Understanding AI Training Data Rights in Music

It's time for artists and rights holders to get in the game! I commend artists taking action by going the legal rout to protect their rights against the onslaught of generative AI companies training on their data. The issue at hand is a metadata issue. Since the dawn of the Napster age, digital music has had no digital provenance. No digital ID. Attribution is an afterthought on platforms. As its not their core business to care if the data is properly tagged. Industry rights-holders and distributor/aggregators have largely contributed to this paradigm by 1. Not publishing an open database containing all metadata about music IP (splits, contact payment info, terms of usage, etc.) and 2. Not prioritizing the importance that metadata information is publicly required for any music IP to exist on the internet. Machines can read this information. This would then make it possible for anyone training AI models on music to have to abide by those rules or else they would be clearly choosing to commit a criminally liable act. So to my original point if you're an artist or rights holder, get in the game and make sure all of your metadata is up to date and open sourced. That's what we did with Grimes and then we built a way for an AI model to be built on top of her metadata and remunerate her when people use her music IP in their creations. The next way to get in the game is to cold-call/email/DM any of these generative AI companies and let them know you want to collaborate. Writing letters leaves it in the hands of lawyers vs. creators. https://lnkd.in/g3N7aMQC

Do you know what you signed away? When you sign up for any online service, you agree to the Terms of Service. (no, saying "I do not grant permission to..." does not revoke what you signed away when you agreed to the terms of service, that's not how contracts work) Many content creators are understandably upset about their works being used to train AI models. But there's a very, very good chance you agreed to exactly that. Go through the Terms of Service for any online service you use. Look for this simple phrase: "create derivative works". A derivative work is a work made from another work. I could paint horns on the Mona Lisa and the new work is a derivative work. If you signed a license with me that permits me to make derivative works, I'm allowed to do that on any copyrighted work. And what is an AI model? You guessed it. It's taking an existing work and making a mathematical representation of it - a translation, as it were, a derivative of it. If you're a musician and you upload your music to YouTube? You have licensed YouTube to use it for AI models. Same for Spotify. If you're an artist and you upload your art to Instagram? You have licensed Meta to use it for AI models. Same for Pinterest, DeviantArt, etc. If you're a writer and you upload your book to Amazon? You have licensed Amazon to use it for AI models. Same for every social network, including LinkedIn. Remember the golden rule of the software age: if you're not paying, you are the product. It's not just ads that companies sell to us. Our content is their product as well, fuel for their AI models - and we agreed to it. Disclaimer: I am not a lawyer. I cannot give legal advice. This is my interpretation of legal documents. If you want to talk to a real lawyer, I recommend Ruth Carter (they/them) and Sharon Toerek. #AI #GenerativeAI #GenAI #ChatGPT #ArtificialIntelligence #LargeLanguageModels #MachineLearning #IntelligenceRevolution

🚨Sony Music Group (SMG) is taking a stand against unauthorized AI training. The music giant is sending formal letters to over 700 generative AI companies and streaming platforms. 🎵 SMG, which includes Columbia Records, RCA Records, and Epic Records, is prohibiting the use of its content for AI training without explicit licensing agreements. 📄 The letters warn against mining text or data, web scraping, and other unauthorized uses of SMG content, covering audio recordings, lyrics, cover artwork, and metadata. 🤖 While recognizing AI's potential, SMG emphasizes that unauthorized use of its content for AI deprives them and their talent of control and appropriate compensation. ⚖️ SMG's letter states that such activities conflict with normal exploitation of works, prejudice their legitimate interests, and infringe on intellectual property rights. 📝 The companies are asked to confirm they haven't used SMG content without permission or, if they have, to disclose how it was used in AI training. 🔗: [NBC News](https://lnkd.in/eNHVXSEc)

SHOTS FIRED -- One of the biggest record labels in the world is issuing notices warning hundreds of companies not to train artificial intelligence models on its content without permission. Sony Music Group — which also owns well-known labels like Columbia Records, RCA Records and Epic Records — has started sending formal letters to more than 700 generative AI companies and streaming platforms prohibiting them from mining text or data, web scraping or otherwise using any SMG content without explicit licensing agreements. • The letter covers a wide variety of content, including audio recordings, musical compositions (inclusive of lyrics), cover artwork and metadata. • The letter also asks companies either to confirm that they have not used SMG content without permission or, if they have, to provide details about how the content was used in AI training. • The European Union’s Artificial Intelligence Act, the world’s first comprehensive AI law, which passed in March, includes a clause requiring providers of general-purpose AI models to make public a “sufficiently detailed” summary of the content used to train those models. New tech. New game. New rules of engagement. 🚀

From Fair Use to Fair Pay: The Changing Tune of AI and Copyright Major music labels - Sony Music Entertainment, Universal Music Group, and Warner Records - have filed copyright infringement lawsuits against two AI music generation companies, Suno and Udio. These lawsuits allege that these AI companies used copyrighted sound recordings without permission to train their AI models. Until last year, one faction of legal scholars argued that AI companies could avoid liability for copyright infringement of training data by claiming fair use. However, recent developments in the publishing industry indicate a move away from litigation and towards licensing agreements. E.g. The Atlantic and Vox Media have signed multi-year licensing deals allowing OpenAI to use their archived content for AI model training. Licensing is not a perfect solution, but it shows that as an industry, we are making some progress in addressing copyright concerns in AI development instead of completely disregarding intellectual property rights. Will the music industry also trend towards licensing agreements, or will they pursue a different strategy to protect their copyrights? Their approach may set a precedent for how the music industry handles AI and copyright. As always, we are tracking AI-related copyright lawsuits on our website. Please follow the link in the comments. Image source: Getty Images #aimusic #copyright #datarights #lawsuits