"Create a bright country song to celebrate my grandfather's birthday."
By inputting a brief command into a music-generating AI, you can complete a song with lyrics, melody, vocals, and accompaniment in a short time, even without professional composition knowledge. You can also extend the song, change the vocal or instrumental arrangement using your own lyrics, recorded phrases, or a hum as material.
Music production has traditionally required many skills and expenses, such as playing instruments, composition theory, recording equipment, studios, and engineering. Generative AI has significantly broadened the entry point. Video creators can make background music that matches their visuals, small businesses can prepare jingles for advertisements, and even those without musical experience can turn their words into songs. For professional composers, it can serve as a tool to speed up idea generation, demo production, and arrangement proposals. (Source 2)
However, the moment technology answered the question of "can it be made," more challenging questions emerged.
Is the song truly original? If it resembles existing recordings or a singer's voice, whose permission is needed? Can users protect works largely created by AI as their own intellectual property? And in an environment where songs can be generated almost infinitely, will people want to listen to that music?
"In the style of Madonna" is not the same as "using Madonna's song"
In discussions about AI music, the differences between "style," "song," "recording," and "voice" are often confused.
For example, instructing "1980s dance-pop style," "dramatic female vocals," or "glittering synthesizers" is not legally or ethically the same as remixing Madonna's actual audio.
General genres or atmospheres cannot be monopolized by any one person. On the other hand, existing song melodies, lyrics, and actual recordings have rights holders. If you input commercially available audio into an AI service without permission and publish or sell the processed work, you may infringe on the rights related to the original song or recording.
Furthermore, when synthesizing a voice that closely resembles a real person, issues of personal rights such as name, likeness, and voice, as well as impersonation and false representation, arise in addition to copyright. Therefore, it is not necessarily safe just because you are not copying the song. If you publish or promote it in a way that misleads people into thinking the person is singing, different legal risks may arise.
The basics of enhancing safety are clear.
Do not input existing recordings, lyrics, or melodies for which you do not hold rights. Do not publish works that could be mistaken for the actual artist's voice. Instead of relying solely on specific singer names, generalize instructions by specifying the era, tempo, instruments, musical style, and vocal characteristics. Before commercial use, check the latest terms of the service used and the rules of the publication destination.
Does a song created by AI have copyright?
According to the U.S. Copyright Office, the output of generative AI does not automatically become a copyrighted work. What matters is whether a human sufficiently determined the expressive elements.
Simply providing instructions in text may not be enough to demonstrate human authorship. However, parts where a human wrote the lyrics, recorded performances, creatively selected and arranged multiple generated results, and edited details to complete the work may be protected. (Source 3)
It is important to note that "commercial use on the service" and "fully obtaining legal copyright" are different.
Even if a paid plan of an AI service grants users commercial usage rights, it is a contractual permission between the service and the user. Whether you can claim exclusive rights if a third party creates a similar song depends on how much human creative involvement is in the work.
Therefore, when using AI music for work, keeping production records becomes important. By documenting your own lyrics, input materials, generation dates, unused ideas, editing history, additional performances or recordings, and mixing work, it becomes easier to explain where human creativity is involved.
Using AI as a "vending machine that produces finished products in one go" is less likely to be strong in terms of creativity and rights than using it as a "collaborative production tool that provides materials."
Lawsuits over training data are shifting from conflict to the era of contracts
The biggest issue with music-generating AI is what the model learned.
In 2024, major record companies sued Suno and Udio, claiming they used a large number of copyrighted recordings without permission for training. AI companies have argued that learning is transformative use to create new expressions, but what constitutes lawful learning and what constitutes rights infringement is still being determined in both courts and contract negotiations. (Source 4)
Subsequently, the industry moved from simple all-out confrontation to partnerships based on licensing.
Warner Music Group resolved its lawsuit with Suno and announced collaboration on next-generation services using licensed music. Universal Music Group and Warner also settled and partnered with Udio, introducing rights-cleared models and new revenue opportunities. (Sources 5, 6)
However, this does not mean the problem is solved.
In July 2026, Sony Music reportedly filed a new lawsuit against Udio, claiming that more than 30,000 songs managed by the company were used for training. Additionally, musicians' unions are contesting that record companies' contracts with AI companies do not provide sufficient consent or compensation for performers. (Sources 7, 8)
This reflects the complexity of the AI music era.
Even if a recording licensed by a record company is used, it involves multiple contributions from singers, performers, composers, lyricists, and producers. A "contract with rights holders" does not necessarily mean a distribution that satisfies all creators.
Future competition will be determined not only by sound quality but also by the reliability of the system design regarding who gets permission and how rewards are distributed.
The issue will be not a lack of songs, but a lack of time to listen
Generative AI makes the supply of music virtually unlimited. What is happening is the so-called "AI slop" problem.
There is concern that similar songs generated at low cost will flood streaming services, overwhelming search results, recommendations, and royalty distribution.
In July 2026, Deezer announced that songs determined to be fully AI-generated exceeded half of the new daily posts at their peak, reaching about 90,000 songs per day on average. Meanwhile, the percentage of fully AI-generated songs in actual playback remained at about 1-3%, with up to 85% of AI song playbacks in 2025 being identified as fraudulent streaming. (Source 9)
These numbers indicate that AI music has not immediately displaced popular human songs. Rather, there is a significant gap between the volume of posts and demand.
Another study also pointed out that the overwhelming majority of AI music in the survey was rarely played, with a tendency to post large volumes aiming for a few hits. (Source 10)
The value of music is not determined solely by the difficulty of making it.
Who created the song and why? What experiences are behind the lyrics? How is it performed live, and what memories do fans share? The reasons people become attached to music include stories beyond the sound itself.
Even if AI can produce a large number of high-quality sound sources, it cannot automatically create a "reason to choose" that one song.
On social media, "democratization," "transparency," and "dilution of value" collide
The New York Times article is newly published, and the Reddit post that automatically reposted the article link had no notable comments at the time of confirmation. Therefore, it is not possible to speak of public opinion based solely on direct reactions to the article itself. (Source 11)
However, several positions have already emerged in social media communities dealing with AI music.
Voices welcoming the "democratization of music production"
The first is the opinion welcoming the ability to shape the music in one's head even without performance experience.
People with disabilities or time constraints can also participate in production, and they can add dedicated music to personal videos, games, podcasts, and advertisements. Among supporters, some argue that if you repeatedly generate and edit rather than just ending with a prompt, human intention and passion exist there. (Source 12)
In traditional music production, performance, composition, recording, editing, and mixing have been divided among multiple people and machines. If AI is considered a new production tool, the use of tools alone should not be a reason to deny the value of a work.
Voices saying "Don't hide AI usage"
The second is the opinion emphasizing transparency.
On social media, there are repeated reactions of feeling deceived after being moved by something only to find out it was AI-generated. On the other hand, there is concern that if it is labeled as AI from the start, it might be rated lower before listening to the content of the song. (Source 13)
A survey by Deezer and Ipsos found that 97% of respondents could not distinguish between fully AI-generated songs and human-produced songs in a blind test, and 80% said AI music should be clearly labeled. (Source 14)
As it becomes harder to distinguish by listening, more people feel the need for labeling, which is a seemingly paradoxical situation.
In the future, it might be more practical to indicate not just "AI song" but also which processes used AI, such as lyrics, composition, vocals, performance, and mixing. In modern music production, there are countless intermediate forms between human production and fully AI generation.
Concerns about "dilution of the value of effort and work"
The third is the concern about the dilution of the value of music and work.
Creators who have honed their skills over many years express feelings of futility as large volumes of sound sources are created with a few seconds of instructions and placed on the same distribution shelf. There is also strong criticism that it is not just a "new instrument" but a system that absorbs past music without permission while dragging human work into price competition.
Especially for BGM for videos, advertising music, in-game music, and demo vocals, which have been the domain of individual composers and small studios, they are more susceptible to price declines due to generative AI. If clients can create "usable enough songs" in minutes, traditional creators will be required to provide value beyond just delivering sound sources.
However, social media posts are not public opinion polls.
In communities where AI music users gather, positive opinions are more common, while in musician communities, concerns about rights and employment are stronger. What matters is not simply comparing the number of pros and cons, but distinguishing what they are for and against.
Many people are not opposed to the technology of AI itself but to unauthorized learning, impersonating voices, lack of labeling, fraudulent playback, and opaque profit distribution.
Will AI music end human music?
In the past, every time synthesizers, sampling, drum machines, Auto-Tune, and streaming emerged, there were debates about "the loss of real music."
New technologies have changed some jobs while creating new genres, professions, and ways of listening.
Generative AI may follow the same path. However, the significant differences from past technologies are the speed and scale of production and the scope of learning from existing works.
In an environment where one person can create hundreds of songs a day and distribute them worldwide, not only the "ability to create good songs" but also a "system to select reliable works from a large volume of generated content" becomes indispensable.
What will be needed in the future is not a binary choice between total prohibition and unlimited use.
It involves consent and compensation for learning data, permission when using real people's voices and names, common labeling to convey the extent of AI use, a system to record human creative parts, elimination of fraudulent mass postings and playback manipulation, and transparent contracts to distribute revenue to stakeholders.
AI has lowered the cost of making music. However, it has not lowered the cost of building trust.
In an era where anyone can generate songs, what will become rare is not the sound source itself, but the relationship that makes people want to listen to "this person's music."
The future of music will not be determined solely by whether AI can sing better than humans. It will depend on how humans use AI with permission, compensation, and transparency, and how they imprint their own experiences and choices into it.
Whether generative AI becomes the next creative revolution or a device that produces countless noise depends more on the rules and culture we establish than on the technology's performance.
Source URLs
1. The New York Times
An article addressing music production by AI, remixing existing artists, copyright, and listener demand in a FAQ format.
https://www.nytimes.com/2026/07/24/arts/music/ai-music-faq.html
2. Suno Official Site
Explanation of features and usage forms that can generate songs including lyrics, vocals, and accompaniment from text instructions.
https://suno.com/
3. U.S. Copyright Office "Copyright and Artificial Intelligence"
Clarification that sufficient human creative involvement is necessary for copyright protection of generative AI output, and that prompt input alone may not be sufficient.
https://www.copyright.gov/ai/
4. RIAA and Reuters 2024 Lawsuit Materials
Background on major record companies suing Suno and Udio for copyright infringement due to unauthorized learning.
https://www.riaa.com/record-companies-bring-landmark-cases-for-responsible-ai-againstsuno-and-udio-in-boston-and-new-york-federal-courts-respectively/
https://www.reuters.com/technology/artificial-intelligence/music-labels-sue-ai-companies-suno-udio-us-copyright-infringement-2024-06-24/