Australia’s music industry has introduced one of the most explicit policy positions yet on the role of generative AI in chart music. The Australian Recording Industry Association, known as ARIA, has updated its Charts Code of Practice to state that recordings must be “substantially human made” in order to qualify for the country’s official charts. The revised rules take effect with the chart dated Aug. 31 and published Aug. 28, placing Australia among the first major music markets to draw a clear line between AI-assisted creativity and fully synthetic music.
ARIA does not describe the move as a blanket ban on AI. Artists may continue to use AI-powered tools during recording, production, mixing, mastering, and other technical stages. The central condition is that human creativity must remain responsible for the essential musical elements of a track. Songwriting, lead vocal performances, and primary instrumental parts must come from human artists. Recordings that depend on AI to generate a lead vocal, a key instrumental performance, or an entire song from a text prompt will no longer be eligible for ARIA charts.
The distinction is intended to be practical rather than philosophical. Many musicians already use AI in their workflows, and ARIA says it wants the charts to evolve alongside technology without allowing algorithmically generated songs to replace human expression. The new wording effectively protects the role of artists while making it clear that AI should serve as an aid or source of inspiration, not as the author of the creative work.
Human creativity and music charts
The phrase “substantially human made” places the burden on labels, distributors, and self-releasing acts to determine how much AI involvement is acceptable. Under the new standards, recordings can still be admitted if a human wrote the song and performed the principal vocals or instrumentation, even when AI was used to clean up audio, generate backing textures, or assist with production decisions. The chart rules do not attempt to punish experimentation or the use of modern production software. Instead, they focus on the origin of the primary creative elements.
Excluded examples include AI-generated lead vocals that mimic a human performer, synthetic drum performances that form the rhythmic core of a track, and songs that are created from a simple text prompt. The rules also suggest that an entire recording generated by an AI service, with no meaningful human contribution, falls outside the definition of a “sound recording” for ARIA chart purposes.
ARIA’s announcement stressed that chart eligibility is determined per release, and submissions require declarations about AI use. Every release sent to the ARIA survey must now include a statement on whether generative AI was used. For artists who use AI as a background tool, this should be a minor administrative step. For works that rely on generative platforms to create the core music, the new declaration process is likely to reveal their synthetic origin and lead to exclusion.
A controversial cover of Like a Prayer prompted action
ARIA’s decision did not happen in a vacuum. The new policy follows a chart controversy involving an AI-heavy cover of Madonna’s 1989 hit Like a Prayer by Australian DJ Josh Fawaz. The track combined AI-generated vocals with drums and other production elements, and it achieved significant commercial traction on Australian streaming services. At its peak, the cover reached No. 2 on ARIA’s overall Australian singles chart and remained one of the country’s most-streamed local songs for several weeks.
The song’s success attracted criticism from fans and industry figures who said that a recording with minimal human performance should not be able to occupy the same chart positions as fully human recordings. Spotify later updated its credits to disclose that AI had been used in the track, and international media reported that the song had received substantial streaming and radio exposure. The cover became a widely shared example of the legal and ethical gray areas created by generative AI, especially when a machine vocal is modeled on an existing artist’s voice.
ARIA CEO Annabelle Herd explained the organization’s intent in comments reported by the BBC. “Artists already use AI tools in their work, the charts can and should evolve to keep room for that, but music generated wholesale by services built on artists’ recordings is a different matter,” Herd said. That statement summarizes the challenge for the music industry: separating assistive from generative use, while also addressing how AI systems themselves may be trained on copyrighted songs.
The debate over AI-assisted versus AI-generated music
The ARIA update is one of the clearest attempts by a mainstream chart operator to establish a boundary between acceptable assistance and sufficient human authorship. The debate is not only about quality or authenticity; it is also about economic fairness and attribution. Songs that use AI to create vocal melodies, lyrics, or instrumental performances cannot be evaluated in the same way as songs written and performed by artists over months or years. If such recordings are allowed on charts, human acts could be displaced by a high volume of automated releases from services that generate music quickly and cheaply.
Industry experts have spent the past two years trying to define terms such as “AI-assisted,” “AI-generated,” and “fully synthetic.” ARIA’s approach is broad enough to encompass all of those categories along a spectrum. If a human provides the melody and lyrics but uses an AI tool to complete the mix, the song can still qualify. If a human records a voice and then replaces it with an AI model trained on that voice, the chart treatment may depend on whether the AI vocal is considered a lead performance. If a producer directs an AI platform to generate a complete track and adds only minor edits, that track is likely ineligible.
Compliance and enforcement under the new code
The new system relies on declarations rather than automated scanning. When a release is submitted, the rights holder or distributor must answer questions about generative AI use. ARIA can also apply enforcement retroactively. If a track was placed on a chart and later found to have failed the “substantially human made” test, ARIA can strip its chart position, revise rankings, and claw back awards associated with a No. 1 entry. This mechanism is intended to discourage artists and labels from giving incomplete disclosures.
Albums are assessed track by track. That means an AI-generated song included as one track on an album will not count toward that album’s chart eligibility, even though the rest of the album might remain eligible. This provision prevents an automated song from riding into the charts by appearing on an otherwise human album. It also gives chart managers a practical method for removing problematic material without penalizing an entire project.
The track-by-track review is significant for a streaming market where deluxe editions and long album cycles are common. A short AI interlude placed between full human performances will not invalidate the entire album, but it will not contribute to the album’s chart position or be recognized as a separate charting track.
Global implications for AI music policies
ARIA’s move may influence chart operators and industry bodies beyond Australia. In the United States, Billboard has already begun testing AI-detection technology with data provider Luminate. The two companies have explored ways to flag AI-generated tracks before they appear on Billboard charts, but no formal public policy has yet been adopted. ARIA’s code provides a workable model that others can adapt: use disclosure, define primary creative elements, and apply retrospective penalties.
International labels and distributors are watching the policies closely because the same recordings circulate globally. A song defined as ineligible in Australia can still chart in other countries unless additional regulators and chart bodies follow ARIA’s lead. The lack of a universal definition for human and AI participation remains a problem for streaming data providers, radio broadcasters, and chart compliance teams.
Another block of concern is copyright and voice rights. AI systems that generate vocals often rely on training data that includes commercially released songs, raising questions about whether the original artists have consented to their voices being replicated. ARIA’s rule does not directly solve copyright disputes, but it creates a business incentive for artists and labels to avoid synthetic lead vocals if they want to participate in the Australian charts. That may reduce the number of AI mimicry tracks submitted to ARIA, while still leaving many questions for courts and government agencies.
What this means for Australian artists
For most working musicians, the new rules will not change the daily routine. AI tools are now embedded in audio editing software, vocal tuning, mastering suites, and even suggestion-based production plugins. These tools are widely considered secondary, and they do not typically replace human authorship. ARIA’s announcement clarifies that producers can continue using them without putting their chart eligibility at risk.
The main shift affects producers and songwriters who create music entirely through generative services. Australia’s official charts are now closed to recordings that cannot demonstrate substantial human contributions. This may push some producers to revise their creative process, adding their own vocal performances, original lyrics, or live instrumentations. It may also create a market for “hybrid” workflows in which an artist uses AI ideation as a starting point but records a new performance and transforms the material into an original human-led work.
ARIA has also signaled that it will continue reviewing the rules as technology changes. The first version of the policy may not answer every edge case, but the structure is flexible. The CEO’s language suggests that ARIA wants to protect human artists while avoiding an overly broad ban that would alienate creators using advanced tools. The distinction between using AI and being replaced by AI is likely to become more important in the coming years, not only in Australia but everywhere music is made.
The broader AI disclosure landscape
Developments beyond Australia show that the issue extends well beyond music charts. Google, for example, is making visible watermarks optional for Gemini-generated images, videos, and music while retaining invisible SynthID markers and C2PA credentials to identify AI-created content. The company’s approach is less restrictive than ARIA’s, because it relies on technical identification rather than eligibility rules. But both systems share the same goal: ensuring that audiences and buyers know when a work has been generated by AI.
The music industry has been slower than other creative sectors to establish clear label requirements. Visual artists and writers have already adapted to AI disclosure rules imposed by major digital platforms. Musicians, record labels, and streaming services must now decide how to handle the influx of AI-generated compositions while still protecting the economic value of human creativity. ARIA’s code is a reminder that technology policy is not just about coding and licensing; it is also about defining what it means to be an artist in an era of machine-made culture.
Source: TechRepublic News