Australia has decided not to create a text-and-data-mining style carve-out for AI developers, with the country’s prime minister publicly branding unlicensed use of copyrighted material to train AI models as theft, according to Tech Times. That framing matters: governments choosing between a US-style fair-use gamble and a licensing-first regime are signaling to the market which side of the ledger AI companies should expect to sit on.
For data companies and AI labs operating in Australia, the practical upshot is that scraping copyrighted text, images, or audio for model training without a license is being treated as an infringement risk rather than a gray area to be litigated later. That pushes model builders toward negotiated licensing deals with publishers, image libraries, and rights holders rather than betting on court-sanctioned fair dealing, and it raises compliance costs for any company training or fine-tuning models on Australian-sourced content.
Rejecting a carve-out doesn’t ban AI training in Australia — it just prices it, and licensing brokers are the ones who stand to collect.
The bigger significance is precedent. Australia joins a small but growing list of jurisdictions declining to hand AI developers a blanket exemption, a stance that contrasts with more permissive interpretations argued for in the US and UK. Watch for how rights-holder groups and publishers in Australia move next — expect pressure for a statutory licensing scheme, and expect AI vendors selling into the Australian market to start disclosing training-data provenance more carefully as this policy line hardens.
Australia Rejects AI Copyright Carve-Out: PM Calls Unlicensed Training Theft
