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 195

Media, Entertainment, and Arts Alliance

Submitter group: Unions.

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 AI is displacing and undercutting creative and media workers and that copyright reform alone cannot address this. Proposes a fee on foreign-owned AI compute and software funding an equitable remuneration scheme and cultural fund, mandatory training-data reporting, take-down notices, AI content labelling, ICIP protections and competition oversight.

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 AI tools are already being used in newsrooms, screen and music, with widespread labour effects.

    AI tools are already being rolled out in newsrooms, on screens, and in streamed music, and the labour effects are now widespread.

  2. Argues that copyright was not designed to deliver compensation or control to original creators and cannot address displacement.

    Copyright reform simply cannot fill the gap. It was never designed to deliver compensation or control to original creators of works, and cannot address issues of redundancy, undercutting, and devaluation in the era of AI.

  3. Proposes a targeted fee on foreign-owned AI compute and software, including GPUs, accelerators, data centres and frontier models.

    To remedy these imbalances, we propose a targeted fee imposed on foreign-owned compute and software across the AI stack, including GPU’s, accelerators, data centres, and frontier models.

  4. Recommends labelling of all AI-generated content through technical mechanisms such as watermarking.

    All AI-generated content should be labelled through a robust technical mechanism like watermarking to ensure audiences are informed about any potential or actual engagement with AI-generated content.

  5. Recommends take-down notices for works misrepresenting a connection with a human creator.

    Government should institute a system of take-down notices to be applied to works that misrepresent a connection with a human creator, including unauthorised digital replicas, surrogates, and derivatives.

  6. Proposes a right to opt out of personal data and content being used to train AI on platforms.

    Enact a GDPR-style right to limit the collection, use, or sharing of personal data on digital platforms. Users should retain the right to opt out of the use of any content shared on platforms being used to train AI.

Recommendations

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

  1. Reform copyright to deliver an inalienable right for creators to be paid for secondary uses of their work, including in broadcast.

    copyright should be reformed to deliver an inalienable right for creators to be paid for secondary uses of their work, including in broadcast.

  2. Levy a targeted fee on foreign-owned compute and software across the AI stack, including GPUs, accelerators, data centres and frontier models.

    To remedy these imbalances, we propose a targeted fee imposed on foreign-owned compute and software across the AI stack, including GPU’s, accelerators, data centres, and frontier models.

  3. Use the AI fee for direct worker payments linked to AI use of their works and a cultural fund for live entertainment, screen, music, public interest journalism and cultural experiences.

    Recommendation 1: Levy a fee on AI companies across the AI stack to support an equitable remuneration scheme that delivers a) direct payments to workers linked to the use of their works in AI systems; and b) a cultural fund to support the delivery of live entertainment, screen, music, public interest journalism, and cultural experiences.

  4. Ensure the remuneration framework supplements the Copyright Act without disrupting developing licensing markets for copyright owners and AI companies.

    Importantly, the proposed framework supplements and supports the Copyright Act 1968 (Cth), without disrupting the developing licensing markets designed for copyright owners and AI companies, while ensuring that the value of Australian labour is included in the cost of doing business.

  5. Make royalties and other AI-related income distributions reflect human involvement in the output and the user's capacity for free and creative choices.

    Royalty payments or other distributions of income related to use of AI systems should reflect the extent of human involvement in creation of the output and the user’s capacity for free and creative choices.

  6. Distribute equitable remuneration annually, adjusting the funds-pool calculation once the required technical infrastructure is established.

    An enforceable obligation on AI providers to provide equitable remuneration for benefits received through scraping would provide some measure of parity for creators. Payments should be distributed annually, with the calculation of the funds pool adjusted once the required technical infrastructure is established.

  7. Phase remuneration with labelling and tracing infrastructure: initially pay for works in training data, then expand to generated outputs and derivative uses and distribution.

    There are three clear categories of use of original works by AI systems where payments could be triggered: (a) the input stage where works form part of training data; (b) the output stage where works are generated by AI systems; (c) derivative uses and distribution of AI-generated outputs. A staged time frame would be required to implement this scheme, given adoption and use of relevant labelling and tracing technologies is required. Initially payments would be limited to where creator works from part of training data, but would expand over time.

  8. Calculate an AI operations and licence fee across foreign-owned GPUs, accelerators, storage, networking, model-serving, APIs, orchestration, data pipelines, model weights and developer tools, accounting for low-benefit offshore processing.

    MEAA recommends a more targeted initiative: An ‘AI operations and licence fee’ that takes into account the extent of the foreign-owned compute and software stack: GPUs, accelerators, storage, networking, model-serving, APIs, orchestration, data pipelines, model weights, and developer tools. The Fee calculation can account for the extent to which data centres and AI infrastructure are simply processing data for offshore entities, offering little or no benefit to Australians.

  9. Have the cultural fund enable public-good, non-commercial arts infrastructure where relevant, including communication and distribution platforms and AI systems.

    A cultural fund is required to help build an Australian sense of belonging that AI technologies cannot provide and appropriately value the contribution of all Australians to the new world-building project of AI. It is also needed to enable the arts sector to create, where relevant, non-commercial infrastructures based on public good principles, such as communication and distribution platforms and AI systems.

  10. Consider coordination of the fund by relevant unions and professional organisations under third-party oversight, without new bureaucracy or AI-company involvement beyond fee negotiation.

    Crucially, this fund would not require new government bureaucracy or AI companies to be involved beyond the negotiation of the proposed AI Operations and Licence Fee, simplifying the administrative burden on both. Instead, the Fund could be coordinated by relevant unions and professional organisations, overseen by a third-party.

  11. Extend training-data transparency beyond general-purpose models to tools and packages marketed to and used by professional artists, artist communities and consumers, with independent research verification and periodic audit.

    Training data content information is required to ascertain which classes of creative and media workers require compensating. An Australian A training-data transparency (TDT) requirement is widely anticipated. To appropriately service the creative and media industries, it should extend beyond general-purpose AI models to include AI tools and packages marketed to and used by professional artists, artist communities, and their consumers. Copyright litigation may also provide relevant information about data inclusions. In turn, the information provided must be verifiable through independent research and periodic audit.

  12. Introduce mandatory reporting of all materials used to train AI and all materials captured under AI licensing deals.

    Recommendation 2: Introduce mandatory reporting of all materials used to train AI and all materials captured under AI licencing deals.

  13. Require disclosure of specific contractual provisions affecting creators whose works are included in AI datasets.

    There must be mandatory reporting of copyright licensing deals with AI companies and mandatory disclosure of specific contractual provisions affecting creators whose works are included in AI data sets.

  14. Institute a system of take-down notices for works that misrepresent a connection with a human creator, including unauthorised digital replicas, surrogates and derivatives.

    Recommendation 3: Government should institute a system of take-down notices to be applied to works that misrepresent a connection with a human creator, including unauthorised digital replicas, surrogates, and derivatives.

  15. Empower creators affiliated with recognised professional organisations to remove unauthorised imitations of arts professionals' image, likeness, voice and style; target sources and commercial distributors.

    A control measure, similar to existing take-down notices utilised to inform carriage service providers of alleged copyright infringements, is required to empower creators affiliated with recognised professional organisations to remove unauthorised works designed to imitate the image, likeness, voice, and style of arts professionals. The obligation to remove content should target the source and all parties engaged in trade and commerce distributing unauthorised material.

  16. Allow take-downs for fraudulent works and AI outputs infringing copyright or overriding a creator's AI opt-out or no-derivative directive, as well as misrepresented human connections.

    Uncontroversial forms of unfair competition that should trigger a take-down include works misrepresenting a connection with a human creator, works motivated by fraud, or AI-generated work that infringes copyright or overrides an AI opt-out or no-derivative directive from a given creator.

  17. Label all AI-generated content through a robust technical mechanism such as watermarking, with take-down notices for works not labelled or whose labels are removed, obscured or altered.

    Recommendation 4: All AI-generated content should be labelled through a robust technical mechanism like watermarking to ensure audiences are informed about any potential or actual engagement with AI-generated content. A system of take- down notices should be applied to works not appropriately labelled as AI- generated, or where labels have been removed, obscured, or digitally altered.

  18. Avoid legislating technical transparency specifications because the measures are evolving rapidly.

    Transparency measures are rapidly evolving. It is thus inappropriate to include in legislation detail about technical specifications.

  19. Mandate additional technical transparency requirements rather than relying on industry self-regulation to incorporate Australian values.

    However, industry self-regulation is unlikely to incorporate Australian values, and additional technical requirements will need to be mandated.

  20. Withhold social licence from models capable of deceiving the public through generated content unless responsibility is taken for resulting harm or deception.

    While new and smaller-scale models may also not currently have the capacity to include robust labelling, there should be no social licence to deploy a model that can deceive the public through generated content without taking responsibility for any harm or deception caused by its use.

  21. Enact an enforceable system of Indigenous Cultural and Intellectual Property and Indigenous Data Sovereignty, with take-down notices for AI-generated culturally offensive or misleading content as defined in relevant protocols.

    Recommendation 5: Enact an enforceable system of Indigenous Cultural and Intellectual Property (ICIP) and Indigenous Data Sovereignty. Establish a system of take-down notices to be applied to works that include AI-generated culturally offensive or misleading content, as defined in relevant protocols.

  22. Protect Indigenous knowledge and culture through culturally appropriate, First Nations-led control measures and ongoing consultation across all relevant government departments.

    Commitments to protect ICIP and Indigenous Data Sovereignty should be supported by control measures so that Aboriginal and Torres Strait Islander knowledge and culture is protected in a culturally appropriate way, led by First Nations People and with ongoing consultation across all relevant government departments.

  23. Provide a simple take-down process for offensive or misleading content distributed in violation of metadata access/use conditions and relevant professional codes and protocols.

    Offensive and misleading 41 Terri Janke (n.d.) Protocols for using First Nations cultural and intellectual property in the arts, Creative Australia, https://www.creative.gov.au/first-nations-arts/protocols-for-using-first-nations-cultural-and- intellectual-property-in-the-arts 18 content made and distributed in violation of access and use conditions contained in metadata and as defined in relevant professional codes and protocols should be subject to a simple take-down process, as currently occurs with online copyright infringement.

  24. Enact a GDPR-style right to limit the collection, use or sharing of personal data on digital platforms, including a right to opt out of content being used to train AI.

    Recommendation 6: Enact a GDPR-style right to limit the collection, use, or sharing of personal data on digital platforms. Users should retain the right to opt out of the use of any content shared on platforms being used to train AI.

  25. Reform Australian law so exploitation of creative and cultural work requires creators' informed consent and compensation.

    It is crucial that the law is reformed in Australia to ensure that the exploitation of creative and cultural work is only possible with the informed consent and compensation of creators.

  26. Ensure the ACCC monitors potential anti-competitive behaviour from platform AI investments, especially mergers, investments and acquisitions.

    The MEAA urges the ACCC to remain vigilant of potential anti-competitive behaviour arising as a result of platform investments in AI. Government should ensure that the ACCC is alert to any anticompetitive behaviours arising in the AI sector, particularly mergers, investments, and acquisitions.

Read the original

Open in the interactive guide