It may be the most promising new technology in music since the rise of streaming, with the potential to revolutionize how songs are made, listened to and promoted.<\/p>\n
But there are also serious questions. What about protecting artists and their copyrights? Will the systems end up overloading us with generic slop?<\/p>\n
Welcome to the world of music AI, which offers dazzling possibilities for creating songs and connecting artists with fans \u2014 but has also been roiling the music industry, as artificial intelligence companies try to figure out how to build and use the technology legally and profitably, while artists and record companies seek to protect their rights. <\/p>\n
Here are some basic questions and concerns as the music industry ventures into a new era of uncharted possibilities and legal complications.<\/p>\n
Can I use AI to make music? <\/p>\n
Sure can. Complete novices can enter a few simple prompts to create brand-new songs on demand: a country-style birthday track for grandpa, for example, or a background tune for a slideshow.<\/p>\n
What\u2019s the first step? <\/p>\n
For the simplest demonstration, pull up Suno or Udio, two of the leading startups in the field, or Lyria 3, a Google product, and enter some musical terms.<\/p>\n
The more specific your prompts, the more interesting the results. I asked Suno for a summer song with lyrics about drinking margaritas.<\/p>\n
Within a minute or two, it generated \u201cMargaritas in the Sand,\u201d an EDM banger with a singalong chant \u2014 though rhymes like \u201cparty with the band\u201d and \u201cput one in my hand\u201d suggest the technology is not yet any threat to Bob Dylan.<\/p>\n
Then I remixed it, asking for an orchestral version and a polka variant.<\/p>\n
These are examples of \u201cgenerative AI,\u201d in which the system creates a complete song out of whole digital cloth. (How the systems gained the ability to generate such tracks can be a contentious issue. See below.)<\/p>\n
Are musicians and songwriters using AI, too? <\/p>\n
Yes, musicians in the studio today frequently use various machine-learning tools for help in fleshing out a bass line or even choosing instrumental sounds. One system, Tamber, uses what it calls \u201csonic intelligence\u201d to find or tweak sounds based on an artist\u2019s gestures or verbal cues, even abstract ones: a synthesizer that sounds the way chocolate tastes, for example. <\/p>\n
An able assistant for someone working alone to craft something without employing a whole band \u2014 or, perhaps, without putting in years of practice.<\/p>\n
What if I want to use AI to play around with a favorite song? <\/p>\n
Fans may soon be able to go on Spotify or other platforms to remix artists\u2019 official tracks. Maybe cut the verse you never liked. Add the sax you always knew it needed. Perhaps even get a nod from the artist herself.<\/p>\n
That sounds cool, but legally complicated. Is it? <\/p>\n
Indeed. Artists may agree some day to let their work be played around with because it would open up new revenue streams and avenues for audience engagement \u2014 or simply to get an edge on the latest high-tech format. But it\u2019s not a simple decision. Allowing your song to be remixed by someone who doesn\u2019t share your values, or to suit the needs of a political candidate, could be upsetting to artists even if they are getting paid. <\/p>\n
How does copyright law figure in here? <\/p>\n
Are you sure you want to know?<\/p>\n
Much of what is considered legal \u2014 or not \u2014 when it comes to AI and music is being vigorously contested in several ongoing lawsuits. Have AI companies, for example, violated musicians\u2019 rights by \u201ctraining\u201d (or teaching) their systems using copyrighted music? <\/p>\n
Legal issues may also arise when something created by AI imitates the individual styles or voices of specific artists. That may implicate copyrights and what are known as rights of publicity, which protect an individual\u2019s name, image and likeness from unauthorized commercial use. <\/p>\n
When I asked Suno for a \u201c1970s folk-rock ballad in Neil Young style,\u201d a message popped up saying, \u201cWe don\u2019t reference artists directly, so we\u2019ve replaced your prompt with similar styles you might like.\u201d The resulting tune sounded a bit generic to my ears. <\/p>\n
What if artists give their thumbs-up? <\/p>\n
The latest deals will test what fans can do when artists give their permission to place their music in the AI sandbox.<\/p>\n
Warner Music, for example, whose catalog includes songs by stars like Madonna, Bruno Mars and, yes, Neil Young, struck a deal with Suno that will allow the service to use specific artists\u2019 music if they opt in. Because \u201cMadonna-style birthday song for my wife\u201d may be much more interesting than \u201c1980s synth-pop song.\u201d Just who has, or will, opt in remains unclear.<\/p>\n
What can I do with my AI creation? <\/p>\n
The industry is still wrestling over whether to allow users to download their AI-generated track and upload it anywhere, or restrict it to the \u201cwalled gardens\u201d of the programs that generated them.<\/p>\n
Isn\u2019t the world of AI music full of scams? <\/p>\n
There are plenty of legitimate uses of AI by real musicians and curious amateurs, but, yes, scams and fraud are a problem.<\/p>\n
Deezer, a French steaming service, recently said that about 90,000 AI-generated tracks are uploaded to its platform every day \u2014 most of them low-quality slop that no one listens to.<\/p>\n
Of the relative few such songs that get clicks, Deezer says, as many as 85% are fraudulent. The scammer\u2019s playbook: create AI songs, then unleash armies of bots to click on them to generate royalties. Under the financial pools used in streaming, the online platforms pay proportionate fees to the companies that provide content, meaning everything paid to bogus creators reduces what\u2019s left for real artists.<\/p>\n
Catching these fraudsters is a constant game of cat and mouse. Recently, music industry groups, including the Grammys, have called for a labeling system to identify music that was made with generative AI.<\/p>\n
What do artists think? <\/p>\n
They are not unanimous. Some, like star producer Timbaland and performer Grimes, have welcomed the technology and signaled their willingness to experiment.<\/p>\n
Timbaland recently introduced a virtual artist named TaTa Taktumi, whose vocals on the song \u201cGlitch x Pulse\u201d were partly generated by AI software, though humans created 80% to 85% of the song, Timbaland said in an interview with The New York Times. <\/p>\n
But many more have raised strong concerns that generative AI in music will undermine the work of humans and dilute their share of streaming royalties. Two years ago, more than 200 major acts \u2014 among them Billie Eilish, Norah Jones and Stevie Wonder \u2014 signed an open letter from the Artists Rights Alliance, calling the \u201cunchecked\u201d use of AI an \u201cassault on human creativity.\u201d<\/p>\n
Does AI threaten musicians\u2019 livelihood? <\/p>\n
To be determined. But some industry organizations have warned of big potential losses for songwriters, studio personnel and other behind-the-scenes workers.<\/p>\n
AI may also pose a particular threat to genres of music that do not rely on the personalities of specific performers, like soundtracks and background mood music \u2014 both lucrative corners of the business.<\/p>\n
Surely there are more legal complications? <\/p>\n
But of course. This is the music industry and new technology, after all.<\/p>\n
A major battleground is whether the AI companies taught their systems how to make music by using the labels\u2019 copyrighted content \u2014 and, if so, whether they needed permission to do so.<\/p>\n
In 2024, the major record labels sued Suno and Udio over this, and the companies responded by saying they were protected by the legal principle of \u201cfair use,\u201d which allows unlicensed use of copyrighted material in limited circumstances. <\/p>\n
Google, after being sued by a group of independent musicians over Lyria 3, offered another defense: that by uploading their music to YouTube and agreeing to the site\u2019s terms of service, they gave Google that permission. That case is pending. <\/p>\n
Does anyone really want to listen to AI music? <\/p>\n
That may be the biggest question of all.<\/p>\n
For most people, the connection to a real, human artist is perhaps the most vital aspect of fandom. We love not only the song, but an artist\u2019s story \u2014 their background, their loves, their ups and downs in life. Can computer-generated music ever replace that? <\/p>\n
So far, the Taylor Swifts and Beyonc\u00e9s of the world seem safe. The few AI \u201cartists\u201d that have made waves, in genres like country and R&B, have been curiosities. But time will tell, and the technology is evolving quickly.<\/p>\n