agents using Google not people. And even if it's not yet, it will be. Last time I searched for something on Google it messed up so bad I quickly returned to GPT-4o+search.
The answer to bad products is not to throw away the idea of people getting to control their own content.
a) Meta are (so far) releasing their models for free.
b) There's nothing stopping non-mega-corps from doing the same, especially if this precedent was established. (Training is of course expensive but this is a challenge, not an absolute block.)
That’s enough to bankrupt individuals but industries fighting industries can see it to the end, if they don’t settle
https://www.omm.com/insights/alerts-publications/trump-admin...
After all, the main people hurt would be Hollywood, which is run by people supporting the Democrats. And it would be popular with many voters (not an issue for Trump but it is for Republicans).
Counter example: ownership of Amazon MGM Studios and its parent Amazon.
They would probably benefit by handicapping Netflix/Disney/WBD/etc.
Historically, copyright cases fell in favor of big media corporations based on the notion that they were very rich and powerful and could fight things endlessly, bribe/lobby politicians, and cause laws to be changed (e.g. the DMCA).
However, AI companies are wealthier still. Some have revenues exceeding the GDPs of most countries. Surely, rich enough to outright buy out some of these media companies. At which point it would stop being copyright infringement because they'd own the copyrights. I'm sure some other arrangement will be found that is less mutually disruptive than a lot of court cases. Both sides are making too much money for anything else to happen. Forget about small book publishers making much of a difference here.
Trump could make Grok, Facebook, Google and OpenAI's actions legal in response to a bribe from Musk.
Or he could step up enforcement actions against Facebook, Google and OpenAI while issuing a pardon to Grok.