A proposal to let AI companies train on Australian copyrighted content after striking a minimum number of commercial deals appears to have been dropped from active government consideration, after it was excluded from a September 17 meeting between the Attorney-General's Department and the country's largest media and creative industry groups8,9.
The move follows weeks of political and creative-sector backlash against two leaked reform options, both of which would have made inclusion in AI training opt-out by default1,4. A coalition of content owners at the September 17 meeting demanded details on timelines, payment values, and opt-out mechanics, saying the proposals heavily favour AI companies.
What the leaked proposals contained
The confidential documents, titled "AI on Australian Terms," were prepared by the Attorney-General's Department and tabled in the Senate by independent Senator David Pocock12. Under the first option, AI companies would gain legal permission to use online material for training after paying a central body. The second would require deals with rights-holder organisations and collecting societies, which would extend access to content created by non-members in the same industry. Under both options, creators would need to "digitally protect" their work to opt out.
The proposals included penalties for ignoring opt-outs, protections for Indigenous Cultural and Intellectual Property, and transparency requirements. "Compensation" and "control" were listed among the document's "key principles," alongside "certainty for all stakeholders, including AI companies".
Anthropic and OpenAI press for conditional access
In submissions to a joint select committee on AI established by both houses of parliament in August, Anthropic and OpenAI acknowledged Australia's October 2025 decision to rule out a text and data mining exception but called for the ban to be relaxed6. Anthropic said "broad exceptions from copyright … have been ruled out by the government" and proposed instead "a narrow form of conditional approval" for model training, saying it was open to investment conditions designed to support Australian creators. OpenAI called for a "balanced framework" enabling models to learn from publicly available information while offering rights holders opportunities to collaborate.
Both companies tied the copyright question to infrastructure commitments. Anthropic was named as a partner for a data centre in Queensland, reportedly leasing part of a $32 billion hub proposed by Zerra DC in the Western Downs5. OpenAI signed an offtake agreement with NEXTDC for a planned data centre in Sydney. Anthropic said its investment in Australia depended on "clarity regarding copyright settings"3.
Australia's public broadcasters, the ABC and SBS, pushed back, calling on the government to enact strict new rules to compensate media organisations and protect public interest journalism. The Joint Select Committee on Artificial Intelligence is due to deliver its report in November. Australia plans broader AI regulations by 2027.
Political and creative opposition
Pocock said the proposals would throw protections for creatives "under the bus"10. Greens Senator Sarah Hanson-Young said AI companies wishing to train on copyrighted material should pay for it14. Former chief competition regulator Rod Sims warned the proposed rules "heavily favour AI companies and risk leaving content creators without payment".
Author Anna Funder said the government had been swayed by "some really false presumptions" put forward by AI companies. Holly Rankin, an artist and executive director of Sentiment Group, argued the opt-out model would give AI companies rights to personal content posted online, from family photos to small business websites15.
Deputy Prime Minister Richard Marles called collaboration with frontier companies an "enormous" economic opportunity11. Industry Minister Tim Ayres denied the proposals would reduce copyright protection. On July 15, 2026, Prime Minister Anthony Albanese said at the University of Sydney that "no company should use Australian books, music, art or news to build or train AI" without the creator's control2.
ANALYSIS The removal of the most permissive option from the September 17 meeting narrows the government's path but does not resolve the core tension: Anthropic and OpenAI have conditioned infrastructure investment on copyright clarity, while creators and public broadcasters are demanding the opposite direction.