Herwig Lux’s Post

Your papers are training AI models Artificial-intelligence developers are buying access to valuable data sets that contain research papers — raising uncomfortable questions about copyright. Anything that is available to read online — whether in an open-access repository or not — is “pretty likely” to have been fed into an LLM already, says AI researcher Lucy Lu Wang. “And if a paper has already been used as training data in a model, there’s no way to remove [it].” Nature | 6 min read #AI #LLM #papers #copyright

"AI firms often trawl or "scrape" sources like fiction and non-fiction books, newspapers, and social media to train their AI models, which has already caused plenty of legal controversies. Alongside Simon & Schuster, Hachette, HarperCollins, and Macmillan Publishers, Penguin Random House is considered one of the "Big Five" English language publishers, controlling 80% of the US book trade as of 2022. Penguin has amended the copyright wording on all its titles worldwide and across all its imprints. It now reads: "No part of this book may be used or reproduced in any manner to train artificial intelligence technologies or systems." According to The Bookseller, the new wording will appear on all its new titles and any reprinted old titles."

“In the event the human authorship requirement is discarded allowing the registration of AI-generated works, additional questions implicating copyright litigants are likely to arise: Who is that author and/or owner of works created by generative AI – is it the AI, its developer, the AI owner/licensee, the human who prompted creation of the work at issue, or perhaps even the owner/author of works on which the AI was trained? What are the implications for the work-for-hire doctrine under these circumstances? And what are the ramifications for where liability arising out of such works can lie? While many questions remain, one thing is certain: as long as the copyrightability of AI-generated works remains in question, companies are well-advised to consider the potential un-registrability – and thus unenforceability – of AI-generated works as they contemplate using AI tools in the creation of works they may ultimately want to protect.”

As I am writing a chapter for a book on AI and Ethics in Higher Education, the three questions I always bring up in presentations and discussions on copyright seem increasingly important: 1. What are the Rights and Responsibilities of the Copyright Owner? (e.g.: should they allow use of their works through the Fair Use Doctrine or can they assert their rights and prevent this?) 2. What are the Rights and Responsibilities of the Copyright User? (the AI user who uses copyrighted works to fine-tune AI tools and/or obtains product on the frontend using copyrighted materials in prompts) 3. Which of the Two is Generative AI? Or, is it Both? or Neither? What are your thoughts on these questions?

As the landscape of artificial intelligence continues to evolve, it’s crucial for us to stay vigilant about protecting intellectual property. Penguin Random House (PRH) has taken a significant step in this direction by updating its copyright wording to explicitly prohibit the use of its books for training AI technologies. This move underscores PRH’s commitment to safeguarding the rights of authors and ensuring that their creative works are not exploited without consent. The new copyright statement will appear in all new titles and reprints, clearly stating: “No part of this book may be used or reproduced in any manner for the purpose of training artificial intelligence technologies or systems.” This approach is a development in the publishing industry, that sets precedent for other publishers but is yet to see if they will follow. This position by PRH highlights the importance of responsible innovation and the need for clear legal frameworks to protect creators’ rights in the digital age. Interestingly, while PRH takes this firm stance, other major publishers like Taylor & Francis, Wiley, and Oxford University Press have chosen to license their works to AI companies. This decision by PRH could influence how AI companies operate. Let's continue to monitor developments and discuss further. Alfredo Medina + AI #Medinalegal #Medinalegal360 #CopyrightProtection #AI #Publishing #IntellectualProperty #Innovation 

🚨 Breaking News in AI & Copyright 🚨 Dutch copyright enforcement group BREIN has successfully taken down a significant language dataset crucial for training AI models. This action underscores the complex intersection of intellectual property rights and the rapidly growing field of artificial intelligence. With more datasets being created and shared globally, it's essential for all stakeholders, including developers, researchers, and IP professionals, to navigate these legal landscapes carefully. 🔗 Read the full article on Reuters: [Link](https://lnkd.in/ef7a_J5m) This development raises important questions: 💡 How can we balance the need for rich data in AI development with the protection of intellectual property? 💡 What frameworks or guidelines could be implemented to address such challenges? Your thoughts and insights on this topic are highly valued. Let’s discuss in the comments! #ArtificialIntelligence #AI #MachineLearning #DataEthics #IntellectualProperty #TechNews #AIResearch #DataScience

Should AI generated art be eligible for copyright? Jason Allen, whose AI-generated artwork "Théâtre D'opéra Spatial" won a state fair competition says yes it should be. He is appealing the U.S. Copyright Office's refusal to register his work. The Office claimed the piece lacked sufficient human authorship, as it was mostly created using Midjourney, an AI tool. Allen argues that crafting precise prompts and spending over 100 hours refining the output qualifies as human authorship. 📚 The AI Policy Newsletter: https://lnkd.in/eS8bHrvG 👩💻 The AI Policy Course: https://lnkd.in/e3rur4ff 🔗 Link to article: https://lnkd.in/estgKScg #AIpolicy #ArtificialIntelligence #TechPolicy #AIGovernance

#AIActUpdate - AI Training Data Disclosure Template Unveiled 🔍 The European Commission’s AI Office just released a template requiring AI providers to disclose detailed data usage—from pre-training to fine-tuning. ⚙️ What’s Required? 1️⃣ Break down data types (text, audio, video, etc.) 2️⃣ List major sources (datasets, open web) 3️⃣ Explain copyright compliance (per the EU Copyright Directive) 💡 Implications for Companies: Right Holders want full transparency, while AI providers fear revealing trade secrets. This tension could spark disputes between right holders and any company using AI solutions. What is your view? Will this boost responsible AI—or create more conflict? 📌 Reliance on the text and data mining copyright exception will become increasingly relevant. Read more on the topic 👉 https://lnkd.in/dmjCBNbb #ailaw #aiact #ai #copyright #artificialintelligence

📣 BREAKING: Historic EU Ruling on AI and Copyright The Hamburg Regional Court has ruled that creating datasets with copyrighted content for AI training is lawful under text and data mining (TDM) exceptions for research. What does this mean? Creating datasets involves collecting large amounts of data—like images, text, or videos—for training AI models. For example, a dataset might include stock photographs or news articles, which are analysed to identify patterns (e.g., pairing an image with its description). This ruling clarifies that downloading and processing copyrighted content for scientific research, like training AI, is permissible under certain legal exceptions. Key Points: 1️⃣ Is downloading copyrighted data for TDM lawful? Yes, for non-commercial research. 2️⃣ What counts as machine-readable opt-outs? Still debated (e.g., “do not use” notices on websites). 3️⃣ Can this boost research-industry collaboration? Likely, by lowering barriers to using copyrighted data responsibly. All in all, I believe this decision paves the way for innovation while raising some eyebrows about copyright and AI ethics. What’s your take? 👇 #AI #Copyright #GenerativeAI #TDM #Innovation #ArtificialIntelligence #EU

Penguin Random House is adding a warning to its books’ copyright pages to prevent them from being used to train AI models. The statement, added to new and reprinted books, prohibits reproducing any part of the books for AI training purposes. This makes Penguin Random House the first major publisher to address AI concerns on copyright pages, though the company is not entirely opposed to AI. They plan to use AI tools selectively, while strongly defending their authors' intellectual property. Subscribe to our Newsletter Plugaiinc.com #AIandCopyright #PenguinRandomHouse #Publishing #AITraining #IntellectualProperty