Submission 40
Australian Writers’ Guild (AWG) and Authorship Collecting Society Australia Aotearoa (ACSAA)
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 that copyright-infringement law is fit for purpose but needs enforcement, while proposing targeted training, moral-rights, First Nations and likeness protections. This joint writers’ guild and collecting-society submission describes threats to creative livelihoods and cultural sovereignty. It seeks ongoing reparative payments, temporary levy-funded relief and funding restrictions for AI replacements of creative workers.
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.
Argues that generative AI is built on mass infringement of copyright, moral rights and Indigenous Cultural and Intellectual Property.
The basis of generative artificial intelligence technology (AI) is mass infringement of copyright, moral rights and – in the case of First Nations’ artists – Indigenous Cultural and Intellectual Property.
Argues that copyright law dealing with infringement is fit for purpose, but creative workers cannot afford enforcement against large foreign technology companies.
Our view has always been that Australian copyright laws, to the extent that they deal with copyright infringement, are fit for purpose. The problem is that foreign tech companies have flaunted those laws without penalty. Australian creative workers are unable to afford legal action against the likes of Google, Meta and Amazon.
Argues that artists should hold an inalienable right to prevent use of their work to train AI models.
It is essential that artists have an inalienable right to prevent third parties from using their artistic work to train an AI model.
Argues that models trained on First Nations cultural material without permission should be forced to disgorge that training data.
AI models that have been trained on ICIP (i.e. in the form of images, transcriptions, and the like) without the permission of the relevant Traditional Owners should be forced to disgorge that training data.
Argues that AI platforms trained on unlawfully obtained content should be prohibited in Australia.
AI platforms that have been trained on unlawfully obtained content should be prohibited in Australia.
Argues that every Australian should own their voice, face and likeness.
Every Australian should be given ownership of their voice, face, or likeness.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Prioritise enforcement of existing copyright laws as the government's priority.
It is the enforcement of existing copyright laws that should be Government’s priority.
Give authors an inalienable right to prevent third parties using their work to train an AI model, with a right to consent provided they are equitably and proportionately compensated.
It is essential that artists have an inalienable right to prevent third parties from using their artistic work to train an AI model. They should have the right to consent to that exploitation too, provided they are equitably and proportionately compensated.
Extend copyright protection to currently unprotected creative works and activities, including the style and technique of screen editors, cinematographers, make-up artists and production designers.
We support the extension of the copyright framework to include creative works and activities that are not currently protected by copyright (e.g. the style and technique of a screen editor or cinematographer, a make-up artist or production designer).
Give every Australian ownership of their voice, face or likeness, which cannot be used in training or outputs without express permission, drawing on developments in Denmark.
Every Australian should be given ownership of their voice, face, or likeness. Their voice, face and likeness cannot be used in training or outputs without express permission. We suggest that Government turn to developments in the EU, particularly in Denmark, where creators, as well as ordinary citizens, can own and license their voice, face, and likeliness.
Deny funding to projects replacing creative workers with AI, at least until copyright concerns are addressed; require grant applicants to actively disclose AI use throughout the grants process.
These problems can be avoided if the federal and state funding agencies deny funding to any creative projects that use AI technology as a replacement (in whole or in part) for work that has traditionally been done by a creative worker at least until the copyright concerns raised in this submission are addressed by government. Any person or company applying for government funding must, throughout the grants process, have obligations to actively disclose any use of AI technology.
Engage in careful consultation with First Nations Traditional Owners and introduce regulations protecting the use of Indigenous Cultural and Intellectual Property (ICIP).
Government must engage in careful consultations with First Nations Traditional Owners and introduce regulations protecting the use of ICIP.
Require AI companies to ensure their models do not contain and are not trained on ICIP, until the legal-recognition issues are resolved.
Until these issues are resolved, AI companies must ensure their models do not contain ICIP, and are not trained on ICIP.
Require AI companies to meet specific transparency obligations on use of ICIP, including what ICIP has already been used to train their models.
AI companies must have specific transparency obligations around the use of ICIP, particularly around what ICIP has already been used to train their models.
Require models trained on ICIP without Traditional Owners' permission to disgorge that training data, with AI companies obliged to provide evidence that this occurred.
AI models that have been trained on ICIP (i.e. in the form of images, transcriptions, and the like) without the permission of the relevant Traditional Owners should be forced to disgorge that training data. The AI companies responsible must have obligations to provide evidence that this disgorgement occurred.
Prohibit in Australia AI platforms trained on unlawfully obtained content; content whose access would be unlawful in Australia must not be part of any model accessible in Australia.
AI platforms that have been trained on unlawfully obtained content should be prohibited in Australia. If the access would be considered unlawful in Australia, it must not be a part of any model accessible in Australia.
Place a positive duty on AI companies to remove unlawfully obtained content from their models and disclose ongoing compliance to the regulator, with strict statutory penalties.
As with ICIP, AI companies must have a positive duty to remove unlawfully obtained content from their models, and make appropriate disclosures to the regulator demonstrating ongoing compliance (with strict statutory penalties for non-compliance).
Extend inalienable authorship rights that persist through commercial relationships, and require equitable and proportionate remuneration for sale or exploitation of economic rights.
Authors and creators should have an inalienable and continuing relationship to their work that persists, despite whatever commercial relationships are entered into along a supply chain. For authors – as the term is broadly used in the Copyright Act – this means extending inalienable authorship rights (sometimes called the neighbouring rights or the right to equitable remuneration) which are rights cannot be sold or alienated and always belong to the author or creator. Furthermore, authors must be equitably and proportionately remunerated for the sale or exploitation of their economic rights.
Ensure new economic and moral rights cover all existing works.
We have argued that affirming the rights for authors and creators can be done through minimal amendments to the Copyright Act to establish both new economic and moral rights, provided that these new rights cover all existing works.
Add a new economic right 'to train a computational system' to section 31(1)(a) of the Copyright Act.
We propose a similar reform, for similar reasons: the inclusion of “(vi) to train a computational system” in section 31(1)(a) of the Copyright Act.
Introduce a moral right prohibiting copying an author’s original work to train any software model without consent.
We support the introduction of a new moral right prohibiting the copying of an author’s original work to ‘train’ any software model without the author’s consent.
As an alternative moral-rights reform, broaden derogatory treatment to encompass non-consensual AI training and production of outputs based on an artist’s work.
Alternatively, we propose that the definition of “derogatory treatment” of a work under the moral rights framework is broadened. “Derogatory treatment” is defined in the Copyright Act as any act “that results in a material distortion of, the mutilation of, or a material alteration to, the work that is prejudicial to the author's honour or reputation”. In our view, the act of training an AI system using an artist’s work and to produce an output based on that work, without the artist’s consent, is a distortion or mutilation of that work.
Fairly compensate all creative workers whose output AI companies copy, even where copyright does not recognise that output as a work or subject matter.
It is critical that all creative workers whose output – which is a product of their individual effort, creative aspirations and skill – has been copied by AI companies should be fairly compensated even if the current copyright framework does not consider their creative work to be a ’work’ or ’subject matter’ for the purposes of copyright subsistence.
Extend moral rights to creative workers without copyright in the Part III literary, dramatic, artistic or musical works comprising Part IV subject matter.
We support, at minimum, the expansion of the doctrine of moral rights to cover creative workers who do not own copyright in any of the Part III literary, dramatic, artistic or musical works that a Part IV subject matter (i.e. a feature film, an episode of a television series) is comprised of.
Require AI companies to obtain creative workers’ consent to breach their moral rights and, if consent is granted, pay equitable and proportionate remuneration.
AI companies would need to seek the creative worker’s consent to breach their moral rights and (assuming the creative worker is willing to grant that consent) pay them equitable and proportionate remuneration for the exploitation of their work.
Include ongoing reparative payments to creators and rightsholders whose rights have been infringed in any AI regulation, collected by their relevant collective management organisation.
In our view, any AI regulation must include reparative payments to creators and rightsholders whose rights have been infringed. These payments are ongoing and collected by the relevant Collective Management Organisation (CMO) that directly represents a class of creative worker: for audio-visual works this includes ACSAA, ASDACS (for directors), and APRA-AMCOS (for musicians and composers)
Have AI companies provide financial support for government creative bodies to deliver AI compliance work on funding and grants.
AI companies should provide financial support for government creative bodies to ensure that they can deliver AI compliance work as it relates to funding and grants.
Have the expanded copyright framework contemplate excluding artists' names as prompts, and full and ongoing transparency about input data.
This expanded copyright framework should also contemplate the specific exclusion of artists’ names as prompts, and full and ongoing transparency around input data
Require AI companies to remove impersonating content or personal data used in training within a strict deadline (e.g. 24 hours) once made aware.
When an AI company is made aware that their platform has been used to impersonate a real person, or where personal data has been used as part of a training data set, that AI company mist have a strict deadline to remove that content (e.g. 24 hours).
Require a warning on media content or advertising developed with AI, and require services using AI recommendations to warn users and offer an opt-out.
Whenever media content or advertising that has used to AI technology in its development or production is communicate to the public, it should be accompanied by a warning that AI technology has been used. Services that use AI to make recommendations to users must warn users that they are doing so and give users the opportunity to opt-out.
Establish a working group principally of First Nations screen and stage creatives, supported by industry organisations, subject experts and collecting agencies, to develop standalone cultural-assets legislation.
A core promise of Revive was to facilitate the implementation “stand-alone legislation to protect First Nations knowledge and cultural expressions”. We supported this goal and proposed the formation of a working group of relevant experts comprised principally of First Nations creatives in the screen and stage sector, supported by a relevant industry organisation, other relevant subject matter experts, and the collecting agencies.
Have the First Nations working group consider cultural assets as copyrightable assets in stage and screen contracts, industry terms of trade and contracts for performed or cinematographic works based on those assets.
We recommended that the working group would consider the translation of ‘cultural assets’ as a class of copyrightable assets into stage and screen contracts, in addition to industry terms of trade and contracts for performed or cinematographic works based on these assets or expressions.
Compel AI companies to disclose whether platforms have been trained on Australian works, including First Nations works.
AI companies must be compelled to share data confirming whether or not AI platforms have been trained on Australian works, including First Nations works.
Enable First Nations Traditional Owners to refuse consent to use of cultural assets for AI training and to have works disgorged at any time, with reasonable assurance that disgorgement has happened.
First Nations Traditional Owners (as with all creators) must have the right to refuse consent for tech companies to use cultural assets for ‘training’ AI models, and to have works disgorged at any time. Tech companies that have used these assets without permission (i.e. in the form of images, transcriptions, and the like) should be forced to disgorge that data and demonstrate this has occurred. The regulation must emphasise transparency and First Nations Traditional Owners must be given reasonable assurances that the disgorgement has taken place.
Enable consenting First Nations Traditional Owners and creatives to set their own terms for use of cultural assets.
First Nations Traditional Owners and creatives who do consent to the use of cultural assets must be enabled to do so on their terms (for example, a tech company may have a restricted licence to use an image; or the right to use a First Nations actor’s image only while they are alive).
Consider the First Nations data sovereignty work of Te Mana Raraunga.
We urge Government to consider the First Nations data sovereignty work of Te Mana Rarunga.
Develop legislation protecting First Nations knowledge and cultural expressions against exploitation without consent with DISR and the AI Safety Institute, which should bear administrative and support costs.
The development of legislation to protect First Nations knowledge and cultural expressions, with specific protections against tech companies exploiting cultural assets without consent, should be done in conjunction with the Department of Industry, Science, and Resources, as well as the AI Safety Institute. These organisations should bear the administrative and support costs to develop this legislation
Compel AI companies to disclose training data and confirm whether First Nations works were used to train LLMs, GANs or other models.
AI data transparency: AI companies must disclose training data and confirm whether First Nations works were used to train LLMs, Generative Adversarial Networks (GANs), or other models.
Have AI companies fund the development of sovereign AI models by and for First Nations creators.
Funding from AI companies to fund the development of sovereign AI models by and for First Nations creators.
Require consent from creative workers, with an appropriate compensation model, before their work is used to train AI datasets.
Requiring consent is given by creative workers before their work is used to ‘train’ AI datasets, with an appropriate compensation model;
Ensure creative workers are made aware that infringement of their work has taken place.
Ensuring that the creative workers are aware that the infringement has taken place;
Implement a compensatory and rectification process for infringements that have already occurred, including removal of work from models.
Implementing a compensatory and rectification process for infringements that have already taken place, including the removal of work from models;
Make a no-cost jurisdiction available to creators seeking remedy for infringement, with transparency required and failure to provide it resulting in default judgement against the AI company.
A no-cost jurisdiction be available to creators to seek remedy where they believe their work has been infringed, where transparency is required and any failure to provide transparency results in a default judgement against the defendant AI company.
Provide an interim emergency relief package to sustain creators during AI disruption.
Government could deliver for creators in the form of an interim emergency relief package for creators in order to ensure the sustainability of the creative sector.
Fund interim relief through a temporary levy on generative AI providers in this market, without interrupting discussion of longer-term structural solutions.
This relief package should be funded either by a temporary AI levy imposed on providers of generative AI services in this market. We refer to our confidential submissions on how these payments might be managed. Such a relief package must be understood as a temporary measure and ought not interrupt ongoing conversations about the longer term structural solution needed from government.