Tag: copyright

The Internet Archive (IA) has recently enjoyed a string of successes, with its Wayback Machine service surpassing the 1 trillion page archive milestone in October and being designated a federal depository library by a U.S. senator. However, in a recent interview with Ars Technica, IA founder Brewster Kahle admitted that while the organization has survived years of copyright litigation, the cost has been a significant blow to one of its core projects, the Open Library. An IA spokesperson confirmed that the organization is currently free of major lawsuits, and the active threats to its collections have been temporarily averted. The "National Emergency Library" during the pandemic became the catalyst for this lawsuit, stemming from the IA's long-standing Open Library initiative. The plan employed a model called "Controlled Digital Lending" (CDL), where for every physical book a library owned, it could simultaneously lend a digital scanned copy to one user. This model had not resulted in major lawsuits over the past decade. However, the turning point came at the beginning of the COVID-10 pandemic in 2020, with libraries worldwide closing. The Online Archive decided to temporarily suspend borrowing limits and launched the "National Emergency Library"...

AI search engine Perplexity AI recently announced a multi-year licensing partnership with Getty Images, a globally renowned image library. Perplexity will gain API access to Getty Images' vast image library, including its complete stock and editorial images, and plans to further integrate this data into its AI search and exploration tools. Key focus of the partnership: Strengthening attribution and educating users on legal use. The key to this collaboration lies in addressing the thorny issues of copyright and attribution in the AI ​​field. According to a Getty Images press release, Perplexity will commit to "improving how it displays images, including image credit with source links," and will further educate users on how to legally use licensed images. This partnership is seen as a crucial step for Perplexity in addressing the numerous legal challenges it has recently faced, given its entanglement in copyright litigation. With the proliferation of generative AI tools, Perplexity has become the target of numerous copyright lawsuits due to its content scraping and presentation methods, including: • Lawsuits from Japanese media: In August, two major Japanese media groups, Nikkei and Asahi Shimbun, sued Perplexity, alleging that it illegally copied and stored content from their servers and provided inaccurate information sources. • Lawsuits from Reddit: In October, Reddit also filed a lawsuit against four AI companies, including Perplexity, accusing them of scraping platform data without proper permission. • Lawsuits from dictionary publishers: Merriam-Webster and its parent company, Encyclopædia Britannica, also filed an infringement lawsuit against Perplexity in September. Getty Images, which previously opposed AI, has now shifted to collaboration...

A U.S. court has issued a new ruling in the copyright infringement lawsuit filed by The New York Times against OpenAI. Federal Judge Ona T. Wang signed a new order on October 9th, U.S. time, formally overturning the previous ruling requiring OpenAI to "indefinitely retain and separate all ChatGPT output data," meaning the controversial data preservation order has officially ended. The controversy began in late 2023 with the copyright lawsuit, in which The New York Times accused OpenAI of using its news content as training data for its AI model without authorization, constituting copyright infringement. To gather evidence, the court ordered OpenAI in May of this year to retain all ChatGPT conversation records and output content, allowing The New York Times to examine whether the model contained its own works. However, OpenAI appealed, arguing that the order was "too broad," interfering with model operation and potentially infringing on user privacy. The new ruling relaxes the scope of preservation, but some data must still be retained. According to the latest court order, OpenAI is no longer required to retain all ChatGPT conversation records from September 26th. However, some data must still be retained, including records previously subject to the retention order and conversation data from accounts marked as related by The New York Times. The court also noted that The New York Times can expand the scope of the marking in the future, requiring OpenAI to retain more specific account data for future comparison or evidence. OpenAI: Alleviating User Privacy and System Burden. For OpenAI, this decision represents a significant relaxation in its operations and privacy management. Due to the massive volume of ChatGPT's conversation records, the previous retention order not only burdened server resources but also raised concerns about the risks of long-term storage of user data. OpenAI had previously emphasized that training data and general user conversation records are not entirely identical, and excessive retention of user interactions would "undermine the trust foundation for AI use." The Tug-of-War Between Copyright and AI Continues. Although the retention order was lifted, The New York Times' lawsuit against OpenAI continues. The core of the case lies in whether AI models "fairly use" copyrighted content and the legal status of media publications in the development of generative AI. The progress of this case not only affects OpenAI but is also considered a key indicator for the AI ​​industry on issues of data transparency and content licensing. As more media organizations and creator groups file similar lawsuits, the line between AI and news publishing is facing redefinition. Conclusion: A New Balance Between AI Development and Regulatory Oversight. This ruling allows OpenAI a temporary respite, but it also highlights the conflict between generative AI and the existing copyright system. How to strike a balance between protecting creative rights, safeguarding privacy, and promoting innovation will be a long-term challenge for global regulatory agencies and the technology industry.

The copyright lawsuit filed against the Internet Archive in 2023 by Sony Music, Universal Music Group (UMG), and several other music labels has now come to an end. Both parties recently filed documents with the court confirming a settlement and expect to formally withdraw the lawsuit within 45 days. While details of the agreement have not been disclosed, the Internet Archive stated in an official blog post that both parties have reached a confidentiality agreement regarding all disputes and will not make any further comments. The focus of the lawsuit: The "Great 78 Project" At the heart of the lawsuit was the Internet Archive's "Great 78 Project." This project aimed to digitize and preserve nostalgic songs and music recorded on vinyl records at 78 rpm, including recordings of American singers such as Frank Sinatra, Ella Fitzgerald, Billie Holiday, Miles Davis, and Louis Armstrong, and make them available online. The original lawsuit targeted 2749 recordings, later expanding to 4142, including classic songs such as "White Christmas," "Sing, Sing, Sing," and "The Christmas Song." The record company argued that these tracks were not at risk of being "lost, forgotten, or destroyed" because they had been sold and licensed for playback through various streaming services; therefore, the Internet Archive's public sharing constituted unauthorized infringement. If the court ultimately rules in favor of the record company, the Internet Archive could face damages of up to $150,000 per track, totaling hundreds of millions of dollars. This is not the first time the Internet Archive has faced copyright litigation. Previously, the Internet Archive faced a lawsuit from several publishers for offering 1.4 million ebooks free of charge through its "National Emergency Library" service launched during the pandemic, and ultimately lost. The court ruled that the service did not comply with fair use principles and ordered its cessation. This settlement signifies that the Internet Archive has chosen to avoid further legal risks in the face of music copyright litigation. However, this move has also raised concerns about how future academic, historical, and music archiving projects will strike a balance between public interest and copyright protection. The tug-of-war between public archives and copyright protection...

With the rapid development of generative AI technology, tech companies are facing a series of copyright disputes. Following accusations against several AI companies for using copyrighted works without permission, Apple has recently been embroiled in a class-action lawsuit filed by two authors. The lawsuit alleges that Apple used pirated ebooks during the training of its AI system, thereby infringing on the copyrights of the authors. The main plaintiffs are American authors Grady Hendrix and Jennifer Roberson, both of whom have published numerous well-known works. The complaint claims that Apple used a large number of unauthorized pirated ebooks, including content from copyrighted works, in its AI training process. Apple allegedly used this content for AI model training and its "Apple Intelligence" service without obtaining the original authors' consent or paying licensing fees. The plaintiffs further allege that Apple copied and used copyrighted works to train AI models that could compete with the original works, thereby diluting the market value of these works. They argue that this not only deprives authors of control over their creative work but also allows Apple to gain enormous commercial profits through illegal means. Due to the sheer number of works involved in the pirated content, this case is currently seeking class-action status, potentially covering thousands of authors. It's worth noting that this is not the first copyright lawsuit related to generative AI. In recent months, AI companies including OpenAI and Anthropic have faced similar lawsuits. OpenAI is embroiled in litigation with The New York Times and the oldest non-profit news organization in the United States, while Anthropic reached a $15 billion settlement earlier this year with a class-action lawsuit involving 50 authors, with each participating author receiving approximately $3000 in compensation per work. Compared to other AI companies, Apple has consistently emphasized its focus on privacy and data protection in its AI training process. However, this accusation highlights another layer of controversy: the reliance of AI model training on the "legality of data sources." If the court finds that Apple did indeed use unauthorized pirated content, it could not only impact its AI development strategy but also further deepen public concern about "massive content infringement" within the industry. Apple has not yet issued a formal response to the lawsuit, but industry insiders generally believe that as the commercialization of generative AI accelerates, establishing a more transparent data licensing and compensation mechanism will become a core point of contention between tech giants and creators. The development of this case may also serve as an important reference case for future copyright regulations in the AI ​​industry.

The U.S. District Court for the Northern District of California has made a significant ruling in a case concerning the use of copyrighted content in training artificial intelligence models. For the first time, the court has ruled that such use falls under the category of "fair use," providing a legal defense for AI companies like Anthropic. This ruling has not only drawn industry attention but has also greatly disappointed creators who have long worried about the unauthorized use of their work by AI. The case was filed in 2024 by several authors, including Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, who accused Anthropic of copyright infringement for unauthorized scraping and use of content from their published books to train large language models (LLMs). However, Judge William Alsup, who presided over the case, ruled that although the content used by Anthropic may have come from unauthorized sources, its purpose and nature in training the AI ​​model met the fair use principle, and such use was legally transformative, therefore not constituting direct infringement. The ruling stated that even if Anthropic initially used a so-called "pirated ebook" and subsequently made a purchase, while not completely excluding its liability, the corresponding statutory damages might be adjusted accordingly. In other words, while acknowledging the reasonableness of using copyrighted content in the development of artificial intelligence, the court did not completely abandon the possibility of pursuing illegally obtained content. This ruling marks the first substantive determination by the US court system regarding the source of training data for generative artificial intelligence, and may provide some reference for other artificial intelligence technology developers, especially in the development stage where generative artificial intelligence models heavily rely on large-scale datasets for training. It is worth noting that while the court did not completely deny the rights of creators, the overall trend is relatively favorable to the artificial intelligence industry. This outcome is expected to spark more discussions about creative and technological ethics. As the application scope of generative artificial intelligence expands, from text to images and music, which can be quickly generated through models, creators will face increasing commercial challenges and risks to the value of their works. Similar lawsuits may continue to increase in the future, and are expected to further promote legislation to establish clearer regulations regarding the source of artificial intelligence data, licensing mechanisms, and compensation systems.

OpenAI and Google recently requested copyright exemptions from the US government for the development and application of artificial intelligence (AI) technologies, aiming to reduce restrictions on AI training. This proposal echoes the Trump administration's call for input from private companies to further clarify the norms for AI development within the US, ensuring safety while maintaining America's AI advantage. OpenAI and Google's respective proposals essentially aim to allow the use of copyrighted content within reasonable limits, and even the use of data in exceptional circumstances for AI training, to train more advanced and complete AI models. Both companies emphasized that this approach would not affect copyright holders, would prevent discrepancies between the training process and real-world data, and would avoid the long waiting times for licensing applications that could hinder AI model development. OpenAI further explained that without using copyrighted data for training, current AI model technology would be impossible. However, many authors and copyright holders are concerned that many current AI models do not merely "reference" the original content, but often directly "copy and paste" it, significantly increasing the likelihood of copyright infringement. Similar controversies have occurred with Google's search engine before, with many authors believing that their work is being recorded by Google's search engine crawlers and subsequently stolen through search engine functions.

Valve's Steam platform recently adjusted its game content sales language, emphasizing that players are actually purchasing a license to use game content, not actual ownership of the game. This change is primarily in response to California's AB 2426 law, scheduled to take effect in 2025. This law requires digital content providers to clearly inform users that the content they are paying for is a license to use, not the digital content itself; otherwise, they cannot use terms like "purchase" or "buyout." In recent controversy, Ubisoft removed The Crew, released in 2014, from its platform and confirmed that it will cease server operations after March 31, 2024. This means that players will no longer be able to purchase the game, and those who already purchased it will no longer be able to download it. Ubisoft even removed the game from its player database via an online update. Similarly, Sony also experienced a situation where Discovery Channel merged with Warner Bros. to form Warner Bros. Discovery. As a result, the Discovery Channel content previously available on the PlayStation platform was removed from player databases due to changes in the cooperation agreement. Content previously included on Netflix may also be removed from the service due to the termination of content cooperation agreements. As previously stated, users primarily acquire usage rights through the service, not ownership of the content itself. Other software content available through subscriptions, such as from Adobe, or from Microsoft's Windows operating system, offers paid licenses, meaning users do not actually acquire ownership of the software. However, GOG, a game service owned by Polish game developer CD Projekt, does not use digital copyright protection for its content, so there is no need to check copyright usage status online during game installation. 　　

Nintendo has recently updated its guidelines to include additional prohibited uses of its copyrighted material on online distribution platforms. These include images not normally visible in the game, items requiring cheating or hacking to obtain, and content obtained by dismantling game files. This aims to prevent further misuse of game content. Previously, Nintendo had established strict regulations for the online use of its copyrighted material, while also allowing some flexibility to ensure freedom of derivative works. This included allowing users to share publicly released Nintendo videos without profit, and to use Nintendo's copyrighted material within reasonable limits, including a certain degree of original content, and to share this content on monetizable platforms like YouTube. The newly added guidelines further prohibit users from sharing images not normally visible in the game, items requiring cheating or hacking to obtain, and content obtained by dismantling game files. Nintendo also emphasized that it will take legal action against users who do not comply with these guidelines and reserves the right to prohibit the use of all copyrighted material. Nintendo maintains an open attitude towards derivative works using its copyrighted materials. Users are only required to create content with a certain degree of originality if the work is for profit. In non-profit situations, users are allowed to share the copyrighted materials, but only those already publicly available by Nintendo. Regarding the use of derivative content, Nintendo emphasizes that if users do not directly use the copyrighted materials, they are not subject to these regulations. However, users must still comply with the relevant laws and regulations of their own country or region and create content within reasonable limits.

A recent report released by the U.S. Copyright Office (CCO) argues that with the increasing speed, sophistication, and scale of digital content creation using artificial intelligence (AI) technology, the government should enact new laws to regulate it, otherwise, the impact will be even greater. The CCO's assessment points out that current laws are insufficient to address the problems caused by digital content reproduction, even extending the impact to the state level, and that current federal law is inadequate to control the effects of such problems. The report also recommends that the U.S. government develop security provisions, including requiring service providers to offer quick removal of unauthorized digital copies. It further calls for respecting everyone's right to control the legality of their likeness usage, arguing that the harm caused by misuse of likenesses is not limited to celebrities but also affects ordinary people, emphasizing that likeness rights apply not only to specific individuals but also to all individuals. This report is just one of many investigations by the CCO regarding the application of AI technology, illustrating that there are many more issues to be discussed in the development of AI technology applications, and that current laws may need to be adjusted in an up-to-date manner. In recent developments in artificial intelligence technology, much controversy has arisen regarding whether content disseminated on online platforms can be legally used for AI training. Many website operators believe that even if content is publicly available online, it is still protected by copyright, and therefore AI companies should not access or use it arbitrarily. However, some AI companies believe that citation is permissible under reasonable circumstances, while many more people have expressed their unwillingness to have their personal images or published content used for AI training.