Skip to content
TrackPodcasts
societySep 15, 202621:05

Where do we draw the AI music line?

About this episode

It sounded like a hit, it charted like a hit, and radio played it like a hit. There was just one catch. No human made it. Now Australia has become the first country to shut fully AI-generated music out of its official charts and its biggest awards. But if a machine makes a song that makes you feel something, does it matter who or what made it? On the show: Steve, Fei Fei & Yushun

Get every episode summarized

Each time Round Table China publishes, we email you a written briefing from the transcript — the topics, who appeared, and any specific claims, with the ad reads skipped.

Email me new episodes

Free for 3 shows. No card needed.

Hosts & guests

Transcript ready

282 searchable segments. Every word is indexed and playable.

Where do we draw the AI music line?

Round Table China

0:00
21:05

Full transcript

Round Table ChinaWhere do we draw the AI music line?. Machine-transcribed; use the interactive transcript above to jump the player to any line.

Discussion keeps the world turning. This is Roundtable. You're listening to Roundtable, I'm Steve Hatharly today with Feifei and Yousian. Coming up, Australia just fired the first shot in the fight over AI and music. Fully AI-generated recordings are now banned from the country's official charts and its biggest awards. And this is the interesting part, AI-assisted tracks can still qualify, provided a human is genuinely behind the creative work. Okay, but there's a problem. No one can agree on what that actually means. Some artists are embracing AI as a tool to push their sound further, but others see it as an existential threat to their craft and their income. Australia has drawn the line in the sand and the rest of the world is watching to see if it holds. You can find us at Roundtable China on Apple Podcasts and don't forget that we love to

hear from you. You can send an email to us, you can send a voice note to us as well. Roundtable podcast at qq.com one more time, Roundtable podcast at qq.com. Australia has just drawn a line in the sand over artificial intelligence and music. The Australian Recording Industry Association, the ARIA, has changed its rules so that fully AI-generated recordings can no longer compete on Australia's official charts or for their awards. Music that uses AI as a tool, though, can still qualify, but only if it's substantially human-made, lawful, and not involved in chart manipulation. The decision follows a controversy over an AI-assisted cover of a well-known pop-ballad that became a major radio and streaming hit in Australia. But Australia is really just one piece of a much bigger story. AI-generated music is already appearing on streaming services and climbing charts and impersonating

real artists, and in some cases it's attracting millions of streams. Meanwhile, some musicians are using AI as another creative tool, but others say it threatens their livelihoods, their careers, and the value of human creativity. It's a fun question to talk about. Where should we actually draw this line? If a human writes the song, but AI sings it, is that really a human song? And if listeners genuinely enjoy an AI-generated track, well then should it even matter at all? We're going to talk about this in detail. What exactly will start here? What exactly has Australia decided with their new rules? Yeah, on the surface, it feels like AI is totally banned from this music chart, but it's not like everything related to AI is banned on this chart. So fully AI-generated music is now excluded from this ARIA charts or ARA. The rules took, in fact, August 28th in this year.

So first of all, AI-generated tracks, and where AI created all or most of the primary creative elements are banned from the charts and awards. But aside from that, AI-assisted tracks where human creative input is the primary driver being eligible. And also AI-assisted tracks that the recording also has to be lawful and free of stream, or chart manipulation concerns. And also ARIA can strip chart positions, revoke awards, or pull recreation retroactively if a track is later found in ARIA. Yeah, and all of this started because of one particular song. Yeah, and there's a well-known 1980s pop ballad that was built. The original track was published in the 1980s, and somehow an Australian producer sort of twisted that ballad with AI-generated vocals and drums and become almost like electronic

piece of music. And that piece became the most played track on Australian radio and reached the top five on the two of the ARIA charts. So people are loving it, but apparently I think it posed, I won't say the rats, but very at least a question for a lot of artists. There were a few issues here. So this song became, is like a remix version of a track called Like a Prayer from a really popular artist from the 1980s. This guy remixed it. He's the performer and the producer. He had his uncle on the synth operator for the track. But when the track was published, there was no mention of generative AI whatsoever. And also when it was published, there was no human vocalist that was credited for the prominent lead singing in the song. And then it wasn't until media and a local media organization pushed for answers that

the person, the artist who made this song finally added generative AI to the credits for the vocals and for the drums. So transparency was one of the major issues here. And it really angered a lot of people because they felt like they were misled. This track was huge in Australia this summer. It spent four weeks as the number one most played song on Australian commercial radio. And that's why too, this wasn't just some song buried in the charts somewhere. This was like the hottest thing to listen to in the whole country. Yeah, that actually reminded me. Yes, transparency is I think one of the triggers that made people discuss it because AI music has been there for a while. And a lot of even worldwide AI music has been in these music banks for a lot of, for a really, really long time and actually a huge amount of music. Now we were listening to maybe they're all generated by AI.

A separate industry analysis found that roughly 40% of new music released worldwide in July 2026 involved AI in some form and about 23% fully AI generated. And another 15 AI 15% AI generated then reworked by a human. Yeah, so that means that you know, you could argue AI is actually becoming popular with listeners. There was and the supply of AI music is absolutely exploding. They've measured it since early 2025 and the number of AI tracks per day has increased and mid in mid 2026 where we are now to 90,000 a day, which is a 50% increase of where we were in January of 2025. And also a lot of the songs are really, really getting popular among listeners. I mean, Australia, that Australian track is one example.

But when we're thinking about either on TikTok, on doing all other social media platforms, we can often hear such remix of a old song with AI technology. And then somehow got really, really popular among people online or even some songs that are purely AI generated that people are loving. I think one part is a lot of listeners can't really tell between the difference of a human remix of an old song versus an AI remix of an old song. And also when it comes to purely fully AI generated song. That I don't, I don't think so because you can still, if you like, listen to music, quite a lot. There are some specific element that you can tell that this music is basically an AI generated one. Whenever I get into like my parents car and they turn on their music app, it feels like half of the music are generated by AI.

What makes you, how can you tell? You can almost tell right away those, you know, first of all, impossibly high vocals and also powerful, yet slightly raspy tone. And then a style that sounds familiar, but somehow different from the one that you listened before. And then with all of these facts, then you can kind of make sure that this song is generated by AI. Well, I mean, there's an underlying question that arises from this whole discussion because if we're talking about music charts specifically and not just AI generated music for our own personal enjoyment, the question about what should be on a chart and what shouldn't be on a chart is very hotly debated because if a chart is supposed to measure the popularity of music, then we should say AI music should be allowed.

But if the chart is supposed to measure the popularity of music made by artists, then AI music should not be allowed. But then the question goes deeper, okay, well, AI assisted is okay. But when it's mostly AI, then that's not okay. So then there's this whole huge gray area that exists and that's why everybody's fighting about this these days. Yeah, I think I can understand on some level like if, for example, a producer or a soundwriter use 10% of AI generated technology in his whole production versus somebody using 90% of the production is AI assisted, then they are competing equally together on the same chart. Can we say it's a fair play is 90% equally important as 10%. What about that soundwriter that decided doesn't use AI at all.

So here's what this guy did in Australia. He went on social media and he made a video and he posted this behind the scenes shot of him showing his workflow. And in the clip, he recorded himself singing the lyrics very poorly, by the way, he sung very poorly, I guess on purpose, into a microphone and then it went into an AI platform. And the AI instantly, of course, spits out this perfect, highly polished R&B style vocal with an American accent. Right. So instead of calming, he was hoping to calm down the critics because he said, I'm the one who sang into the microphone. About a lot of people are saying, no, that's not you who is singing even in the song. In fact, he was just, it seemed like he just poured gasoline onto the fire because musicians pointed out it wasn't the standard pitch correction like an auto tune would do.

The AI completely invented a brand new voice. And the critics said to him, this, you're not helping yourself. You're only proving our point that this is a core problem in the industry is that this is almost entirely automated. I guess my question would be, why doesn't matter? I mean, if one person who can't say now with the help of AI can produce his own or her own track, compared with somebody who learned music, for example, for 10 years and is also producing his own music, if they are competing on the same market, why doesn't matter? Well, isn't that the listener's choice is more important? Well, see, so when it's music charts, right? And you, you're at the top of the charts, then you get paid and you get paid per listen. So if it's a, I generated, then the money is going to the machine instead of other artists.

And this is an Australian example, but this, you could have this conversation in many other parts of the world, many artists who poured their heart and their soul and their time into actually making the music themselves into making their art themselves. And this guy put a, a prompt into a computer and boom, instant hit, instant paycheck. And that's why on August 20th, August 25th, the ARIA officially rewrote its rules completely banning wholly or largely AI generated songs. But I agree with what you're saying, Fei-Fei, because we've, it seems like we've had this conversation on round table before when we were talking about authors who were using AI to assist their work, to in their defense, assist in their writing, not do the writing for them. Yes, I think a lot of writers are, well, on the fence about whether they should be,

first of all, using AI to assist their work in the first place. Some are using for purely just research of a certain story. Some are probably using more of such technology. And then they're also critics saying that if, for example, that author really used AI to assist with that, with that piece of work, and they put a little badge on the book cover saying this novel or this book is AI assisted. Would the readers still want to pay by the book? And I remember answering that question, and my answer was no. Yeah, publishing has its own version of this. Shorthor videos has its own versions too, right? And what we kind of see now on these kind of Shorthorne video platforms, I think also a huge amount of even, if it's not more than 50% of the videos that we scroll on their phone is AI generated.

And I think I've, I have told this example before I got into the subway and then whenever I stare at somebody's phone, they're looking at a AI generated video. It's just, you know, that's, I think also reflected and all of these incidents, I think they are reflecting a fact that for me, at least, it makes me feel like our lives and the content that we consume are increasingly being flooded with these kind of AI generated content. And what is the problem is they are repetitive and low quality content that you are kind of just chewing the same thing over and over again. And these kind of Shorthorne dramas give you plus that you can predict from the start and these AI music give you that kind of vague sense of familiarity that you can get from other music. And this Ben is, I think, also reflecting on attitude of, okay, we just don't want that

gray area anymore. We just want to ban something that is so obvious. Do you think it's a quality issue too, though? Yes, of course. I'm a little bit of, or kind of sick of looking at all of these, you know, same plot and same content over and over again. Yeah, I don't think, yeah, we have different issues here because we have the book is instance, we have the video instance and then we have the music instance. Now if you go and listen to this song, I don't know if you've had a chance to hear this particular song from Australia or not. But I think in dance music, it's really hard to tell whether something is created by AI or not. Because in dance music, there's a lot of effects on vocals anyway. They did a study or some research and they wanted to know if people could tell the difference or not.

This was done by D'surf. A research 9,000 different people in eight different countries and 97% couldn't distinguish between AI generated music and human made music. 97, that's almost everybody. They couldn't tell the difference. I think when you're using your eyes, it's easier to tell. But I think when you're using your ears, it's maybe even harder and that's why such extreme measures were deemed necessary. For me, I think it's a little overreaction till this new technology. Because I listened to that song, that Australian song. It's pretty good, right? Yeah, it's pretty good. But I also think it's just like every other normal dance music on the market that each could climb on the top five of a certain chart. But it won't stick when we're thinking about the whole of music industry and even the history of pop music. So I mean, why bother?

What's the big deal? So when we are having a lot of music artists who are already using more than technology, like Autune, not only Autotune, but... Yeah, Autotune. That's back what? Late 1990s, early 2000s. But a lot of artists are already using the same keyboard to produce a whole track with different instruments. They're using synthetic drums, synthetic guitar sounds. Then what's the difference between AI-generated versus artists using synthetic machines to produce music? First of all, Autotune is different from how you just sing really badly and turn it into a perfect toned music. The second of all, actually now people are also complaining about singers when they're doing live using Autotune because if a fan just go to that specific place to listen to a performance, they want to hear some real stuff.

Has that been an issue in any Chinese performances I wonder where people have seen a lot of news? People complain about that. Yeah, audiences complain about these kind of specifically auto-tuned performance and saying that it is quote unquote, unreal. So that's one thing. And another is, yes, the transparency maybe is the focus in this kind of topic. And another thing is when we're talking about music, it is really very, very subjective, right? When we're just discussing how good it is. And I remember one another, really popular, I'm not sure it is AI-generated or just very robotic one. It's called K-pop Demon Hunter. It's really popular anime and song that is... Yeah, that music was incredibly popular. Basically in every music chart. And now there are real singers singing that song. So that's why I'm thinking maybe, of course, the music can be popular, but maybe people

just like real people singing. So I think the real question here is really on some level about copyright. For example, about that Australian song that we discussed in the show, did that producer pay the original singer and producer the amount of money he should be paying because it's basically using their production and doing a remake on that. And I think in China at least, this is where the main focus is about. But when it comes to the transparency, is it okay to use AI assisted to generate your music? I think a lot of artists are really having fun with it. We're seeing fully AI-generated songs getting popular online. And then a real human singer starts to sing that song. But another popularity boom. So I think just having fun with it is a new technology. I remember a performance on a TV show in America where there was a rapper and he was supposed... He always used auto tune in his songs and in this performance that it wasn't working. So his real singing voice came out.

Oh, that was a big story in the news the next day. That's for sure. Yeah, this is very debatable. Interesting to talk about. Alright, that'll do it for Round Table 4 this episode. We had fun today. We hope you did as well. I'm Steve for Feifei and Youshan. Have a good one and join us next time.

More episodes

More from Round Table China

View all episodes →