AI Inquiry Atlas

An unofficial, independent guide to the public records of the Joint Select Committee on Artificial Intelligence. Not connected with the Parliament of Australia or the committee. Its makers also made a submission (no. 481; see About). Summaries and scores are written by AI and can be wrong: check the original before relying on them.

Submission 306

Australian Recording Industry Association (ARIA) and Phonographic Performance Company of Australia (PPCA)

Submitter group: Creative sector.

An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.

Open in the interactive guide

Summary

Read the original submission for the full text.

Argues that Australia's copyright framework, including its application to AI training, is consistent with major markets and that voluntary licensing of recorded music is functioning, so a text and data mining exception, statutory licence or opt-out scheme is unjustified. Recommends transparency of training use, AI labelling, application of Australian law to offshore-trained models and protection of First Nations cultural material.

Key points

Quotes are the submitter’s own words. Text taken from documents, including scanned pages read by text-recognition software, can differ from the PDF. Check the original. Listed in the same order as on the submission’s page in the interactive guide.

  1. Argues that Australia's copyright framework does not hinder AI investment, citing $10 billion of data centre investment in 2024.

    AI investment in Australia is not hindered by Australia’s copyright framework. Australia attracted $10 billion in data centre investment in 2024, making it one of the leading global destinations - second only to the United States.

  2. Argues that government should not override functioning licensing markets without evidence of market failure, which it says has not been provided.

    Governments should not intervene to override functioning commercial markets without clear and demonstrable evidence of market failure - evidence which clearly has not been provided as LLM companies refuse to negotiate with major rightsholders, instead seeking favourable commercial deals from government.

  3. Says the music industry has long licensed its content for new technologies, including streaming platforms, through voluntary licences.

    The music industry has for decades been adept at investing in and licensing its content for use in new technologies. Following the digital disruption of the 2000s, streaming platforms like Spotify, Apple Music, TikTok and YouTube all utilise voluntary licences for the use of music on their platforms.

  4. Argues that opt-out mechanisms relying on digital protection measures give copyright owners little real control and are unworkable for music.

    The “control” being offered by requiring copyright owners to apply digital protection measures at an asset level everywhere a track is made available digitally anywhere in the world is largely illusory, and the requirement for “digital protections” is unworkable and unreasonable for sectors such as music.

  5. Citing Deezer, says AI-generated songs account for 44% of daily streaming uploads, most streams are fraudulent, and bad actors dilute human artists’ royalties.

    Bad actors are taking advantage of AI music generation tools and using AI tracks to dilute the pool of royalties available to genuine human artists. AI-generated songs represent 44% of the daily uploads to streaming platforms, with a majority of streams of these songs detected as fraudulent.

  6. Argues that Australian laws and regulation should apply to all AI models put on the market in Australia regardless of where they were developed, trained or deployed.

    Australian laws and regulation need to be applied to all AI models put on the market in Australia, regardless of the jurisdiction in which the model may have been developed, trained or originally deployed.

Recommendations

Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.

  1. Not intervene in the commercial licensing market without evidence of market failure.

    Without evidence of market failure government should not intervene in a functioning commercial market simply because some of the world’s most valuable technology companies would prefer not to pay the market price or accept conditions on their use of other people’s property.

  2. Require AI developers to show they attempted to license copyright content in Australia and could not, as the minimum before any regulatory intervention in their favour is considered.

    Critically, AI developers have yet to provide any evidence that they have attempted to license copyright content in Australia but have been unable to do so. This should be the minimum requirement before any regulatory intervention in their favour is even considered.

  3. Ensure copyright reforms comply with free trade agreements and treaty three-step tests: special cases only, no conflict with normal exploitation, and no unreasonable prejudice to legitimate rightsholder interests.

    Government also must be mindful of any changes to copyright law that fall foul of free trade agreements or the three-step test incorporated into international treaties which Australia is party to (including the TRIPS Agreement and the Berne Convention). The three-step test requires that any exceptions to the exclusive rights of rights holders must be limited to special cases, must not conflict with the normal exploitation of the subject-matter, and does not unreasonably prejudice the legitimate interests of the copyright holder. This threshold may be crossed if Government unnecessarily interferes in rights holder licensing markets.

  4. Absent demonstrated voluntary-licensing market failure or evidence copyright hinders AI investment, do not subsidise developers’ music access through statutory licensing or other means.

    With no demonstrated failure of the voluntary licensing market, nor any evidence to suggest that AI investment is hindered by copyright, there is no justification for giving AI developers subsidised access to music whether through statutory licensing schemes or other means.

  5. Exclude sound recordings from any statutory AI-training licence.

    At the minimum, copyright in sound recordings should be excluded from the scope of a statutory licence, as there is clear evidence that direct licensing of music copyright material already exists and there is no market failure nor any public interest benefit that necessitates the need for a statutory licence for AI.

  6. Introduce meaningful training-input record keeping and disclosure without onerous burdens on AI companies, such as ARIA’s two-tier approach.

    Meaningful record-keeping and disclosure measures (such as the two-tiered approach proposed by ARIA in its submission to the Productivity Commission 24) would enable rightsholders to better manage and protect their works without imposing onerous burdens on AI companies.

  7. For ARIA’s illustrative two-tier approach, publish detailed summaries of copyrighted material used; let rightsholders request identifying information and sources/legal bases where summaries indicate their works were used.

    The two tiered approach involves publicly available summaries with sufficient detail on what copyrighted material has been used (tier one), with the ability for rightsholders to request more detailed information, including identifying information, where and on what legal basis the content was obtained, where the publicly available summary indicates their work/s have been used (tier two).

  8. Introduce labelling obligations on developers, deployers and users where AI-generated content could mislead, helping consumers distinguish human, AI-assisted and wholly AI-generated works.

    Labelling requirements would assist individuals in distinguishing between human, AI-assisted and wholly AI-generated works, allowing them to make informed decisions about the content they consume. Overseas, other territories such as the EU (through Article 50 in the EU’s AI Act25) and China26 are implementing labelling requirements, on developers, deployers and users, where AI- generated content could mislead persons. Implementing similar obligations in Australia would protect Australian consumers, while also giving an opportunity for human artists to distinguish themselves and their works from the large volume of competing AI slop.

  9. Apply Australian laws and regulation to all AI models put on the market in Australia, regardless of where they were developed, trained or first deployed.

    Australian laws and regulation need to be applied to all AI models put on the market in Australia, regardless of the jurisdiction in which the model may have been developed, trained or originally deployed.

  10. Consider an Australian Consumer Law-style approach covering businesses that carry on business in Australia.

    In Australia an approach similar to that taken with Australian Consumer Law should be considered, where businesses regardless of where they are based or where products are made, are subject to the ACL if they “carry on a business” in Australia.

  11. Include the National Cultural Policy (Revive) commitments in any AI legal or policy considerations and in any deals between Government and global AI companies.

    It is essential that the Government’s commitments as detailed in the National Cultural Policy: Revive27 are included in any AI legal or policy considerations or deals between the Government and global AI companies

  12. Safeguard First Nations art, stories and cultural knowledge against misuse and appropriation in AI development in any government policy or AI deals.

    Government must also safeguard First Nations art, stories, and cultural knowledge from misuse and appropriation in AI development in any policy or AI deals that it enters into.

Read the original

Open in the interactive guide