Written by Julian Watchorn, Vice-Chair, Electronic Frontiers Australia.
On World Intellectual Property Day, we are reminded that copyright is not a barrier to innovation: it is one of the systems that helps make innovation possible.
In 2026, that system is under pressure. As governments accelerate investment in artificial intelligence, there is a growing push to weaken or sidestep copyright protections in favour of training large language models (LLMs). At its core is a familiar but flawed assumption: that faster technological progress requires fewer constraints on how data is accessed and used. It doesn’t.
Australia should absolutely support AI innovation. The real question is what kind of system that innovation will sit within, one that is lawful, fair, and sustainable, or one that depends on the large-scale extraction of creative work without meaningful consent, transparency, or accountability.
For Australian creators and communities, this is not an abstract policy debate. It is already shaping how work is used, valued, and understood in a digital economy increasingly driven by automated systems.
Australia’s current policy direction risks prioritising rapid AI development over the long-term sustainability of its creative and digital ecosystem. That is not a necessary trade-off, but it is a dangerous one.
Copyright is infrastructure—but it must remain balanced
Copyright is often framed in policy debates as a constraint; something that slows down access to data, complicates training processes, or creates compliance burdens. In the context of AI, this framing has become increasingly prominent. That framing doesn’t really hold up.
Copyright is part of the infrastructure that supports creative production, cultural development, and knowledge sharing. It enables creators to participate in the digital economy and provides incentives for ongoing innovation.
At the same time, it has to remain balanced. Copyright should not operate as a blunt tool of control or enforcement, nor should it prevent legitimate access, reuse, or the development of new technologies. Australia’s current copyright framework is widely recognised as inflexible and in need of reform, particularly to better support modern, digital uses of content.
So the challenge isn’t whether to protect copyright or reform it—it’s how to ensure it
continues to serve both creators and the public in a rapidly evolving technological environment. In the context of AI, that balance is starting to shift.
When AI systems are trained on copyrighted material without meaningful transparency, consent, or accountability, the issue is no longer just legal, it becomes one of fairness, power, and control. At scale, this is not simply reuse. It is extraction, where value is transferred from creators to those controlling data, compute, and distribution.
The impacts are already emerging:
- Creators are losing control over how their work is used,
- Large technology companies gain a disproportionate advantage through access to vast datasets,
- Individuals have far less visibility over how their work, data, and identity are incorporated into automated systems.
For example, generative AI systems can now produce content that closely mimics the style of identifiable authors, journalists, and artists, often drawing on training data used without their knowledge or consent. In practice, that’s not a sustainable foundation for innovation.
Moving beyond the false trade-off
A persistent narrative in policy discussions is that Australia must choose between being competitive in AI and maintaining a fair and balanced copyright system.
Weakening safeguards in the hope of accelerating AI development may offer short-term advantages to certain actors, but it also creates longer-term risks:
- Market concentration, where a small number of firms benefit from large-scale data access,
- Ongoing legal uncertainty, as unresolved questions lead to disputes and instability,
- And, ultimately, a loss of public trust in how these systems operate.
Innovation built on unstable or contested foundations tends to be reactive rather than resilient. A better path is to ensure AI development is both innovative and accountable. In that model, access to data is lawful, transparent, and aligned with community expectations.
The limits of Australia’s current approach
Australia’s current trajectory on AI governance reflects a preference for soft regulation, voluntary standards, and institutional signalling rather than enforceable rules. The proposed national AI institute is emblematic of this approach.
While the creation of such a body signals recognition of AI’s importance, there are legitimate concerns about how effective it will be in practice.
Without legislative authority, independence, and enforcement capability, the institute risks becoming a “lame duck” without legislative authority and enforcement capability, unable to meaningfully influence industry behaviour, enforce standards, or respond effectively to harm. What that leaves is governance in appearance, rather than governance in substance.
Key limitations include:
- A lack of binding powers, meaning compliance relies largely on goodwill.
- Unclear accountability mechanisms, particularly for higher-risk or harmful systems.
- Limited ability to address systemic issues, including how data is sourced and used at scale.
At the same time, stronger regulation needs to be designed carefully. Concentrating too much power in a single regulator—without safeguards, transparency, and review mechanisms—can create new problems while trying to solve existing ones.
Good governance requires both capability and constraint.
Learning from the European model
If Australia is serious about building a trusted and sustainable AI ecosystem, it should look to more comprehensive regulatory frameworks, such as the EU AI Act.
The contrast is fairly clear: where Australia is signalling, Europe is regulating.
The European approach is structured, risk-based, and grounded in fundamental rights. It recognises that different AI systems and use cases carry different levels of risk and applies proportionate obligations accordingly.
Key features include:
1. Transparency in data and design
Developers of advanced AI systems are required to provide meaningful information about how models are trained, including the nature and sourcing of training data. This enables proper scrutiny of whether those practices are lawful and fair.
2. Integration with existing legal rights
Rather than bypassing existing frameworks such as copyright and privacy law, the European approach reinforces them. Technological development does not override fundamental rights.
3. Risk-based obligations
Higher-risk systems are subject to stricter requirements, ensuring regulatory attention is focused where harm is most likely.
4. Independent oversight and enforcement
Regulators are equipped with real powers—but those powers are paired with transparency, accountability, and review.
5. Rights and redress
Individuals are given avenues to challenge decisions, seek explanations, and pursue remedies where harm occurs.
Taken together, these elements create a system that supports innovation while maintaining public trust.
Why fairness, consent, and control must be central
At the heart of the AI debate is a broader question: who benefits from these technologies, and under what conditions?
If AI systems are built on large-scale data extraction without meaningful consent or control, they risk reinforcing existing power imbalances. That’s not simply a copyright issue—it’s a digital rights issue.
Privacy, consent, and control over information remain fundamental. Where individuals or creators have no real ability to understand, influence, or refuse how their work or data is used, claims of “consent” become legally and ethically unsustainable. In these conditions, consent functions less as a safeguard and more as a mechanism for legitimising practices that individuals cannot realistically refuse.
A fair system should ensure that:
- Data is not used in ways people cannot reasonably anticipate,
- Consent is genuine, not implied or effectively coerced,
- Power asymmetries between individuals and large organisations are actively addressed.
Without these principles, AI development risks becoming extractive rather than generative.
A path forward for Australia
Australia has an opportunity to take a more balanced and sustainable approach.
This includes:
1. Reforming copyright to support both creators and the public
Copyright law must evolve to reflect modern digital realities, including more flexible mechanisms such as fair use, while ensuring that large-scale, commercial data use is subject to clear standards of fairness and accountability.
2. Embedding transparency and accountability in AI systems
Developers should provide meaningful information about data use, system design, and potential impacts.
3. Moving beyond voluntary governance
Voluntary standards are not sufficient on their own. Clear, enforceable rules are needed—designed to be proportionate and subject to oversight.
4. Establishing independent, accountable regulation
Any regulatory body must have appropriate powers, but also be subject to transparency requirements, independent review, and clear limits.
5. Adopting a risk-based approach
Regulation should focus on systems that pose the greatest risk, rather than applying blunt or overly broad measures.
Conclusion: building systems that deserve trust
World Intellectual Property Day is not just a celebration of creativity—it is a reminder of the systems that enable people to create, share, and participate in the digital world.
As Australia navigates the rise of AI, it should avoid framing copyright, privacy, and digital rights as obstacles to progress. They are not barriers—they are the foundations of a system that people can trust.
The challenge is not simply to innovate, but to do so in a way that is fair, transparent, and accountable. EFA believes that copyright laws must strike a balance between the interests of rights holders, public institution uses of content, and consumer ability to freely engage with works for personal enjoyment, education and creation.
The question is not whether Australia will lead in AI, but whether it will do so on foundations that respect the rights, agency and interests of the people whose data, creativity, and participation make that innovation possible.
Image credit: Unsplash