Brandon Butler
LightNews LightNews
bb.usefairuse.com
Brandon Butler
@bb.usefairuse.com
840 followers 530 following 700 posts
Copyright lawyer at Jaszi Butler PLLC, Exec Director @recreatecoalition.bsky.social, dad. Press inquiries: press@recreatecoalition.com.
Posts Replies Media Videos
bb.usefairuse.com
ACCESSIBILITY ALLIES! A fresh addendum to the @arlpolicy Code of Best Practices in Fair Use for Academic and Research Libraries brings the 2012 document up to date, adding positive caselaw and statutory updates! www.arl.org/wp-content/... #a11y #fairuse
Screenshot shows the first page of the ARL Code addendum, which includes confirmation of fair use's application to accessibility under HathiTrust, Section 121's applicability to research libraries, and updates to Section 121 from the Marrakesh Treaty Implementation Act. Screenshot of the second page from the new Addendum, which explains that the HathiTrust opinion also confirmed fair use's availability for accessibility work involving non-textual materials, and improvements to Section 121 in this regard from the MTIA.
October 30, 2025 at 1:00 PM
bb.usefairuse.com
Again, via PK's excellent post, here's what happens when you train an AI model:
publicknowledge.org/piracy-vs-f...
The critical point: Al models do not store or distribute the copyrighted works they are trained on in the way pirates do. Instead, the training process works more like this: • The system ingests the training data and converts it into mathematical representations (vectors, weights, and parameters). • The original works are not preserved in the final model. Instead, the system captures statistical information about how words follow one another, how images are composed, or how sounds combine into melodies. • Once training is complete, the model generates new outputs based on probabilities, not by reproducing specific works or parts of them. This training process almost always requires making temporary copies of copyrighted works during the process. But the end product is not a copy of those works; it is a statistical model capable of generating new, original content.
September 17, 2025 at 4:11 PM
bb.usefairuse.com
Does AI training look like that? Not at all. It doesn't do the two most important things a pirate does: make and sell unauthorized copies of works. As Patrick explains, what AI training does instead is radically different:
September 17, 2025 at 1:00 PM
bb.usefairuse.com
We all know what piracy is: making and selling copies of entire works without payment or permission. Piracy hurts rightsholders because people buy or download a copy of their work from the pirate instead of buying one from them.
September 17, 2025 at 1:00 PM
bb.usefairuse.com
This is #copyfraud. Like the bullshit FBI warning on DVDs and the "No copying allowed without express permission" warning in the front of books, it's literally false. But OK, sure, why not.
# NOTICE: all crawlers and bots are strictly prohibited from using this # content for AI training without complying with the terms of the RSL # Collective AI royalty license. Any use of this content for AI training # without a license is a violation of our intellectual property rights. License: https://rslcollective.org/royalty.xml
September 10, 2025 at 6:59 PM
bb.usefairuse.com
Bobbs-Merrill tried this in 1908, putting a tag on every copy of the book The Castaway that read "The price of this book at retail is $1 net. No dealer is licensed to sell it at a lower price, and a sale at a lower price will be treated as an infringement of the copyright."
September 10, 2025 at 6:40 PM
bb.usefairuse.com
2️⃣ We also have an infographic for YouTubers worried that they can't be confident that their videos are fair use in order to issue a counterclaim: Flex your fair use rights like a pro by relying on the fair use best practices developed by pro filmmakers. www.recreatecoalition.org/infographic...
September 10, 2025 at 1:43 PM
bb.usefairuse.com
If YouTubers want to rally around a one-word fix that would help them clap back against UMG and other copyright over-claimers, we have a suggestion: remove the word "knowingly" from Section 512(f) of the Copyright Act. That would *start* to put fair use on a level playing field.
September 10, 2025 at 1:43 PM
bb.usefairuse.com
Section 107 fair use rights have no such protection. The one part of the law meant to punish bogus takedown notices only applies if you can prove the sender *subjectively knew* their takedowns were bogus when they sent them. It's impossible.
Screenshot text from newsletter explains the high bar to recovering damages when fair users are harmed by excessive copyright claims.
September 10, 2025 at 1:43 PM
bb.usefairuse.com
First, Rick and his followers ask how this can be legal, and we explain why: Copyright has a double standard for liability. Section 106 exclusive rights trigger crushing penalties with strict liability, meaning you're responsible regardless of your intent. Just ask Anthropic:
Screenshot text explains the high statutory damages in copyright law and how that led to an exorbitant settlement against Anthropic in its legal battle over AI training.
September 10, 2025 at 1:43 PM
bb.usefairuse.com
This week's @recreatecoalition.bsky.social newsletter has 2 suggestions for YouTubers like Rick Beato, who raised the alarm over Universal Music abusing the ContentID system to send a steady stream of seemingly bogus copyright claims. us21.campaign-archive.com/?u=b444b538...
September 10, 2025 at 1:43 PM
bb.usefairuse.com
wonderful law review article on fair use's least-loved factor - the nature of the work used. In "Is That All There Is? Reflections on the Nature of the Second Fair Use Factor," Kasunic gets fair use exactly right.
kasunic.com/Articles/CJ...
While American copyright law grants a property interest in original works of authorship, that property interest is limited, not just in duration, but also in scope. An essential limitation on the scope of the copyright owner’s exclusive rights is the fair use of the work by others. Thus, while copyright provides limited property rights, the enforcement of these rights is a tort – the requirement of proving that the use of expression by another is wrongful. A fair use of another’s expression is not wrongful when it does not impede an author’s incentive to create. Notwithstanding the exclusive rights of the copyright owner, a fair use is not an infringement of those rights.
September 10, 2025 at 12:27 PM
bb.usefairuse.com
The first one, filed early, comes from longtime friend of Recreate, Professor @charlesduan.bsky.social, whose briefs have been cited in two previous SCOTUS opinions on IP. As usual, Charles' arguments are lucid and compelling. Find his brief here:
www.supremecourt.gov/DocketPDF/2...
This is not a patent case, yet patent law is crucial to it. Copyright law’s contributory infringement doctrine is closely linked to the parallel doctrine of inducement of patent infringement. Indeed, the “material contribution” test at issue is shared between the two. Patent inducement can thus guide this Court’s analysis. Patent law’s material-contribution analogue has long required a specific, unambiguous, affirmative act; that requirement is appropriate for copyright as well. And recent decisions by the U.S. Court of Appeals for the Federal Circuit, which have ignored this rule and “made a mess of patent infringement theory and doctrine” as a result, illustrate the dangerous consequences of failing to adhere to that rule, for both patent and copyright law.
September 5, 2025 at 3:31 PM
bb.usefairuse.com
Indeed! So, like, why break the internet for the rest of us?
www.nytimes.com/athletic/65...
September 4, 2025 at 2:32 PM
bb.usefairuse.com
Just how totally insane is the 4th Circuit's secondary liability ruling in Cox v. Sony—a decision now up on appeal at the Supreme Court? Combined w/ ©'s statutory damages, it's truly nuts. 😵 @CCIAnet.bsky.social has the numbers, and they will curl your hair. 🫠
ccianet.org/research/ca...
Screenshot shows excerpt from research report that concludes "* This resulted in a statutory damages to actual economic harms ratio of at least 1,445 to 1, and potentially as high as 481,695 to 1 in the Cox case." Also, in the recent Hachette v. Internet Archive case, plaintiffs "could have sought at least $690 billion (greater than the GDP of 172 IMF-tracked countries in 2023, and greater than the combined sum of 77 IMF-tracked countries' GDP) — and potentially up to $6.6 trillion dollars-one-fifth of U.S. GDP and more than the combined GDPs of about 144 IMF-tracked countries-from legitimate service providers who did not terminate the Internet Archive from their services"
September 4, 2025 at 2:23 PM
bb.usefairuse.com
I think there's plenty of room for fair use of copies sourced from shadowy sites, for the reasons Judge Leval explained over 30 years ago when he laid out the framework that still governs fair use today:
www.jstor.org/stable/pdf/....
Screenshot shows an excerpt from Pierre Leval's article "Toward a Theory of Fair Use," which culminates in this observation: ...The inquiry should focus not on the morality of the secondary user, but on whether her creationclaiming the benefits of the doctrine is of the type that should receivethose benefits. This decision is governed by the factors reviewed above - with a primary focus on whether the secondary use is productive and transformative and whether it causes excessive injury to the market for the original. No justification exists for adding amorality test. This is of course not an argument in favor of immo-rality. It favors only proper recognition of the scope and goals of abody of law.
August 28, 2025 at 2:21 PM
bb.usefairuse.com
A more accurate headline would be "Universal Pictures to frightened EU media studies professors: we'll sue you if you study our movies." Very brave, very good for culture. Stay classy, Universal. www.hollywoodreporter.com/business/bu... via @thr.com
In some countries, the warning cites a 2019 European Union copyright law largely opposed by major tech companies that allows creators to opt out of having their material used in scientific research by explicitly reserving their rights.
August 8, 2025 at 9:09 PM
bb.usefairuse.com
Wow, Patreon just hit an extraordinary benchmark, paying out more than $10billion to creators since its founding in 2013. Founder @jackconte.bsky.social's optimism is a breath of fresh air.
www.axios.com/2025/08/05/...
In terms of being a creative person, having control, having leverage, getting to keep your IP [intellectual property], getting to keep your copyright and building a business around what you love and what you're passionate about — This is the best time in the history of humanity to do that
August 6, 2025 at 3:49 PM
bb.usefairuse.com
I made a schedule for Birmingham's Sidewalk Fest (August 19-24) with only Jaszi Butler clients on it. 😎 You could spend the whole fest just watching movies we worked on - and I can vouch, they're all amazing.
sidewalkfest.com/festival/si...
August 1, 2025 at 12:04 AM
bb.usefairuse.com
I suspect we will see more and more of these little conversions re AI and art, kinda like how most grumpy rock dudes eventually recognized hip hop is legit art. But they may not all be as eloquent in their flip-flopping as Nick Cave is here. pitchfork.com/news/nick-ca...
Roughly once a week, Nick Cave blogs about music, religion, grief, and the infernal state of world affairs through the medium of fan correspondence in his Red Hand Files newsletter. His meatier responses become news stories, and few have traveled so far, or with such fervor, as his 2023 diatribes eviscerating ChatGPT. First, that January, he portrayed a lyric “in the style of Nick Cave” as an omen of the apocalypse and “a grotesque mockery of what it is to be human.” A few months later, Cave elaborated that any lyricists using artificial intelligence were “participating in [the] erosion of the world’s soul and the spirit of humanity itself.” But that was before his friend Andrew Dominik made a music video for Nick Cave & the Bad Seeds’ “Tupelo” featuring a floating AI Elvis. Dominik—a longtime Cave collaborator and the director of Blonde, The Assassination of Jesse James by the Coward Robert Ford, and One More Time With Feeling—had been at work on the video behind Cave’s back. Presented with the result on the song’s 40th anniversary, Cave decided it was “a soulful, moving, and entirely original retelling of ‘Tupelo,’ rich in mythos and a touching tribute to the great Elvis Presley.” Cave found that Dominik’s AI-animated photographs of Presley “had an uncanny quality, as if he had been raised from the dead, and the crucifixion-resurrection images at the end were both shocking and deeply affecting.” He and his wife, Susie Cave, he added, “were blown away. As I watched Andrew’s surreal little film, I felt my view of AI as an artistic device soften. To some extent, my mind was changed. ‘It’s a tool, like any other,’ said Andrew.” Far from coming away from this religious experience with his tail between his legs, Cave emerged with a renewed understanding of his own brilliance. “I believe that the ability to change one’s mind is the very definition of strength,” he wrote. “We pursue the truth wherever it may lead, remaining flexible and humble enough to adjust our views as new evidence emerges, regardless of how uncomfortable that may feel. It is ultimately a form of resilience, not a sign of weakness. Rigidity breaks; flexibility endures.” Watch the video below.
July 30, 2025 at 2:59 AM
bb.usefairuse.com
2c cont'd And finally, the coup de grace, even if the competition is effective, is it more effective because of the inclusion of the plaintiff's work than it would be without it? This, to me, could be fatal to many, perhaps most claims of market dilution.
June 26, 2025 at 3:53 PM
bb.usefairuse.com
2c cont'd If the tool does make works that compete, what is the impact of that competition? Will people really buy what amounts to slop rather than the works in the training data?
June 26, 2025 at 3:53 PM
bb.usefairuse.com
2c cont'd If the AI can make works of the same kind as the training data, do those works compete with the training data works? Not necessarily!
June 26, 2025 at 3:53 PM
bb.usefairuse.com
2c. Indeed, in chastising Meta's lawyers for somehow fumbling the market dilution argument, Judge Chhabria lays out what actually looks like a pretty daunting series of factual showings that this argument would require. Does the tool actually generate competing works?
First, is Llama capable of generating such books? If it isn’t currently, will it be capable of doing so in the near future? Presumably the answer is yes, but that’s not a foregone conclusion. An LLM could, for instance, be configured to be unable to produce book-length or book-style outputs. So the fact that books are being created by some LLM does not automatically mean that 33 Case 3:23-cv-03417-VC Document 598 Filed 06/25/25 Page 34 of 40 Llama can create them or will be able to do so soon.
June 26, 2025 at 3:53 PM
bb.usefairuse.com
2b. But, to his credit, Chhabria tries to redeem the theory by setting limits to its scope, recognizing that this harm, if it exists, won't impact all works in the same way, or to the same extent. Plaintiff Sarah Silverman, for example, won't likely suffer this harm.
Of course, not all copyrighted works would have their markets diluted equally by AI- generated competitors. It seems unlikely, for instance, that AI-generated books would meaningfully siphon sales away from well-known authors who sell books to people looking for 28 Case 3:23-cv-03417-VC Document 598 Filed 06/25/25 Page 29 of 40 books by those particular authors. But it’s easy to imagine that AI-generated books could successfully crowd out lesser-known works or works by up-and-coming authors. While AI- generated books probably wouldn’t have much of an effect on the market for the works of Agatha Christie, they could very well prevent the next Agatha Christie from getting noticed or selling enough books to keep writing.12 Third, what impact does this competition actually have on sales of the books it competes with? Does it drown out those books entirely? Does it just chisel at their sales at the margins? Or, as discussed above and seems likely, does it depend on the book—are readers of romance novels happy to buy AI-generated ones, while all the people who want to read Sarah Silverman’s memoir still want to read it over AI-generated comic memoirs? Whatever the effects have been thus far, are they likely to increase in the future, as more and more AI-generated books are written, and as LLMs get better and better at writing human-like text?
June 26, 2025 at 3:53 PM