For many genAI tools, information you share can end up being used as training data, which could then be included in the responses/outputs the tool provides to other users.

Research has shown that many genAI tools plagiarise from their training data, so there is potential for parts of material you include or upload in genAI prompts being used verbatim in outputs elsewhere.

Issues around privacy and use of information have also been flagged with some platforms. Anything you share with genAI platforms, like questions, prompts, or attachments, is typically stored by the third-party service provider. This means your information could be at risk from cyberattacks or other security threats.

When using genAI tools you should be careful not to share any personal, confidential, sensitive, proprietary, or other commercial information with genAI without understanding the platform’s privacy and security policies.

This is why data security is so important – where data security is in place, your prompts (and any materials included within) will not be shared with the broader system or used as training data. The University’s enterprise access to Microsoft Copilot includes data security and is free for staff and students. This provides greater certainty if you wish to upload your work as part of a prompt.

Using platforms that have data security and limit training data may also allow for greater use of copyright materials under exceptions in the Copyright Act 1968 and/or other licensing. More information around using copyright materials in prompts can be found in the Using copyright materials with genAI section below.

Have a question? Contact the Copyright Advisor

There is much discussion around whether current genAI tools were legally trained using copyright material scraped fom the web. In the U.S., a large number of lawsuits are still pending against OpenAI, Meta, and other genAI/LLM owners, claiming that AI training is infringing in nature and not covered under the ‘fair use’ doctrine in U.S. legislation.

OpenAI has publicly admitted the necessity of using copyright materials in training data, suggested that useful genAI would be "impossible" without it. The main issue around this is that no consent was sought for this use, and in some cases, training was done illegally using pirated/infringing material.

The first U.S. trial to be concluded in this space (February 2025) found that the training of the AI tool from the case was not fair use, as it was in direct competition with the source material it was trained upon. A determination of fair use will often require a decision from the courts and this can be very case-by-case in approach. So while the determination of this first case is important, the outcomes of further cases may vary due to the circumstances involved.

This is still an evolving space, and many platforms are now actively licensing content to help mitigate against further legal action. There has also been growth in the so-called 'ethical AI' space. Some genAI tools, such as Adobe's Firefly image generator have only been trained on licensed content. Training datasets containing only public domain materials are also gaining prominence.

In Australia, fair use does not exist in our legislation for potential use with AI training. Our local version, 'fair dealing', is much more limited in scope. As a result, use of copyright materials to train AI tools would likely require permission from the copyright owner/s.

Have a question? Contact the Copyright Advisor

Where prompts are sufficiently original and meet the usual requirements to be copyrightable, prompts may qualify for protection under copyright.

Where you upload your own material as part of a prompt (e.g., a document or an image), you will continue to hold the copyright in that material, regardless of changes made by the tool. For more considerations around uploading your own material, see the accordion for The importance of data security and privacy above .

This is the same for third-party copyright material – ownership does not change because material is uploaded into genAI. When reusing copyright materials, there are requirements you must follow so that you do not infringe copyright, as outlined in our Copyright Compliance Policy.

More information around using copyright materials in genAI prompts is included in the section below on Using copyright materials with genAI.

Have a question? Contact the Copyright Advisor

Who owns the copyright in genAI outputs currently depends on the situation, as different countries have different approaches.

The U.S. Copyright Office (USCO) has issued guidance to state that prompts alone "do not provide sufficient human control" for users of to be considered the creators of the output. They note that "prompts essentially function as instructions to convey ... ideas", which are not protectable under copyright. While highly-detailed user prompts could contain expressive elements, at present humans cannot control how an AI system will processes them to 'create' outputs. Essentially, it is the (black box) algorithm doing the 'creative' work and copyright can only be held by humans.

There is, however, potential for copyright protection where humans contribute more to the 'creative' process, such as for:

• Assistive uses of AI systems - where use of the tool is not "expressive" or a "stand in for human creativity", i.e., the 'human-first' approach for creation. Note that using multiple prompts would not qualify here. The USCO suggests that such human-created work should be protectable by copyright.
• Expressive inputs - where a human creator uploads their creative expression into an AI tool and asks it to modify or add to it. The USCO suggests where human "expressive elements are clearly perceptible in the output", limited copyright may be granted, notably excluding any "non-human expression".
• Modifying or arranging AI-generated content - where a human (i) modifies AI-generated content "to such a degree", or (ii) selects or arranges AI-generated material in a "sufficiently creative way", that the resulting work may "meet the standard for copyright protection", the USCO notes limited copyright may apply. Any underlying AI-generated content would be excluded from this protection.
• Inclusion of elements of AI-generated content in a larger human-authored work - for example, including AI-generated special effects or background artwork in a film. The USCO notes that doing so does not affect the copyrightability of the larger human-created work as a whole, even if the AI elements included are not copyrightable separately.

While U.S. legislation differs from Australia's by requiring registration for copyright, the human authorship requirement is the same for both countries. The considerations above should provide useful guidance for Australian creators using genAI until local legislation is updated.

UK law provides protection for works generated by a computer without a human creator, but there are several interpretations around how the wording of this legislation will apply to genAI, including who will own the copyright in works generated - the platform, the developers, or the users who input the prompts - which confuses the matter somewhat.

While Australia shares some characteristics of both the U.S. (extended copyright duration, requirement for human authorship) and the UK (fair dealing) systems, there is no official direction in our legislation around genAI as yet. It has been suggested in multiple legal opinion pieces that AI-generated works will not be protected in Australia. In terms of intellectual property more generally, the High Court has already ruled that AI cannot be named as an inventor for patents.

Where copyright material is added to prompts, such as documents, images, etc., the original owner will likely retain copyright in the outputs where this is reproduction or an adaptation of the original. There may also be potential issues where a genAI tool creates a work with substantial similarity to an existing work (such as with image generation). For these reasons it’s important to consider copyright/ownership when using genAI tools, especially where third-party materials are involved.

For more information on see the section below on Using copyright materials with genAI.

Have a question? Contact the Copyright Advisor

GenAI platforms may have specific terms noted around how outputs from the platform can be used, and/or how user inputs will be utilised.

As noted in the accordion The importance of data security and privacy above, it’s not uncommon for these platforms to add user prompts and materials to their training data. As part of their terms, user content can also be sublicensed to external parties or even shared with foreign governments.

Some online resources may include specific clauses in their site Terms of Service/Use that explicitly forbid the use of their content with genAI tools. Where this is the case, permission should be sought from the copyright owner to make use of the material.

Have a question? Contact the Copyright Advisor

Copyright protects the expression of ideas, but not styles or techniques. As a result, it is possible for genAI users to create works similar in nature to popular songs, artworks, photographs, etc., by adding "in the style of [X]" to a prompt.

While this isn't technically copyright infringement, creators are often unhappy about this functionality as it highlights their works have been scraped into training data. Some 'style'-based outputs may also fail the substantial similarity test for copyright or include "protectable expressive elements" and become infringing.

Style has become one of the ethical discussions around the use of various genAI tools, including for text (stories or lyrics in the style of...), images (artwork in the style of...), and music (songs in the style of...), especially around the lack of acknowledgement and potentially lost work for the creators being imitated.

It's interesting to note that many genAI tools now ignore or block this kind of prompting, perhaps in relation to the large number of copyright cases currently being litigated. OpenAI has even announced that the latest model (GPT-4o at time of writing in late March 2025) has policies in place that will prevent generation of outputs that "directly mimic any living artists’ work". However, initial reporting has suggested that these policies may not be quite working as intended, especially where a studio style is closely linked to an artist's style.

Have a question? Contact the Copyright Advisor

The short answer is – it depends. The amount that users can potentially upload without infringing copyright will vary based on the use, the licensing/permissions in place for the material, whether the output would be considered a derivative work, and what you plan to do with the output.

How likely is that that I can upload an entire copyright work to genAI?

Again, short answer – not very likely, but it will depend on the type of material and the use case.

For most copyright material, uploading or otherwise including an entire copyrighted work as part of a prompt will likely infringe copyright. There are some exceptions noted in other accordions below. For example, uploading a single article/chapter to a genAI tool may be acceptable in some situations, whereas uploading a whole ebook will likely remain an infringing use unless there is a major change to Australian copyright legislation.

Do not assume that because something is publicly available you can just use it with genAI. The same is true for resources from the Library’s subscriptions such as ebooks and journals – not all material is suitable to include in genAI prompts due to licensing or other factors.

The GenAI Basics section above discusses the training of genAI tools using copyright materials and the ownership of outputs.

Have a question? Contact the Copyright Advisor

The University has enterprise access to Microsoft Copilot that is free for staff and students.

When logged in through the University's access, users receive data security that covers anything they share with the AI tool, whether in a prompt or uploaded as an attachment. This data security is important as it ties into a number of licensing and other potential options for the use of copyright materials with genAI, as outlined in the accordions below:

• Use under the educational statutory licence (S113P in the Copyright Act 1968 [the Act]) requires a closed enterprise system endorsed by the University.
• Use under publisher subscription/licensing agreements may also require a closed access system.
• Reliance on the fair dealing exceptions in the Act is more likely to apply to uses where a closed-access system is in play.
• University-owned materials can only be uploaded to an endorsed system that does not share material externally.
• Feedback for student materials can only be generated via an endorsed platform.
• Copyright owners are more likely to grant permission for use of materials with genAI where they know the materials won't become part of the tool's training data.

Where data security is in place, such as for Copilot, you also have greater certainty around uploading your own work if you would like to do so as part of a prompt.

For more information on Copilot see the Copilot FAQs from DTS (login required for SharePoint).

More around the use of genAI can also be found on the University's AI Tools guide.

Some publishers, such as Wiley, have already included clauses in their subscription agreements that allow for use of “limited portions” of their content with genAI for the purposes of non-commercial research and academic teaching.

This permission does not allow for systematic download of the licensed resources, or use with an external party. There is also a requirement for use within a local, secure network environment, meaning that permitted use would likely be limited to the University’s enterprise access to Copilot.

Note that this is an evolving space – we are still waiting to confirm allowed use with genAI under many of the publisher subscription agreements.

Do not assume that all publishers will allow for this type of use.

Have a question? Contact the Copyright Advisor

The educational statutory licence is an exception in the Copyright Act 1968 that allows the University to make use of certain copyright materials.

The Copyright Agency has announced limited use of materials covered by the statutory licence will be allowable with certain genAI tools. This licence would, therefore, potentially cover some materials not already included in publisher agreements around genAI (detailed in the accordion above), but only for the “educational purposes” of the University.

The statutory licence also places limitations on the amounts of material that can be used, usually 1 chapter or 10% of a work, or 1 article per issue of a journal. It does not allow for the use of an entire ebook or for some types of copyright material.

There is a further restriction of the statutory licence only applying where use is made of materials with 'closed' tools / enterprise access, where data-security is in place. For the University, this would be our access to Microsoft Copilot.

Note that materials used under the statutory licence will also require some form of attribution/referencing.

Distinctions around student use of copyright materials with genAI in courses and assessments have, as yet, not been confirmed with the Copyright Agency. There is potential that, where students are directed to use genAI as part of coursework or assessment, the use could become reliant on the statutory licence rather than fair dealing (as outlined in the accordion directly below). As a result, students should be directed to use Copilot over other tools to help assist with compliance around the licence.

AI Tools guide: Issues and considerations

Have a question? Contact the Copyright Advisor

The Copyright Act 1968 (the Act) includes an exception for the University to use third-party content for examination purposes without infringing copyright.

Depending on the use, coordinators may be able to rely on Section 200(1A) of the Act for copying and communicating ‘the whole or a part of copyright material’ with genAI tools, as long as the material is used to generate and/or as part of the questions in an examination. S200(1B) includes the ability to adapt or localise materials for use in an examination.

Students may also be able to rely on the options under this exception when answering examination questions.

Best practice would be to use a 'closed' genAI tool like the University's enterprise access to Microsoft Copilot that does not train its model on user prompts.

Use of copyright material within an examination still requires attribution. To learn more, access the Examinations copyright guide (PDF).

Have a question? Contact the Copyright Advisor

The fair dealing exceptions in the Copyright Act 1968 may allow for certain uses of copyright materials with genAI tools, with a large caveat. That being, fair dealing is less likely to apply where copyright materials are used with platforms that treat prompts as training data. Therefore use under fair dealing could be limited to platforms with 'closed' / enterprise access such as the University’s access to Microsoft Copilot that includes data security.

Fair dealing for ‘research or study’

Students may be covered under this exception where they use copyright materials with genAI as part of their own study activities. Where students are directed to use copyright materials with genAI as part of coursework or assessments, the educational statutory licence may apply instead, as detailed in the accordion directly above (this distinction is still to be confirmed with the Copyright Agency). Either way, there will be a limit on how much material can be used.

It could be harder to justify certain staff uses under ‘research or study’, as this exception can be limited in terms of how staff would likely need to use/share potential outputs.

Fair dealing for ‘criticism or review’ and ‘parody or satire’

These fair dealing exceptions may be applicable for certain uses with genAI. ‘Criticism or review’ may have teaching or research applications, depending on the potential use. Use under ‘Parody or satire’ may be limited due to the definitions of the terms involved.

Have a question? Contact the Copyright Advisor

The Copyright Act includes access provisions for persons with a disability to provide more equitable access to copyright material:

1. fair dealing for the purpose of access by persons with a disability (S113E), and
2. use of copyright material by organisations assisting persons with a disability (S113F).

These provisions include allowances for persons with a disability to have access to copyright materials in the same way as other users. For application with genAI, this could be using tools to generate alt-text for images or making accessible copies of material.

Dealings under S113E by students or staff for their own use must be ‘fair', which imposes restrictions similar to the fair dealing options included in the accordion above.  The easiest way to be 'fair' in this situation would be to use a 'closed' genAI tool like the University's enterprise access to Microsoft Copilot that does not train its model on user prompts.

While staff assisting persons with a disability under S113F do not need to consider whether the use is 'fair', they will need to satisfy a commercial availability test. S113F cannot be employed where the material is already available in the format required. Where no suitable version is commercially available to meet the needs of the person with a disability, it may be reasonable to apply S113F to generate alt-text or create accessible copies of materials using genAI tools.

For more information about these access provisions, please contact the Copyright Advisor

Materials created by staff and students can be extremely broad in nature. Note that the ownership of these created materials can vary, based on the distinctions made in the University's Intellectual Property Policy.

For articles or other publications you’ve written, any use with genAI will likely be determined by your publishing agreement. You may still need permission from the publisher where the use is not otherwise covered in the options presented in the accordions here.

Where you have retained copyright, or the material is licensed under Creative Commons, you will have more options for reuse. See the accordion on Creative Commons and other open licences below for more.

Note that the University holds ownership and copyright for many materials created by staff, including course materials. See the accordion directly below for more on the use of University-owned materials.

Have a question? Contact the Copyright Advisor

Materials owned by the University are outlined in the Intellectual Property Policy.

Course materials created here are owned by the University and, under University policy, can currently only be used with our data-secure enterprise access to Copilot, unless otherwise approved.

The same is true for other  materials owned by the University - it's important not to share or use any University-owned materials with unsecured, non-endorsed platforms without first gaining approval to do so. This is, in part, due to the issues outlined in the GenAI Basics section above.

Approval for the release of course materials can be provided by the Head of School. Other materials may require approval from the DVC (Research & Innovation) if no other delegations are in place.

Have a question? Contact the Copyright Advisor

As per the University's Intellectual Property Policy, students own the materials they submit for assessment or other projects (unless run by, or signed over to, the University). This includes theses for HDR students.

Student-created materials should not be uploaded into genAI without prior informed consent from the student/s involved, and the use must be limited to endorsed genAI systems.

For guidance on the use of genAI for marking and feedback purposes, refer to the Policy on the Use of Generative AI in Teaching, Learning and Assessment.

AI Tools guide: AI for Staff

Have a question? Contact the Copyright Advisor

You may be able to rely on the Creative Commons (CC) licensing attached to copyright materials for certain uses with genAI tools, as CC licences generally allow for sharing and reproduction.

Some uses with genAI may be limited, depending on the CC licensing involved, for example:

• the ‘NC’ element limits potential commercial use. This may restrict use with some genAI tools, especially where prompts are used to train commercial AI models. It may also limit certain uses of outputs,
• the ‘ND’ element doesn’t allow for the sharing of derivative works / adaptations, and
• the ‘SA’ element requires the same or compatible licensing to be attached to any derivative works.

Also remember that CC-licensed materials incorporated into genAI outputs will require attribution.

Note that Creative Commons works within copyright. Where a copyright exception exists that would allow for the use of copyright materials with genAI, any restrictions associated with a CC licence on the material will not apply. This is because where a user can rely on an exception in legislation, the CC licence will not be triggered by the use. See the flowchart reproduced below and the FAQs for CC licensing for more information. Use under a copyright exception will generally still require attribution so the 'BY' element should always be honoured as best practice.

For materials released under CC0 or flagged with the Public Domain Mark (the CC public domain tools), see the Public domain materials accordion below.

Requirements for other open licensing schemes (such as for software) may vary – check the licensing terms to confirm what uses will be allowed.

Have a question? Contact the Copyright Advisor

AI and CC Licenses v5 flowchart by Creative Commons is licensed under CC BY 4.0

The language 'public domain' can cause confusion as, for copyright, it does not mean materials that are publicly available. Most material found online, for example, is very much still in copyright and may have restrictions around use with genAI. True public domain is specifically where material is not protected by copyright, whether due to age, form, or function.

Where materials are released or flagged under the CC public domain tools (CC0 and the Public Domain Mark) they are free to be copied, distributed, displayed, performed, modified and or otherwise used for any purpose, even without attribution. This makes them perfect for use with genAI tools as the usual restrictions associated with copyright won't apply.

Our guide to public domain tools (PDF) has more information on CC0 dedication and using public domain works.

Looking for public domain materials? Our curated lists on the free/open resources page includes some options which are public domain.

Have a question? Contact the Copyright Advisor

You may also be able to rely on direct permissions from the copyright owner to make use of copyright materials with genAI.

Some online resources, such as many of the sites on our free and open-licensed resources page, already grant bespoke usage licences that are compatible with use of their materials with genAI. Other resources on this page may be released under Creative Commons licences (see the accordion above). The curated PDF guides in each section provide links to the licensing/terms for each resource - it is your responsibility to check for any usage restrictions. Note that free licences for some media types can be limited to particular uses.

Where licensing or usage terms do not cover use of materials with genAI, permission may need to be requested directly from copyright owners (assuming a copyright exception above does not apply). Our PDF guides for teaching use and research purposes provide guidance on how you can approach seeking permission. Where your desired use includes genAI, requests should ask for explicit permission around this, plus check what exactly the copyright owner will allow – don’t assume that a ‘generic’ permission grant will cover all (or any) use with AI.

It is important that you meet any requirements for the allowed use (e.g., providing attribution/credit, or only using a data-secure platform). Best practice is to keep this permission on file for later referral if needed.

See also 'Terms of Service/Use' in the Copyright basics for genAI section above.

Have a question? Contact the Copyright Advisor

It’s openly acknowledged that the entirety of the open internet has already been scraped by AI platforms for use as training data - see the GenAI Basics section above for more on this.

As a result, using a link to openly-accessible material online, or quoting the title for a well-known policy/strategy/theory/etc., may be a better option than uploading a file.

For example, to create a summary of the Australian Government's ‘Closing the Gap’ strategy, it’s possible to simply include the strategy title in a prompt, rather than needing to upload a file.

When adding links to prompts users should still be careful around the terms of service/use for websites, in case use of the site's content  is not permitted with genAI. Consideration should also be made that any online material being added to prompts is not infringing in nature to begin with.

Have a question? Contact the Copyright Advisor

Key copyright considerations for students with genAI:

Key copyright considerations for staff involved with teaching or courses:

Key copyright considerations for researchers and HDR students:

Key copyright considerations for professional staff: