Every AI tool you use, from a chatbot to an automated document summariser, was built by feeding enormous amounts of text into a computer system until it learned patterns in language. That text came from somewhere: books, articles, websites, databases. Most of it was taken without asking, and almost none of it was paid for.
Australia's debate is about whether that practice should be made legal.
In October 2025, Australia's Attorney-General Michelle Rowland formally ruled out what is called a "text and data mining exemption": a legal change that would let AI companies use copyrighted material to train their models without needing permission or paying a fee. The government said it would instead explore a paid licensing model, where AI companies negotiate with content creators to use their work.
That decision was welcomed globally by creative industries. But it did not close the argument.
Anthropics reported plan to build up to 1.4 gigawatts of datacentre capacity in Australia, worth an estimated $150 billion, has reignited the pressure. That scale of capacity is roughly equivalent to the entire Australian datacentre industry as it stood at the end of 2025. Industry sources have told Australian journalists that frontier AI companies view the country's copyright laws as the primary barrier to training their models locally.
A whistleblower tip-off to independent senator David Pocock in late June described a proposal where AI firms would offer at least $50 billion in datacentre investment in exchange for a copyright carveout, sweetened by a fund for Australian creators worth $350 million per year. The government flatly denied the deal. The Australian Financial Review then reported that Anthropic was pushing for exactly such an arrangement as part of a plan to make Australia its second home outside the US. Anthropic declined to comment publicly on the specifics.
The tension inside the Australian government is real. The industry minister and the assistant minister for the digital economy are focused on attracting investment. The attorney-general and the arts minister are focused on protecting creator rights. The prime minister has said creators should be "compensated" if their work is used, but has not committed to a timeline for how that would work in practice.
What makes Australia's position interesting, compared to the rest of the world, is that it has held firmer than most. The EU allows AI companies to train on copyrighted content unless a rights holder explicitly opts out. Japan allows it for AI training outright. The US has no clear law and relies on courts to decide case by case, with the direction of rulings gradually moving in favour of AI companies. The UK proposed a broad exemption in 2022, shelved it after creative industry resistance, then revived a softer version of the idea in late 2024. Australia, by contrast, has twice rejected the exemption and is exploring mandatory paid licensing instead.
There is a real cost to that position. AI companies looking to train models locally need legal certainty. Without it, they can build the physical infrastructure in Australia but train their models elsewhere, in countries with more permissive rules. The power, land, and economic activity from construction still flows into Australia; the intellectual property and the training work does not.
For business operators, the practical takeaway is this: if your company produces any kind of written or visual content, the rules being set right now will determine whether AI companies can use it freely or must pay for it. A paid licensing model, if it actually gets built, would mean content producers may eventually receive some share of the value their work creates for AI systems. A free exemption would mean the opposite: your content, used without consent, at scale, with no return.
The Australian government has until now held its ground. The financial pressure being applied is significant, and the political incentives are pulling in two directions. What is clear is that other governments are watching. The decision Australia makes in the next few months will be cited as a reference point in copyright debates in every country that follows.