Artificial intelligence tools for generating music are not new While there are dozens of companies operating in this field, the music industry has taken aim at two prominent companies and initiated legal action The two services provide commercial licensed output of vocals and music
Suno and Udio have raised millions of dollars in venture capital over the past year to train them to provide complete, complex, natural sounding songs based on vast amounts of data More recently, they have added new features such as using your own music and sounds as a starting point
Since its launch, questions have been raised by both companies over the source of the data used to train the AI models, with speculation that copyrighted material, including songs from major music labels and big artists, was used
Neither company has commented on this, but in both cases moderation and restrictions had to be placed on the tool to prevent it from creating music that sounds like a living artist
The Recording Industry Association of America (RIAA) says this is not enough Several labels, including Sony and Universal, have filed lawsuits against major startups for "stealing music to spit out similar works" and are seeking $150,000 per song in compensation
"Unauthorized services like Suno and Udio that claim it is 'fair' to copy an artist's life's work and use it for their own benefit without consent or compensation set back the promise of genuinely innovative AI for us all," RIAA CEO Mitch Grainger warned
Lawsuits have been filed against Suno AI in federal court in Boston and against Uncharted Labs, the developer of Udio AI, in New York In both cases, the claims are nearly identical: that training with copyrighted data is copyright infringement
The RIAA-supported court case noted that there was "mass infringement of copyrighted sound sources" and alleged that they were "copied and used without permission by two multi-million dollar music generation services"
The lawsuit further states that "AI companies, like all other companies, must comply with laws protecting human creativity and ingenuity" and that "nothing exempts AI technology from copyright laws or exempts AI companies from following the rules"
Many of the claims are similar to lawsuits against AI image generation companies, bringing to the forefront the debate over whether training data used for generated AI constitutes fair use or should be licensed
Mikey Shulman, CEO of Suno AI, told Tom's Guide that Suno's mission is to enable everyone to make music Our technology is transformative and designed to generate entirely new outputs, rather than storing and playing back existing content," he said in a statement And he said in a statement: "That is why we do not allow user prompts that refer to specific artists"
"Suno is built for new music, new uses, and new musicians We value originality," he added
Shulman said he tried to explain this to the label, but they would not listen He said, "We were willing to explain this to the corporate record labels that brought this lawsuit (and indeed, we did), but they would not listen
Some of the largest AI companies, including OpenAI, Google, and Anthropic, have claimed "fair use" exemptions for data used to train artificial intelligence models, and OpenAI has previously stated that without vast amounts of data, models like the GPT-4 stated that it would not be possible to obtain a model like the GPT-4 without using vast amounts of data
The threat of litigation may have been factored into the funding when Udio and Suno first started their business There are two industries that have been in conflict for some time, and generative AI is bringing them to the forefront In other words, it's about how tech can use the data created by creators
If the labels are successful in arguing that copyright applies to the training data and that there is no fair use exemption, Udio and Snow could be awarded $150,000 for each work used in the training data
If there are no changes in copyright law before this litigation is completed, the outcome of this case will most likely take years to reach a Supreme Court decision on fair use
If the fair use claim is rejected, and the RIAA argues that the AI tools should be rejected because they are attempting to output content that replaces input, it would have far-reaching implications for the entire AI field, perhaps significantly slowing the development of these tools and increasing their use would increase their costs
However, companies like OpenAI, Meta, and Google are already working on licensing agreements with publishers and other content creators, and Adobe is training Firefly with licensed data
Music labels are not opposed to AI in music, and many have already signed agreements with Google to allow artists' songs to be used to generate sound-alike tracks on YouTube, essentially want an opt-out and licensing agreement
Glazier says: "The music community is embracing AI, and we are already partnering and collaborating with responsible developers to build sustainable AI tools around human creativity that put artists and songwriters in charge
Music licensing company Rightsify's Hydra II is trained on its own tracks, and Casette AI says its creator uses licensed content in its datasets, among other AI that focus on this aspect and use licensed music There are a number of startups
The problem is that despite some impressive features, and more ethical training data, neither allow vocals
Universal Music Group and Roland, a musical instrument company, have formed AI For Music to promote musicians' rights, and as part of this ethical initiative, they offer key principles for the use of AI in music production
The principles include:
AI for Music describes this as a "new era of music creation," but argues that guidelines have been established that emphasize the importance of leaving humans in the creative process and the need to protect artistic integrity
To me, the lawsuit, what is being alleged, and the promise of working with AI evoke memories of Napster, the last great battle between techs who want to move fast and the music industry who want to make sure their content gets paid
The file-sharing service, which ended in 2001 after facing multiple copyright infringement lawsuits, was called the next big thing for the music industry while it was going strong
The lawsuits did not end digital music, and this spurred the industry to sign deals with Spotify, Apple Music, and others, leading to today's music streaming sector Whether that was good for musicians is another story
As for Napster It's now a pretty good legitimate music streaming service, even offering AI-generated playlists
Comments