- Atlassian co-founder Scott Farquhar says AI copyright reform is widely misunderstood and urgently needed in Australia.
- Farquhar argues that without AI copyright reform, Australian developers will fall behind global competitors building on freer legal ground.
- The debate centres on whether AI companies should be allowed to train models on copyrighted material without direct licensing fees.
- Australia’s response to this issue could set a precedent for how Asia-Pacific nations regulate generative AI development.
- Atlassian co-founder Scott Farquhar says AI copyright reform is widely misunderstood and urgently needed in Australia.
- Farquhar argues that without AI copyright reform, Australian developers will fall behind global competitors building on freer legal ground.
- The debate centres on whether AI companies should be allowed to train models on copyrighted material without direct licensing fees.
- Australia’s response to this issue could set a precedent for how Asia-Pacific nations regulate generative AI development.
AI Copyright Reform and the Voice Nobody Expected
AI copyright reform has an unlikely champion in Australia right now — and it’s not a lobbyist, a lawyer, or a politician. It’s Scott Farquhar, the billionaire co-founder of Atlassian, the enterprise software giant behind Jira and Confluence. Farquhar has stepped into one of the messiest, most emotionally charged debates in tech policy today, and his message is blunt: almost everyone is getting this wrong.
Speaking publicly on the issue, Farquhar described the current conversation around AI and copyright as deeply misunderstood — a debate that has collapsed into opposing camps of artists fearing for their livelihoods and tech optimists promising utopia, with very little serious policy thinking happening in between. He’s calling for Australia to undertake meaningful legal reform that clarifies the rules around how AI systems can be trained on existing creative works.
It’s a position that takes real nerve to stake out in public. The copyright debate around AI is politically toxic. Creators — writers, illustrators, musicians, filmmakers — are furious that their work is being scraped and used to train models without compensation or consent. Tech companies, on the other hand, argue that AI training is transformative use, no different in legal spirit from a human reading widely and then creating something new. Neither camp is entirely wrong, which is precisely why the policy gap keeps widening.
Why Farquhar Thinks Australia Is Falling Behind
Farquhar’s concern isn’t abstract. Australia is trying to position itself as a serious player in the global AI economy, and right now its legal framework is working against that ambition. While the United States has the doctrine of fair use — a flexible, judicially interpreted standard that gives American AI companies significant room to train on publicly available data — Australia has no equivalent. Its copyright framework is older, more rigid, and hasn’t been meaningfully updated to account for machine learning at scale.
That asymmetry matters enormously. When an Australian AI startup wants to train a model, it faces a legal landscape that’s far more hostile than what its American or British competitors deal with. Australia’s Attorney-General’s Department has acknowledged that the existing Copyright Act was never designed with AI in mind, but legislative movement has been slow. Meanwhile, companies like OpenAI, Google DeepMind, and Anthropic are building at speed under legal regimes that, while contested, at least offer more clarity.
Farquhar’s point is that this isn’t just a technical legal question — it’s an economic competitiveness question. If Australia’s AI sector can’t train on the same breadth of data that overseas competitors can, the country risks locking itself into a permanent second-tier position. That’s bad for startups, bad for enterprise software companies like Atlassian that increasingly rely on AI features, and bad for the broader tech ecosystem.
What AI Copyright Reform Actually Means
Here’s where the debate gets genuinely complicated, and where Farquhar’s call for clarity is most valuable. AI copyright reform doesn’t mean simply declaring open season on creative work. What reformers are generally pushing for — and what several countries are already experimenting with — is a form of text and data mining exception. Japan, Singapore, and the EU have all introduced or considered specific carve-outs that allow AI training on copyrighted material under defined conditions, sometimes with compensation mechanisms attached.
The idea isn’t that creators should get nothing. It’s that the current all-or-nothing framing — where every piece of training data either requires an individual licence or can’t be used at all — is practically unworkable and legally incoherent. No AI company can licence billions of individual pieces of content. And if they could, the cost would make training large foundation models the exclusive domain of trillion-dollar corporations, which is arguably worse for competition than the current situation.
What thoughtful AI copyright reform looks like in practice is probably something like a statutory licence — a legally mandated right to train on copyrighted data, paired with a collective compensation system that channels money back to creators through royalty pools. It’s not a perfect solution, but it’s a functioning one. Germany and France have explored similar models for music streaming, and the logic translates reasonably well to AI.
The Creative Industries Aren’t Wrong to Be Angry
It would be easy to read Farquhar’s position as tech-industry boosterism dressed up as policy pragmatism. But dismissing the concerns of creators entirely would be a mistake, and to his credit, that’s not what Farquhar appears to be doing. The grievance from writers and artists is real: their work was used, without permission and without payment, to build commercial products that now compete with them directly. That’s not an abstract philosophical complaint. That’s a concrete economic harm.
The question isn’t whether creators deserve recognition and compensation — they clearly do. The question is what mechanism delivers that most effectively. Right now, the legal system isn’t delivering anything for most creators. Litigation is slow, expensive, and uncertain. Class action lawsuits against OpenAI, Stability AI, and others are working their way through US courts, but the outcomes will take years to materialise and will likely be shaped by US-specific law in ways that don’t transfer cleanly to Australia.
A properly structured AI copyright reform framework in Australia could actually give creators more reliable recourse than the current patchwork of uncertain legal claims. That’s not a consolation prize — it’s potentially a better outcome.
Scott Farquhar and the Broader Industry Responsibility Question
There’s a bigger question lurking behind Farquhar’s intervention, and it’s worth sitting with for a moment. Why is a tech billionaire the one making this argument? Where are Australia’s policymakers, its cultural institutions, its universities? The fact that the most prominent voice calling for AI copyright reform right now comes from the co-founder of a software company says something uncomfortable about the state of Australian AI governance.
Farquhar isn’t a disinterested party here. Atlassian builds products that use AI, and clearer copyright rules benefit companies like his. That’s fine — having skin in the game doesn’t make your argument wrong. But the vacuum he’s stepping into is concerning. Australia’s AI strategy documents are full of ambition and short on mechanism. The government has been consulting on AI regulation broadly, but the copyright question — which is arguably the most immediately consequential legal issue in the AI space — has received less attention than it deserves.
Other jurisdictions are moving. The UK’s Intellectual Property Office ran extensive consultations on AI and copyright before the current government paused the process amid fierce creative industry pushback. The EU’s AI Act touches on transparency obligations for training data but stops short of resolving the underlying copyright question. The US is waiting to see how the courts land before Congress acts. In that context, Australia actually has an opportunity — not just to follow, but to design something better than what other countries have managed so far.
What Comes Next
Farquhar’s public advocacy probably won’t immediately shift Australia’s legislative calendar. But it raises the profile of AI copyright reform in a country where tech founders have genuine cultural credibility, and it signals that the business community — not just Silicon Valley-adjacent voices — sees this as urgent.
The most productive outcome from this debate isn’t a binary win for either tech companies or creative industries. It’s a legal framework that actually works: one that gives AI developers the certainty they need to build, gives creators a genuine slice of the value their work helps generate, and gives Australia a competitive footing in a technology race that’s already running at full speed. Whether Canberra has the appetite to deliver that is the real open question — and the answer will say a lot about how seriously Australia takes its AI ambitions.
