Computational Music. The Legal Implications Of AI In the Music Industry: A Case Study Of Suno.ai & what It offers To India

By understanding the legal landscape and adopting best practices, AI companies can innovate responsibly, ensuring that their creations amplify human creativity without infringing on the rights of others. As AI continues to evolve, striking this balance will be key to fostering a vibrant and legally compliant music industry.

The recent lawsuit filed against Suno.ai in the United States by major record labels, including UMG Recordings and Sony Music Entertainment, highlights the complex intersection of artificial intelligence and copyright law in the music industry. This case underscores the challenges and mistakes that AI companies may face, offering valuable lessons for future innovators and entrepreneurs.

This article has a few blurbs for all stakeholders viz., AI developers / entrepreneurs, copyright owners, and regulators / government, on how to prepare for and embrace the AI waive in the music industry, especially in the Indian context.

The Suno.ai Case: A Summary

In the case filed in the United States District Court for the District of Massachusetts, several major record labels have accused Suno, Inc. of copyright infringement. Suno.ai’s generative AI service allows users to create digital music files that closely mimic existing copyrighted sound recordings. The plaintiffs allege that Suno.ai unlawfully copied and used their copyrighted sound recordings to train its AI model, resulting in outputs that compete directly with the original works. The plaintiffs are seeking injunctions and damages for what they describe as “massive and ongoing” copyright infringement.

Apparent Missteps by Suno.ai and Lessons for AI Innovators:

1. Training with Proprietary and Public Data: Suno.ai claimed publicly that it used both proprietary and public data for training its AI model. The public data, in this case, it will be music or sound recordings that are available in public domain (i.e. without a paywall), is also subject to copyright protection. While this statement looks good to attract more users to the platform, however, in absence of any legislative, regulatory and judicial clarity on the legitimate use of the “Proprietary and Public Data”, the model trained on such data will be up for legal challenges by the copyright owners.

Lesson: Either AI developers must ensure that all data used for training is legally acquired and appropriately licensed with clear documentation and agreements with copyright holders can prevent legal disputes, or have a carefully crafted cheat-sheet on “what-to-say” and “what-not-to-say” about the model training process, when making any statements in the public or in any other medium which could be discoverable under the law for the purpose of evidence or trial. An experienced techno-media lawyer will come very handy for creating this cheat-sheet.

2. Asserting ‘Fair Use’ Prematurely: Suno.ai’s defense strategy included asserting ‘fair use,’ implicitly acknowledging that the use of copyrighted material was unauthorized. Fair use is a complex defense that does not automatically apply to all uses of copyrighted material and requires careful consideration of the context and purpose.

Lesson: Relying on fair use without thorough legal analysis can be risky, as fair use recognizes an existing alleged infringing use. AI companies should seek legal advice to understand the nuances of fair use and explore alternative legal justifications for their use of copyrighted material. Again, the cheat-sheet of “what-to-say” and “what-not-to-say” given the ambiguity on whether training AI model from Proprietary and Public Data is fair-use or not.

3. Investor Communications: Statements from investors indicated a willingness to underwrite legal risks, with additional statements discussing legal arbitrages as market opportunities. While disruptive business models do test the settled legal boundaries, but publicly admitting that investors and builders have a reasonable belief that the business may face legal heat or that funding from investors will be used to underwrite legal risk, would surely give enough fodder to the copyright owners.

Lesson: It is always good that investors and companies should prioritize legal compliance and risk management. However, if the business is edgy and needs to operate in grey or unsettled arena, it is important for investors and builders to carefully articulate these grey areas in their public statements. Additionally, the communications between builders and investors aren’t protected under legal privileges. Therefore, it may be prudent to wrap these discussions with an advocate / legal professional in such a manner to protect its disclosure or reliance in a legal proceeding, which may bring the business to a total standstill.

4. Not being smart with Producer Tags: A producer tag is the element of a song, typically at the beginning of it, inserted there by the song’s producer – a short sound meant to familiarize the listener with who is responsible for the production of the song. A simple example is the producer tag of the sound “Yo Yo Honey Singh” in most of the Honey Singh’s songs. Suno.ai was found to have used producer tags in its training material, replicating these tags in the output even when not prompted. This suggests unauthorized use of copyrighted material, further strengthening the plaintiffs’ case.

Lesson: Train the trainer. AI companies should train their scraping and training tools to avoid using identifiable elements like producer tags of copyrighted works without permission. Ensuring that AI outputs do not inadvertently reproduce protected elements can help mitigate infringement claims.

5. Direct Mimicking of Existing Songs: Prompts used by Suno.ai often led to the generation of music that closely resembled specific existing songs, raising questions about whether the service was merely facilitating access to copyrighted material rather than creating new, transformative works.

Lesson: AI developers should design systems that create truly original and transformative content. Avoiding prompts that directly fetch or replicate existing copyrighted works can help demonstrate innovation and originality.

Legal Parameters for Music AI Development in the Indian Context

To navigate the complex legal landscape, AI builders and entrepreneurs in the music or media industry can discuss with their legal advisor on whether the following strategies align with Indian laws:

1. Intermediary Role: AI companies could position themselves as intermediaries, facilitating the creation of content without directly infringing copyrights. Implementing supervised learning and human-in-the-loop capabilities by independent third parties can help ensure that the AI operates within legal boundaries with this safe harbour under Information Technology Act, 2000 and fair use principles under Copyright Act, 1957, such as transient or incidental storage. There are judgements that have protected e-commerce companies, even when they were providing managed warehousing and logistics services of infringing goods, from being held liable for selling the infringing goods by independent sellers on the e-commerce platforms.

2. Search Engine Analogy: Drawing parallels with how search engines operate i.e. they also web crawl or web scrape billions of webpages and create their index and caches for ranking and publishing based on the search queries, similarly music AI companies can argue that their systems scrape the public data to build vectors or tokens for AI transformer technologies rather than repositories of copyrighted content. By facilitating creation of a newly generated content from the stored vectors and tokens, but without storing or reproducing the copyrighted material, AI services could possibly stay within legal limits.

3. Judicial precedents: Music AI builders can carefully assess how they can take benefits from various court judgments on the personality rights vis-à-vis constitutional right to do business and certain fair use principles around contents that are mimicry, parody and satire.

How copyright owners can start guarding against music AI disruptions to their rights.

Copyright owners need to buckle up to embrace AI tools while continuing to relish the economic benefits of their creativity. Musicians and singers can start putting producer tags in their music and re-open all their licensing contracts or online terms of use and add clauses to restrict the world to use the content for AI training. In order to build for the future, copyright owners may consider converting their own repertoire into AI trainable vectors or tokens and make them available to music AI builders at mutually beneficial commercial terms. This will help AI builders save time and money for burning tokens, using high cost GPUs, and a lot of electricity and most importantly expensive human capital.

Proactive steps by the govt. to make India’s music copyright AI available without any ambiguities and business disruptions.

Certain legislative steps were taken by the govt. in the past to support the business models of cover versions. This could be the right time for the govt. to work closely with the copyright board, copyright owners, copyright societies, music AI builders, and AI investors, to frame adequate laws or regulations – including the regimes of compulsory licensing – to ensure copyright owners get their rightful share considering the commercial realities of music AI revenues, AI builders can build with legal and cost certainties, and there are no music AI monopolies i.e. to prevent 1 or 2 large music AI companies to sign deals with large music companies, and deprive many start up music AI companies to train their AI.

Judicial Considerations.

Till such time we get legislative clarity and predictability, in resolving disputes involving new technologies like AI, courts should be cautious about granting immediate injunctions that could stifle innovation. Instead, allowing for revenue-sharing arrangements or royalties – in case the court sides in favour of copyright owners and not music AI builders, can balance the interests of copyright owners and technology innovators. This approach ensures that copyright holders are compensated without hindering technological progress.

Conclusion

The Suno.ai case serves as a crucial learning opportunity for AI developers, investors, copyright owner, regulators, and legal professionals in the music industry. By understanding the legal landscape and adopting best practices, AI companies can innovate responsibly, ensuring that their creations amplify human creativity without infringing on the rights of others. As AI continues to evolve, striking this balance will be key to fostering a vibrant and legally compliant music industry.

If you are curious about how suno.com or suno.ai works, here is a slow hind romantic rap song on breakfast on a beautiful Sunday morning in Mumbai on a rainy day, that I prompted suno.com to create for me https://suno.com/song/019c7597-8c55-48a3-84a5-089b54767a9f. It took less than 1 minute.

Disclaimer – This article is for informational purposes only and should not be construed as legal or professional advise. The copyright is owned by the author.