When Anna Funder stood up at Australia's Parliament House on Wednesday and called AI training without permission a "copyright crime," she was not pitching a new lawsuit. She was explaining why her country's authors and musicians have stopped trying to win in court and have started trying to win in Canberra instead.
The math behind that pivot is in a single number: roughly $US3,000. That is what an individual Australian author expects to receive from the September 2025 settlement between US AI lab Anthropic and a class of writers whose books were used to train its models. Anthropic's overall settlement was about $US1.5 billion spread across an estimated 500,000 titles, according to NPR's recap of the agreement. For creators based on the other side of the Pacific, the per-book payout does not even cover the cost of flying to a deposition. Funder framed the deal as the model she is now refusing: copyright confirmed, then rendered meaningless by the way damages are distributed.
At the press conference, Funder, children's author Andy Griffiths, and industry figures from APRA AMCOS and ARIA asked the federal government to make Big Tech pay for AI training up front rather than after the fact, as reported in the Canberra Times wire and WA Today. Griffiths said 67 of his books had been scraped without permission, against the 43 he has actually written. Funder reached for a housing metaphor: AI labs are the new tenants in a block of flats she built, and she has been told to take it up with management in another country.
The structural problem is jurisdictional. Under current copyright rules, infringement suits against US-headquartered AI companies effectively have to be brought in the United States. That is also where Australia's own tech industry is now lobbying. Tech Council of Australia chair Scott Farquhar, the Atlassian co-founder, has been pressing US policymakers for what he calls a "fair solution" to the training-data question, a phrase Australian creators read as code for a permissive licensing regime, according to The Conversation's analysis. The same day Australian authors appealed to their own Parliament, the architecture of any deal was being negotiated a hemisphere away.
The federal government has already drawn one line. In October 2025, Attorney-General Michelle Rowland ruled out a text-and-data-mining copyright exception that would have let AI companies train on Australian work without payment, as ABC News reported. APRA AMCOS chief executive Dean Ormston and ARIA's Annabelle Herd, joined by musician Holly Rankin (Jack River), backed that position at the press conference. The lobbying push is therefore aimed at holding that line against future pressure, both domestic and foreign, not at opening a new front.
The creators' calculus now treats the courts as a last resort and Canberra as the only venue left. About 300 commercial AI content licensing deals have already been signed with publishers and labels, including Condé Nast and Warner Music, and roughly 270 active AI copyright lawsuits are grinding through various jurisdictions. The Productivity Commission's interim estimate is that AI could deliver about A$116 billion to the Australian economy over a decade, per ABC News. None of those numbers settles the question of who pays whom. What the Canberra press conference signals is that Australian creators have decided the answer should be set in their own Parliament, on terms they can vote on, rather than in US courtrooms where their presence is at best symbolic.
The next trigger to watch is whether the federal government hardens its October 2025 position into binding rules, and whether any of the roughly 270 active lawsuits produces a settlement with a payout structure more generous than the Anthropic precedent. If neither moves, expect the press conferences to keep coming.