Submission 312
Australian Society of Authors (ASA)
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.
Summary
Read the original submission for the full text.
Argues AI training without permission or payment harms Australian creators, while supporting AI investment without weakening copyright. Proposes enforcing existing law for onshore training and reforms covering models offered in Australia wherever trained, including Copyright Act clarifications, statutory damages, transparency, voluntary collective licensing infrastructure, content labelling and First Nations cultural and intellectual property protection.
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.
The ASA states copyright law already requires permission for AI training done within Australia.
The law is simple: when copyright works are used for AI training undertaken within Australia, the activity is subject to the Copyright Act and must therefore be conducted with the permission of the copyright owner.
Argues that creators must be able to challenge AI models offered in Australia that rely on unauthorised use of their work, wherever training occurred.
If AI models offered in Australia rely upon unauthorised use of their work, Australian creators must be able to challenge this activity, no matter where the training took place.
Argues that the importation provisions of the Copyright Act should be updated to expressly cover AI models.
To put this prohibition beyond doubt, the government should update the importation provisions with AI models in mind. It could clarify the definition of ‘article’ in the Copyright Act to expressly include AI models and make clear they are ‘imported’ when they enter Australia by any means.
Asks for AI-developer transparency obligations so Australian rightsholders can identify their works in datasets used to train models deployed in Australia.
We request new transparency obligations be imposed on AI developers via the Copyright Act (or other legislation) so that Australian rightsholders can know if their works were included in datasets relied upon for AI training of models deployed in Australia.
Argues that voluntary collective licences negotiated with collecting societies should supplement direct licensing for smaller rights holders.
To supplement direct licensing and to represent the interests of smaller rights holders, voluntary collective ‘AI training’ licences must be negotiated with collecting societies.
Argues that Australians have a right to know when they are engaging with AI-generated content.
Australians have a right to know when they are engaging with AI-generated content.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Compel AI companies to license copyright work as a condition of doing business in Australia, through a mandatory code of conduct or, alternatively, a strengthened Copyright Act.
We propose the government intervene to compel Big Tech to license copyright work as a condition of doing business in Australia. Previously, we have suggested to government this might be achieved through a mandatory code of conduct but, if that is unpalatable, we suggest strengthening the Copyright Act.
Require multinational companies training AI onshore to abide by Australian law, stating no new legislative intervention is required for onshore training.
the Australian government must simply require multinational companies looking to train here to abide by our laws. No new legislative intervention is required with respect to onshore training.
Clarify the Copyright Act to expressly allow rights holders to bring a claim where their works were copied into an AI model made available in Australia, if that copying would have infringed Australian law had training occurred here.
The government could clarify the Copyright Act to expressly allow rights holders to bring a claim where their works have been copied into an AI model which is then made available in Australia, if such copying would have infringed Australian copyright law had the training occurred here.
Clarify the importation provisions of the Copyright Act so the definition of 'article' expressly includes AI models, which are 'imported' when they enter Australia by any means.
the government should update the importation provisions with AI models in mind. It could clarify the definition of ‘article’ in the Copyright Act to expressly include AI models and make clear they are ‘imported’ when they enter Australia by any means.
Deem AI models to contain copies of ingested copyright works and clarify that reproduction rights cover both training-dataset copying and continued reliance on embedded works.
The government could therefore make clear in legislation that AI models will be deemed to contain copies of the copyright works ingested as training data. The reproduction right must guard not only against copying of a work into a training dataset but also expressly against the continued reliance on embedded works within an AI model.
Introduce statutory damages into the Copyright Act at a level that dramatically increases careful adherence to the law.
A third approach is to introduce statutory damages into the Copyright Act at a level which dramatically increases careful adherence to the law.
Impose transparency obligations on AI developers, via the Copyright Act or other legislation, so rightsholders can know if their works were in datasets used to train models deployed in Australia.
We request new transparency obligations be imposed on AI developers via the Copyright Act (or other legislation) so that Australian rightsholders can know if their works were included in datasets relied upon for AI training of models deployed in Australia.
Support voluntary collective AI-training licences negotiated with collecting societies to supplement direct licensing for smaller rightsholders; copyright owners may choose whether to opt in.
To supplement direct licensing and to represent the interests of smaller rights holders, voluntary collective ‘AI training’ licences must be negotiated with collecting societies. Copyright owners may opt in should they wish to be represented by their relevant collecting society in such deals.
Build digital infrastructure now for future high-volume AI licensing; assist collecting societies with searchable databases enabling efficient licensing and content delivery.
Given that AI training is necessarily ongoing, it is important for the government to build digital infrastructure now that would support high-volume licensing into the future. As part of its investment into and vision for the AI economy, the government must assist collecting societies to develop sophisticated, searchable databases in formats that allow efficient licensing and delivery of content for AI training.
Introduce a classification system distinguishing AI-free, AI-augmented or assisted, and AI-generated goods and services.
To give consumers clarity and choice a classification system can be introduced to clarify services or goods that are: AI-free; AI-augmented or assisted; and AI-generated.
Develop sector-specific mandatory AI transparency codes with industry input and penalties for non-compliance.
Transparency requirements are likely to be sector specific so mandatory industry codes could be developed with industry input, with penalties for non-compliance.
Ensure AI companies using Indigenous Cultural and Intellectual Property obtain permission from both copyright owners and community cultural leaders and observe cultural protocols.
Indigenous Cultural and Intellectual Property (ICIP) must be respected. AI companies wishing to exploit material containing ICIP will require permission not only from copyright owners but also community cultural leaders, and will need to observe cultural protocols.
Impose positive duties on AI companies to exclude Indigenous Cultural and Intellectual Property unless they demonstrate compliance with First Nations protocols during ingestion and in safeguards against inappropriate outputs.
there ought to be positive duties imposed on AI companies to ensure their models do not contain ICIP, unless they can show they have complied with First Nations cultural protocols, both in the acts of ingestion and in safeguarding against inappropriate outputs.