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 507

Mr Sam Moran

Submitter group: Individuals.

Position labels are not shown for individual submitters.

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.

Proposes CARA: opt-in training licences, auditable records and royalties using broader licensed pools and an uncertainty fee. Says preliminary tests do not establish reliable attribution or royalty allocation. Recommends proportionate compliance and consultation and testing, with independent confidential audits, creator choice over public acknowledgement, and preserved rights to challenge use and payments.

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. Proposes CARA, a framework that would shift the burden of training data transparency from creative workers onto model builders and price the uncertainty of their systems.

    This Creative Attribution & Royalty Architecture (CARA) is a framework that shifts the burden of training data transparency from the workers onto the model builders and prices the uncertainty of their systems; financially incentivising them to increase their accuracy of provenance and attribution.

  2. Says provenance opacity burdens creative workers, who often lack the technical access, resources or bargaining power to challenge use of their work.

    Provenance opacity falls on creative workers, who are left to find out whether their work has been used to train AI models, understand that use and challenge it; often without the technical access, resources or bargaining power to do so effectively.

  3. Proposes that CARA begin with creator-controlled opt-in licensing and require model providers to keep auditable records of the works used in training.

    CARA begins with creator-controlled, opt-in licensing and requires model providers to maintain auditable records of the works used in training.

  4. Argues that royalty payment need not wait for perfect attribution; uncertainty can be handled through broader licensed pools and an uncertainty fee.

    The architecture does not depend upon perfect attribution before recognition or payment can begin. Where attribution remains uncertain, royalties can be distributed across progressively broader licensed pools, while an additional uncertainty fee creates an incentive for providers to improve their systems.

  5. Proposes that training-data manifests be given to an independent auditor rather than published, to protect commercially sensitive information.

    The detailed training-data manifest would not need to be publicly accessible; instead, developers would provide it to an authorised, independent auditor operating under confidentiality obligations and regulatory oversight. The proposed purpose is to protect commercially sensitive information while enabling verification of provenance, licensing permissions and remuneration obligations.

  6. States that initial results suggest attribution responds to engineering choices, without establishing reliable individual-work attribution, generalisation to commercial models or effective royalty allocation.

    These initial results suggest attribution performance can respond to engineering choices; it does not establish reliable individual-work attribution, generalisation to commercial models or effective royalty allocation.

Recommendations

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

  1. Ensure any future regulation requires transparent and auditable use of works, with a path to compliance for existing companies.

    It must also be considered that any future regulation ensures the use of these works must be transparent and auditable with a path to compliance for existing technology companies.

  2. Keep any compliance requirements proportionate, avoiding burdensome upfront costs that create barriers to entry for local startups and entrench well-resourced players.

    Additionally, any compliance requirements should be proportionate, avoiding unnecessary barriers of entry for competitive local startups through burdensome upfront costs that only work to entrench the existing, well-resourced players.

  3. Require developers to document and disclose training use in ways creators can understand and act on, supported by audit and redress mechanisms.

    Shifting this burden requires developers to document and disclose training use in ways creators can understand and act on, supported by audit and redress mechanisms.

  4. In CARA, use creator-controlled opt-in licensing and auditable training-work records, with provenance and attribution accompanying outputs to support royalties at the highest available confidence level.

    CARA begins with creator-controlled, opt-in licensing and requires model providers to maintain auditable records of the works used in training. Provenance and attribution information then travels with generated outputs, allowing royalties to be distributed according to the highest available level of attribution confidence.

  5. In CARA, pay royalties despite imperfect attribution through progressively broader licensed pools, adding an uncertainty fee to encourage improvement.

    The architecture does not depend upon perfect attribution before recognition or payment can begin. Where attribution remains uncertain, royalties can be distributed across progressively broader licensed pools, while an additional uncertainty fee creates an incentive for providers to improve their systems.

  6. In CARA, provide detailed manifests to authorised independent auditors under confidentiality and regulatory oversight, without requiring public access.

    The detailed training-data manifest would not need to be publicly accessible; instead, developers would provide it to an authorised, independent auditor operating under confidentiality obligations and regulatory oversight. The proposed purpose is to protect commercially sensitive information while enabling verification of provenance, licensing permissions and remuneration obligations.

  7. Keep training permission separate from permission to publicise a creator’s participation or present it as endorsement.

    Permission to license a work for AI training should not automatically constitute permission to publish the creator’s participation or present it as an endorsement.

  8. Let creators choose public acknowledgement or confidential verified attribution for rights administration and payment through their collecting society or licensing representative.

    Creators could elect public acknowledgement, while those preferring confidentiality would retain verified attribution for rights administration and payment through their authorised collecting society or licensing representative.

  9. Preserve creators’ access to check use, challenge unauthorised use and dispute payments; confidentiality must not override public reporting or lawful regulator and court access.

    Confidentiality would not prevent creators from checking whether their works were used, challenging unauthorised use or disputing payments. Nor would it override applicable public- reporting requirements or lawful access by regulators and courts.

  10. Report audit scope, findings and unresolved compliance gaps without automatically exposing creators or the full manifest.

    Public reporting would describe audit scope, findings and unresolved compliance gaps without automatically exposing individual creators or the complete manifest.

  11. Support consultation and practical testing of frameworks such as CARA with creative workers, researchers, rights organisations and technology providers, collectively addressing payment rates, independent oversight and fair distribution.

    I encourage the Committee to support further consultation and practical testing of frameworks such as CARA, bringing creative workers, researchers, rights organisations and technology providers into that process. Questions of payment rates, independent oversight and fair distribution need to be worked through collectively.

Read the original

Open in the interactive guide