Anthropic asks Australia to loosen AI training copyright ban

Anthropic and OpenAI are urging Australia to soften its ban on using creative works to train AI models, putting copyright rules on a collision course with the country’s plans to attract data-centre investment. The companies want a narrower, conditional route to training rather than a broad exemption for commercial AI.

Anthropic Watch analysis

What happened

In submissions to an Australian parliamentary inquiry, Anthropic acknowledged that broad copyright exceptions have been ruled out by the government. It proposed conditional approval instead, potentially tied to investment or other support for Australian creators and cultural organisations. OpenAI called for a framework allowing models to learn from publicly available information while giving rights holders opportunities to collaborate.

Reuters reports that Australia is preparing wider AI and data-centre rules for next year. The country has already linked the debate to major infrastructure projects, including OpenAI’s planned Sydney data-centre arrangement with NextDC and Anthropic’s recently announced Queensland partnership. The parliamentary Joint Select Committee on Artificial Intelligence is due to report in November.

Why it matters

This is not just a fight over whether a model may read a book. Australia is trying to decide whether AI investment, local creative rights and access to training data can share the same policy document without one of them being quietly shoved into the boot.

For Anthropic and OpenAI, Australia’s position could limit where they train models and under what conditions. For authors, publishers and other creators, a conditional system could offer leverage, but the value would depend on what counts as fair payment, consent or cultural investment. The submissions are lobbying positions, not a settled compromise.

Our read

The useful development is that Anthropic has moved from opposing a broad exemption to proposing a specific alternative: conditional approval. That gives lawmakers something more concrete to interrogate, but it does not answer the difficult questions about who qualifies, who pays and who gets a veto.

Australia’s November report should reveal whether the government can turn those competing interests into workable rules, rather than producing another elegant framework that collapses when someone asks where the money goes.

What to watch

• The November committee report:
  whether it recommends conditional approval, licensing or a continued ban.
• Creator safeguards:
  whether payment, consent and transparency requirements are made explicit.
• Infrastructure leverage:
  whether data-centre investment changes the government’s copyright position.
• Industry response:
  whether other AI developers support Anthropic’s narrower proposal or seek broader access.

Discussion spark: Should Australia permit conditional AI training on creative works if developers fund or support local creators, or should consent remain the non-negotiable rule?

Anthropic Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by Anthropic.