Copyhype

AI training of copyrighted material not fair use: Third Circuit — “The Third Circuit rejected an artificial intelligence company’s fair use defense after it trained its AI on a competitor’s product, finding the training constituted copyright infringement — and potentially setting a landmark precedent for artificial intelligence.”

U.S. Copyright Office Celebrates 50th Anniversary of the 1976 Copyright Act — “Across speakers’ remarks from the three panels, it was evident that the U.S. Copyright Office, its staff, leadership, and copyright supporters and champions in Congress and in the private sector have been instrumental in creating copyright legislation that has weathered and survived incredible challenges over the last 50 years.”

A Lego copyright fight over a jacket worn on ‘Queer Eye’ is headed toward trial — “Concannon sued Lego in 2021 in the U.S. District Court for Connecticut, where the company was previously based in the U.S. before moving its Americas headquarters to Boston last year. He alleged the company infringed his copyright, and he also brought claims for trade dress infringement and violations of Connecticut’s unfair trade practices law. This week, U.S. District Judge Omar A. Williams granted Lego’s motion for summary judgment in part, dismissing Concannon’s trade dress claim but allowing his copyright claims to proceed to trial.”

Vetter Says Wait. The Supreme Court Shouldn’t — “BMG and the publishing arms of Universal, Sony and Warner then bought out the losing defendant’s interest in the song for the express purpose of seeking Supreme Court review. Now Kappel has teamed up with one of the country’s top Supreme Court advocates, Orrick’s Josh Rosenkranz, to argue that this isn’t the case, and now isn’t the time. Their brief in opposition to cert (read here) argues that there’s no circuit split, and that the issue needs to ‘percolate’ before the Court weighs in.”

The Heart of the Matter: Copyright, AI Training, and LLMs: Part 4: Licensing as a Key Part of the Path Forward — “Copyright law has adapted to every major technological disruption since the player piano. When photocopying arrived in the 1970s, direct and collective licensing together built a functioning market. The early internet required a similar recalibration. Text and data mining raised the same questions a decade ago. Each time, the answer was not litigation to exhaustion. It was a licensing framework that allowed innovation to proceed while ensuring that creators were compensated. AI is not a different problem.”

“An astonishing theft of unprecedented proportions”: Court records show what Microsoft and OpenAI actually thought about AI training — “Internal Microsoft communications identified the ‘real risk’ that generative AI could ‘significantly disrupt the employment of the very people who generated the data on which the foundation model was trained.’ A separate policy document put it like this: ‘LLMs are a product that destroys its supply chain.’ The current approach has created a ‘doom loop,’ according to another internal Microsoft document.”

Fair Use Is Not a Shield AI Companies Can Use to Shelter Their Use of Pirate Copies: Analysis of Fair Use Factor 1 — “The fact that AI companies’ intentional use of these illicit sites does not presumptively disqualify them from claiming fair use is antithetical to the foundations and goals of our copyright system and the rights it guarantees.”

Afro Caribbean Creative Bridge forum: Stakeholders call for stronger intellectual property, copyright protection for creatives — “The forum brought together stakeholders to explore ways of strengthening intellectual property protection and deepening creative ties between Africa and the Caribbean.”

MLB Loses Most of Cable Royalty Fight, but Court Still Throws Out Copyright Board’s Split — “The U.S. Court of Appeals for the District of Columbia Circuit on Tuesday vacated the Copyright Royalty Board’s allocation of cable retransmission royalties collected from 2014 through 2017 and sent the matter back for further proceedings. The ruling came even as the court upheld most of the methods challenged by MLB and Public Broadcasting Service.”

AI Bots Are Stripping European Booksellers’ Shelves; EU Copyright Law May Ban It — The industrial-scale destruction of books for training LLMs has reached Europe. The practice emerged following the decision in Bartz v. Anthropic, a lone district court decision in the US that ignored clear contrary precedent.

Meta Fails to Block Zuckerberg Deposition in AI, Copyright Suit — The Meta CEO will be deposed in the copyright suit before the end of the month. Counsel for one of the plaintiffs had argued that Zuckerberg “has personal knowledge of Meta’s AI initiatives including Project Vellum, a program to purchase and scan physical books to train AI models.”

The French Association of News Media has accused Google’s AI summaries of copyright infringement and called on regulators to launch an investigation. — Google’s AI summaries are in the spotlight in France, which has press publisher rules requiring payment and good-faith negotiations by digital aggregators to display article extracts.

Shein loses UK copyright lawsuit against rival Temu in key e-commerce ruling — The clash beteween the two ecommerce giants began with allegations by Shein that Temu engaged in “industrial-scale” copying of photos of its clothing offerings. The UK court rejected the claims, finding that Temu had not authorized infringement by third-party merchants that uploaded the images to its platform, nor that it had the requisite knowledge of the infringement.

Tech Bro Scrapes Anti-AI Photo App Cara, Then Gloats About It — The site was created to allow photographers to share their work online on a platform that would prevent it from being scraped to train AI. The site was scraped anyway. The site’s founder asks, “What are artists supposed to do? We already have anti-bot and anti-scraping measures. We already set robots.txt telling bots not to scrape for AI.”

AI Developers’ Trail of Book Destruction — There is no “loophole” in copyright that permits scanning books if you destroy the original. There is a single federal district court decision that did not acknowledge the body of case law that says otherwise, including binding precedent from its own Circuit Court. Keith Kupferschmid of the Copyright Alliance takes a closer look at the stories around AI developers scanning and destroying physical books for training materials.

The Underappreciated Third U.S. Fair Use Factor in Copyright Infringement Cases Concerning AI Training — Law professor Marketa Trimble discusses the third fair use factor, which directs courts to consider “the amount and substantiality of the portion used in relation to the copyrighted work as a whole” when analyzing whether a particular use is a fair use, and observes that in most cases, while AI companies use a vast number of works to train models, no particular work is required for an AI model. This raises an issue under fair use since it indicates a lack of the type of justification to use a particular work that is needed for fair use.

Google Earth Pulls AI Image Generator After Users Created Misleading Images — There are several explanations for why seemingly every week one of the major tech companies introduces a new AI feature that no one asked for and is clearly harmful (example). One is that they are beta testing and need to collect large amounts of real world user data.

AI-assisted music now eligible for copyright registration in South Korea — The Korea Music Copyright Association reversed its earlier policy that made such works ineligible for registration. Songs generated purely from text prompts without any additional human involvement will remain ineligible under the new policy.

China’s MiniMax curbs overseas access to new AI video model over copyright disputes — The developer has limited access to users in the US, EU, UK and South Korea following ongoing copyright litigation, including a federal lawsuit in the US brought by several film studios.

German court rules AI music firm Suno broke copyright rules — The final text of the decision is not yet public, and the ruling can be appealed. Nevertheless, first-hand reports suggest the court held both that the training of the model with copyrighted works is not fair use under US law and that outputs retrievable in Germany can be domestic acts of infringement.

Only humans can be ‘authors’ say Canadian music publishers in test case on copyright status of AI-generated works — The association for Canadian music publishers has formally intervened in a challenge of a copyright registration for an AI-generated work which has already been denied registration by the US Copyright Office.

ChatGPT starts blocking direct requests to copy an author’s style — Ars Technica reports the generative AI model is now preventing outputs that copy authors’ styles, which suggests the company may foresee exposure to liability for the outputs it model generates independent of liability for training.

Meta Says It Has Scanned Every Single Public Instagram Post With its AI — Another week, another tone-deaf admission from the social media giant. The benefit of the scanning touted by Meta is that users spend more time on Instagram.

It’s a wrap: Constructive discovery under copyright statute of limitations — Attorney Vincent Li analyzes a recent Second Circuit decision in Hayden v. Koons in which the court dismissed a copyright infringement claim as untimely rejecting plaintiffs constructive discovery arguments.

Bad Bunny asks judge to reconsider reggaeton ‘dembow’ copyright ruling or certify it for appeal to the Ninth Circuit — The motion comes in response to a July 1 ruling denying defendants’ motion for summary judgment on claims that over 150 artists infringed on works in which foundational reggaeton elements can be found.

OpenAI faces sanctions bid as copyright case escalates — News plaintiffs have moved for remedies related to what they call the “deliberate and systemic effort to obstruct discovery” by the generative AI startup. The news lawsuits are running in parallel with author class actions against OpenAI.

Publishers, Authors File Class Action Lawsuit Against Google — The lawsuit alleges the tech giant engaged in widespread infringement to train its Gemini AI models, including copying from books provided by publishers for other Google services, such as Google Books.

Designers beware: Register your building designs as architectural works — A West Virginia federal court dismissed a copyright infringement claim because the works, building designs for a bubble tea shop and restaurant, were registered as “pictorial, graphical, or sculptural works,” rather than as “architectural works.” Sounds like a formality that acts as a trap for the unwary.

Mexico Reforms Copyright Law to Protect Against AI, Voice Cloning — Analysis of amendments to Mexico’s Federal Copyright Law introduced in May, which would strengthen image and likeness rights for individuals, including with respect to generative AI uses.

New York Times-led group asks court to sanction OpenAI in US copyright dispute — In litigation originally filed in 2023 alleging the AI developer infringed copyright in news publishers’ works when it trained its Large Language Models, the group of newspapers this week asked the court to impose sanctions on OpenAI for allegedly misrepresenting its ability to search its training datasets and deleting other key evidence during discovery.

Judge Chhabria denies Kadrey book authors’ motion to seek an interlocutory appeal of downloading part of ruling in fair use decision — The order means that, for now, the Ninth Circuit will not weigh in on the question of whether copying works from notorious pirate sites can be excused if the works are later used for something that qualifies as fair use.

Outcry as Meta lets users make AI images from public Instagram profile pics — The feature, which is turned on by default, facilitates the creation and manipulation of user photos with Meta’s artificial intelligence tools.

Taylor Swift Granted Motion to Dismiss ‘Absurd’ Copyright Lawsuit Where Woman Claimed Pop Star Copied Her Poems — Days after her Madison Square Garden wedding, the pop star won a motion to dismiss claims she copied from a Florida poet to create lyrics “for more than a dozen songs featured on her albums Lover, Folklore, Evermore, Midnights and The Tortured Poets Department.”

French regulator slams Meta over failure to pay media outlets — A 2019 French law requires online service providers, including social media companies, to obtain authorization from publishers of online news services or press agencies before any reproduction or communication to the public of all or part of their press publications in a digital form and compensate them for exploitation of those rights. This week, the French competition authority ordered Meta to resume negotiations with two news agency groups in order to meet its obligations under the law.

Reggaeton artists’ unoriginality argument falls short in Jamaican producers copyright suit — At issue is “Fish Market”, a 1990 musical work and sound recording that was used without authorization by numerous reggaeton producers. The court denied summary judgment, saying, “The record presents genuine disputes of material fact regarding the identification, characterization, and protectability of the Fish Market elements, precluding summary judgment in either side’s favor.”

Kim Dotcom Loses Court of Appeal Bid to Block Extradition to the U.S. — The alleged operator of Megaupload continues to fight, unsuccessfully, his extradition from New Zealand to the US to face criminal copyright infringement charges originally filed in 2012.

Hearing Confirms Congress is Moving Ahead with Legislation to Protect Creators Against Piracy and Digital Replicas — A summary of the hearing, held this week by the House Judiciary Committee, Subcommittee on Courts, Intellectual Property, Artificial Intelligence, and the Internet. Witnesses included actor SAG-AFTRA president Sean Astin.

New copyright shareholder derivative suit v. Microsoft, its directors & officers. Microsoft becomes 2d company hit with new legal theory — As reported by Prof. Ed Lee, the new lawsuit alleges “that corporate officers or the board of directors approved copyright infringement, exposed the company to substantial risk of copyright liability, or made material misrepresentations about such conduct.”

AI-generated search summaries and personality rights: Munich court grants preliminary injunction against Google — Writing for IPKat, Marcel Pemsel analyzes a recent decision from a court in Germany that held Google liable for AI-generated answers that incorrectly claimed the plaintiff was known for fraudulent or dubious business practices.