Anthropic asks Australia for opt-out AI copyright model as ABC objects

Anthropic has urged the Albanese government to consider giving "conditional approval" for big tech companies to train their AI models on Australian copyrighted works under an opt-out model, after conceding that it will not secure a blanket copyright exemption.

The maker of the Claude AI models made the proposal in submissions to the Australian government's ongoing policy process on artificial intelligence and copyright. The company argues that AI could transform the economy, according to its submission, but it has accepted political and legal reality: a full exemption from copyright law, of the kind it previously advocated, is not on the table in Australia.

Under the opt-out model Anthropic now favors, rightsholders would by default see their works used for AI training unless they actively opted out — a reversal of the standard copyright bargain, in which creators control use unless they license it. "Conditional approval" would leave the practice lawful subject to conditions the government might impose.

Australia's public broadcasters have pushed back hard. The ABC and SBS have strongly criticised AI firms in their own submissions, calling on the government to enact strict new rules to compensate media organisations and to protect public interest journalism. The ABC has warned of what it describes as the "cannibalisation" of news: AI systems that ingest journalistic content, then serve synthesized answers to readers who never visit the original publisher.

The confrontation matters beyond Australia's borders. Anthropic, along with other major AI developers, faces a wave of copyright litigation and regulatory scrutiny in the United States and elsewhere over the use of scraped creative and journalistic material in model training. How a mid-sized, English-speaking democracy like Australia resolves the question — whether through opt-out schemes, licensing mandates, or compensation funds — will be watched closely by publishers and AI companies operating in larger markets.

For the ABC and SBS, the stakes are institutional. Public broadcasters funded by Australian taxpayers argue that AI firms extracting value from their archives and reporting without payment undermines the business and civic function of journalism. Their call for strict rules aligns them with commercial publishers worldwide who have demanded that AI companies negotiate licensing deals rather than train on content by default.

For Anthropic, the opt-out proposal is a pragmatic retreat. The company has shifted from seeking broad protection for training on copyrighted material to asking for a regulatory framework that would make such training lawful under defined conditions, with an opt-out mechanism for rightsholders who object.

The Albanese government now has to weigh those competing positions. A decision to adopt an opt-out model would please AI developers and frustrate publishers; a decision to mandate compensation or licensing would do the reverse. The outcome of Australia's process will set an early precedent for how democratic governments balance AI industry claims of economic transformation against media organisations' warnings about the erosion of the news ecosystem.

The ABC's "cannibalisation" warning and Anthropic's "conditional approval" pitch now sit before ministers as the competing frameworks. Which one the government adopts will shape how Australian creative and journalistic works are used — and paid for — in the training of the next generation of AI models.