AI Music, Copyright and APRA AMCOS: Your Legal Questions Answered
AI can help a songwriter turn an idea into a finished track, but it cannot answer the legal questions that follow. In Episode 29 of Legal Matters, Andrew Bell and Colin Balewski explore the human contribution behind AI-assisted music, copyright and moral rights, APRA AMCOS and ARIA issues, voice-cloning risks, and the practical records creators should keep.
This podcast accompanies Bell & Senior Lawyers’ detailed Australian Technology Law guide for songwriters, producers, performers and community broadcasters using AI tools. Begin with the one-page practical guide, then use the detailed guide when you need a fuller explanation of authorship, royalties, industry rules and release risks.
Key Topics
- The distinction between conventional digital production, machine learning and generative AI.
- Why a song can involve separate rights in lyrics, musical composition, recordings and performances.
- Human authorship and creative contribution in Australian copyright law.
- What IceTV can and cannot tell creators about computer-generated outputs.
- The role of APRA AMCOS, industry eligibility rules and the uncertainty around a “substantial portion” generated by AI.
- The continuing relevance of copying disputes, including Kookaburra Sits in the Old Gum Tree and Down Under.
- Why cloning a recognisable voice creates a higher-risk issue than ordinary AI-assisted production.
- Evidence creators should keep from the first lyric draft through to release.
Listener FAQ Highlighted in This Episode
If I wrote the lyrics but AI created the music, do I own the song?
The answer may differ for the lyrics, composition, arrangement, sound recording and performance. The strongest position generally comes from being able to identify and prove the human creative decisions behind each element. If another person used the tool, or if collaborators contributed different parts, record those contributions and agree on ownership before release.
Can community radio play AI-assisted music?
The legal and practical answer is not a simple yes or no. A station must consider its licences, policies and the rights in the material it receives. A creator should not represent a track as wholly human-made if that would be misleading, and should give accurate information about writers, performers, samples, rights and AI assistance when asked.
What records should I keep?
Keep the first lyric drafts, musical sketches, project sessions, prompt and revision history, exports, communications with collaborators, licence terms, performer permissions and dated version files. A sealed, dated copy may add a small piece of timing evidence, but it does not create copyright and should not be the only record.
Key Takeaways
- AI is a spectrum of tools. A reverb effect, virtual instrument, remix workflow and generative music service may raise different factual and legal questions.
- A song is not a single right. The lyrics, music, recording, performance and visual assets can have separate creators, owners and licences.
- Human creative input remains central to the copyright analysis. The quality and evidence of that input matter more than a simple label of “AI” or “not AI”.
- Industry eligibility, chart rules, collecting-society requirements and copyright protection are related but distinct questions.
- Do not use a recognisable singer’s or public figure’s voice without checking permissions and applicable laws.
- Keep contemporaneous records from the creation process, and agree on co-writing, credits and ownership before a song earns money or causes a dispute.
Annotated Transcript
AI music is a live question for creators and community radio
⚖️ Topic: AI-assisted music, authorship and industry practice
Relevant law: Copyright Act 1968 (Cth)
Andrew: Andrew introduced the topic as a practical question for musicians, songwriters and community radio: if an AI tool helps write or produce a song, who owns it, can it be played, and how should it be treated by APRA AMCOS and the industry?
Colin: Colin said that broadcasters are increasingly receiving tracks made with AI assistance. The first practical question is often the extent of AI use, but the industry has not reached a single consensus on what audiences will accept.
Andrew: The discussion distinguished older statistical or machine-learning tools from current generative tools. Andrew and Colin compared the new technology with familiar production techniques, including reverb, virtual instruments, synthesisers, mixing and remixes. The relevant issue is not simply whether a computer was involved, but the creative choices made by people using the tools.
📎 See also: AI-Written Music, Copyright and APRA AMCOS: A Practical Guide for Australian Songwriters | One Page Practical Guide PDF
A song is made of several separate rights
⚖️ Topic: Lyrics, musical works, sound recordings and performances
Relevant law: Copyright Act 1968 (Cth)
Andrew: Andrew explained that copyright questions about a song need to be separated. The lyrics, musical score or composition, performance and sound recording may be distinct works. Different people may own or control different rights, particularly where one person writes words, another arranges music and another records or performs the result.
Human creative contribution and computer-generated output
⚖️ Topic: Authorship, originality and evidence of creative choices
Relevant law: Copyright Act 1968 (Cth) ; IceTV Pty Ltd v Nine Network Australia Pty Ltd
Andrew: Andrew referred to IceTV as an example of the distinction between information compiled by a computer and human creative expression. The episode did not treat that case as a final answer for AI music. Rather, it illustrated why a purely automated output may raise different questions from work shaped through selection, editing, remixing, arrangement and other human decisions.
Andrew and Colin: The camera analogy was used to explain the point. Photography once prompted debate about whether the camera simply recorded reality. Today, the photographer’s framing, timing, selection and editing are widely recognised as potentially creative. AI-assisted music may similarly require a close factual assessment of the human contribution to the finished work.
Sampling, substantial parts and the Kookaburra dispute
⚖️ Topic: Copyright infringement and taking a substantial part
Relevant law: Copyright Act 1968 (Cth) ; Larrikin Music Publishing Pty Ltd v EMI Songs Australia Pty Ltd
Colin: Colin raised the familiar dispute over Kookaburra Sits in the Old Gum Tree and Men at Work’s Down Under.
Andrew: Andrew explained that copyright disputes often turn on whether a substantial part of a protected work has been reproduced. The discussion is relevant to AI because a creator, broadcaster or distributor may face a claim if a generated output reproduces protected material. The question is not confined to whether a whole song is copied.
Annie’s favourite AI-assisted dance track
⚖️ Topic: Disclosure, audience expectations and AI-assisted release
Relevant law: Copyright Act 1968 (Cth)
Annie: Annie described Just Friends by Magic Brown featuring Manyus and Dario Guida as a favourite track from her dance class. The discussion turned to the fact that the music was produced using AI tools and whether that prevented a song from being released or played.
Andrew: Andrew’s practical advice was that a creator should be up-front about the extent of AI use. If a person wrote the tune or lyrics but used AI for an arrangement, that should be understood in the same spirit as other production credits. He also noted that questions about training data, performer rights and commercial use are developing internationally.
🎵 Listen to Annie’s favourite track: Play “Just Friends” by Magic Brown (feat. Manyus & Dario Guida) on YouTube (if video embedding is restricted by your browser).
📎 See also: AI-Written Music, Copyright and APRA AMCOS: A Practical Guide for Australian Songwriters | One Page Practical Guide PDF
APRA AMCOS, ARIA and the meaning of “substantial portion”
⚖️ Topic: Collecting societies, chart eligibility and AI disclosure
Relevant law: Copyright and industry rules are separate questions
Andrew: Andrew said that APRA manages performance rights and AMCOS deals with other music rights, while noting that recent industry developments had raised questions about the eligibility of music where a substantial portion was generated by AI. The programme observed that the practical meaning of a substantial AI contribution is an area that needs clearer guidance.
Editorial note: Copyright protection, APRA AMCOS registration and royalty administration, chart or award eligibility, and platform acceptance are related but not identical. A creator should not treat an industry rule as a definitive answer to every copyright question.
Voice cloning and preserving evidence
⚖️ Topic: Voice-model permissions and contemporaneous creation records
Relevant law: Copyright Act 1968 (Cth) and potentially applicable contractual, consumer and overseas laws
Andrew: Andrew identified cloned voices as an area of particular concern. A recognisable voice, whether it belongs to a well-known performer or another person, should not be used on the assumption that it is legally available. The relevant risks can extend beyond Australian copyright law, especially if music is distributed internationally.
Andrew and Colin: The discussion then turned to evidence. Creators should save drafts, notes, session files and material showing what was created and when. Colin suggested the long-standing practice of signing, dating and posting a sealed copy to oneself. Andrew agreed that records can assist in establishing a timeline, while emphasising the value of preserving the underlying creative process.
📎 See also: AI-Written Music, Copyright and APRA AMCOS: A Practical Guide for Australian Songwriters | Download the One Page Practical Guide
This page provides general legal information only; it is not personal legal advice. Everyone’s situation is different, so please seek independent legal advice for your own circumstances.
Related Episodes
- AI Law
- Technology Law Part 2: Consumer Law and AI
- Technology Law Part 1: Business and Cyber Security
- Drones, Rental Cars and Parenting Relocation: Your Legal Questions Answered
AI music creator resources
Start with the One Page Practical Guide
Before you release, submit, pitch, broadcast or monetise an AI-assisted track, download the AI Music Copyright and APRA AMCOS One Page Practical Guide .
Read or download the detailed guide
- Read online: AI-Written Music, Copyright and APRA AMCOS: A Practical Guide for Australian Songwriters
- Download: Full AI Music Copyright and APRA AMCOS Guide PDF
These resources provide general legal information only. They are not personal legal advice. Seek advice for your particular song, release, licence, dispute or commercial arrangement.
Full Broadcast Transcript
The following is the full 4CRB broadcast transcript for 1 September 2026. It has been lightly edited for punctuation and readability. It is provided as a record of the programme and remains general legal information only, not personal legal advice.
Read the complete transcript of Legal Matters, 1 September 2026
00:00 - Programme introduction
Narrator: 4CRB now presents Legal Matters, proudly brought to you by Bell and Senior Lawyers. Call them today for all your legal needs on 07 5532 8777. This program provides general legal information only. It is not personal legal advice. Everyone’s situation is different, so please seek independent advice for your own circumstances.
00:25 - Introducing AI and music
Colin Balewski: As the promo states, it is a very good morning to Andrew Bell from Bell and Senior Lawyers.
Andrew Bell: Thank you for having me back, Colin.
Colin: We have been discussing this topic heavily off air. It is very interesting. I will throw to you, Andrew, and then add some colour.
Andrew: This week I thought we would do something different because we are on a community radio station and play music. Musicians, songwriters and listeners keep asking: if you use an AI tool to write or produce a song, who owns it? If I am a songwriter, what does APRA think? Can it be played on radio? Should there be limits on how much AI is used in music, and where it is used?
Colin: Each week I receive emails from people saying, “Here is my new song” or “Here is a song I generated”. Our first question is how much of it is AI. The industry does not have a consensus. We have asked listeners what they think about AI-generated music and whether it is acceptable to them.
Andrew: There is no current Australian case that decides copyright or AI use in a song. I am writing about AI law and how rapidly it is developing. We can look at copyright’s historical treatment of creative expression to see where it might land legally. There is also the separate question of whether people want to listen to something generated by a machine.
02:37 - AI as a creative tool
Colin: This is your opportunity to ask Andrew Bell a legal question. Ian, our AI contributor, has AI-generated content. He writes all the lyrics himself but could not find anyone to voice the lyrics or put music to them, so he used AI to generate music. He may have country, rock or jazz versions. He has completed something he is passionate about by using a tool.
Andrew: What counts as AI is itself open to debate. AI previously covered machine learning, using statistics to develop models. We are now in the era of generative models, generative AI and large language models. ChatGPT is a generator that looks at training material and attempts to predict a response.
If a computer makes something clever, people increasingly call it AI. But much of that is known mathematics with a standard output. Digital music production has expanded greatly over the last 20 years, and much contemporary music has digital filters or virtual instruments associated with it.
Colin: Even an early reverb pedal added a sound that was not generated by the voice or instrument itself. Is that AI, or is it a tool doing something useful?
Andrew: It is similar to a Marshall amplifier being pushed into distortion, even though it was not designed for that use. Is this simply another generation reacting to a new kind of music, or does AI make something fundamentally different?
06:05 - The legal question: who owns an AI-assisted song?
Colin: If someone writes lyrics and plugs them into an AI system that generates music, and a vocalist voices those words, where do we stand?
Andrew: We do not have a settled answer. Copyright in a song is not one right. There can be copyright in lyrics, in the score or composition, in a live performance, and in the audio mix released to radio or retail. Different people could own one or all of those rights.
Copyright requires creative expression and a human author. Andrew referred to IceTV, in which the High Court considered compiled television-program information. The point drawn from the case was that plain information compiled by a computer did not itself attract copyright in the same way as a human creative work.
If a person only presses run and a track is generated without input, Andrew’s view was that there may be a weak case for that person owning copyright in the lyrics, music or performance. The facts change where a person writes lyrics, selects outputs, edits, layers, changes arrangements and directs the creative process.
09:27 - Cameras, mixing and human creative choices
Andrew: Modern composition often involves multiple tracks, multiple takes, splicing, mixing and editing. People use AI for inspiration, generate drums or a voice track, cut material out, change lyrics, alter the hi-hat and reshape the song into something they want to hear. That is creative input.
The camera is a useful historical analogy. Photography once prompted debate about whether it could attract copyright because the camera recorded reality. Over time, it became clear that two people could use the same camera on the same subject and create different works. Framing, timing and post-processing matter.
Colin: Synthesizers and remix artists also involve creativity even where machines do much of the technical work. The operator’s ear, thought and consideration shape the finished product.
Andrew: People can take the same starting song and produce completely different remixes based on the choices they make. Modern music often involves EQ, reverb, effects, re-recording and layering. The legal analysis may focus on the creative human expression rather than treating every use of a tool as disqualifying.
15:02 - Copyright disputes and substantial parts
Colin: Copyright often comes back to whether someone will sue because a product was generated by someone else, or because a protected work was breached. A famous example is Kookaburra Sits in the Old Gum Tree and Men at Work’s song.
Andrew: The song was Down Under. In 2010, the courts considered whether a substantial part of Kookaburra Sits in the Old Gum Tree was replicated in Down Under. To recover damages for copyright infringement, a substantial part of the work needs to be reproduced. A simple song can still be substantially taken if a recognisable part is used in another arrangement.
If a person uploads a track to YouTube or it is played on community radio, and it infringes another person’s copyright, the copyright owner may seek compensation or ask for the track to be removed. Larrikin Music owned the rights to Kookaburra, and that is the case people can search for.
17:41 - Annie and Just Friends
Colin: We have our first caller. Good morning, Annie. Ask your question of Andrew.
Annie: I am finding the program interesting, and it is lovely that a legal program can be musical. I like AI-generated music because I used to go to a dance class and the teacher played a song that was actually an AI one. I loved it. It was called Just Friends by Magic Brown, featuring Manyus and Dario Guida. It is my favourite one to dance to.
Andrew: Music historically is about having fun.
Annie: I thought I saw headlines saying that Australia was the first to ban AI music on radio.
Andrew: Andrew said that APRA and AMCOS had announced an industry position concerning prizes where a substantial portion of music was generated by AI. He noted that the meaning of “substantial portion” was legally interesting and that clear practical guidance would be valuable.
Colin: Annie had written material herself, while a friend used AI to generate music around it.
Annie: I did not know whether it was legal to use that or whether it could be made into a record.
Andrew: Andrew’s view was that the work could be published, but that the creator should declare what was generated by AI. If the person wrote the tune and lyrics but AI arranged the music, listeners should understand the extent. Questions about training data and compensation for source artists are developing internationally, but Andrew said that transparency about the generation process was sensible.
Colin: Be loud and proud. You wrote the lyrics and used a tool to generate the rest.
Andrew: Andrew said Bell & Senior would publish guidelines for writers, producers and performers using AI-generated music.
🎵 Listen to the track: Play “Just Friends” by Magic Brown (feat. Manyus & Dario Guida) on YouTube
22:31 - Practical steps for creators
Andrew: Australia has no copyright-registration system like a trade mark or patent register. Creators should record and save a definitive version of their work. Copyright is generally asserted by the person claiming it unless another party challenges that claim, but evidence becomes important if royalties or ownership are disputed.
Creators should document the extent of their involvement and preserve contemporaneous notes. Co-writers should record what each person developed. Someone who wrote lyrics may own significant rights in those lyrics even if another part of the production later raises an AI issue.
Andrew recommended against cloning recognisable voices without permission. Australia may not have the same identity-rights framework as some overseas jurisdictions, but a distributed song can be subject to laws beyond Australia. Anyone using a virtual voice should ensure they have the rights to use it.
Colin: One simple way to preserve evidence is to sign and date a page, have the signature witnessed, place it in an envelope and post it to yourself by registered mail. Keep it sealed unless ownership is challenged.
Andrew: That may assist in showing that material was not public before a particular date and can help establish a timeline. Creators should also preserve recording-session information, notes and production records. The more information available about the recording or generation process, the better if the work ever has to be defended.
26:15 - Close
Colin: Andrew, that was a fantastic discussion. I think the number of calls proves that people were listening. Thank you to Annie, and thank you, Andrew, for clarifying the topic.
Andrew: Thank you, Colin.
Narrator: You have been listening to Legal Matters, proudly brought to you by Bell and Senior Lawyers. Call them today for all your legal needs on 07 5532 8777. A copy of this and past programs and relevant resources will be available to replay or download on bellsenior.com.au and 4crb.com.