On August 27, Australia’s News Media Bargaining Charge Act, passed by the Federal Parliament a week before, came into force. The regulations require specific social media or search services earning more than A$250 million per year from relevant Australian digital advertising revenue—that is, effectively, Google and Meta—to pay a 2.75 percent charge calculated on the statutory revenue base. The Australian government will then pass the revenues on to selected Australian news media content providers to support public interest journalism.

The targeted firms can reduce their liability by 200 percent of any money they give voluntarily to small and medium qualifying Australian news media publishers, and by 150 percent of sums paid to large operators. The over-crediting is deliberate: The mechanism is calibrated so that at the margin and within the statutory eligibility and offset limits it is cheaper for platforms to fund journalism directly than to remit the statutory charge to the government.

The act is the successor to the infamous 2021 News Media Bargaining Code, which enabled the government to impose a mandatory arbitration process if Google and Meta failed to come to “voluntary” agreements to compensate Australian news media companies for access to their online content. While non-permissioned use of copyrighted content was the superficial justification for the law, its real purpose was to compensate Australian media companies for the loss of advertising revenues that had traditionally been used to support the creation of news content. The loss of these revenues and Australian-generated media content was held to be of significant concern for the functioning of the country’s democratic institutions, under the view of journalism as a public good. The subterfuge of the copyright claim was exposed: If payment for use of all unsanctioned Australian copyrighted material was the objective, then why limit the compensation agreements to deals with only news media companies?

Initially, no mandatory arbitration processes were needed, as Google and Meta came to “voluntary” agreements with the relevant Australian media firms. The first contracts lasted for three years, but as they came up for renegotiation, the arrangements started to unravel. Meta announced it would not be renewing its deals—a credible threat, as it had subsequently refused to participate in a similar set of arrangements in Canada. With Australian media firm revenues—and the nation’s democracy—under threat, a new mechanism was required. The result is the News Media Bargaining Charge Act.

The act broadens the scope of the previous arrangements, extending it to all companies meeting the statutory digital advertising revenue threshold. It maintains a preference for “voluntary” agreements by way of the generous rebates against the liability when such payments are made. It also explicitly assists smaller companies—either by way of voluntary payments or government-distributed funds—as the bulk of the monies under the earlier agreements had been paid to large national and state-level entities rather than small local or regional publishers.

But the issue of large international internet platforms compensating Australian copyright holders for use of their copyrighted material has not gone away. Indeed, it has returned with a vengeance under Prime Minister Anthony Albanese’s National AI Plan and associated measures, announced in July. Legally, because unlike the US Australia does not have a fair use doctrine, using Australian copyrighted material to train AI models requires a license unless a narrow statutory exception applies. The application of those exceptions to commercial, large-scale model training remains legally uncertain and untested in definitive Australian case law.

In practice, however, there have been few impediments to powerful international AI firms scraping Australian digital content and using it in their training processes, just as the internet platforms utilized news media content. Indeed, if the AI models are to be representative of relevant Australian content and useful for Australian users, then they must include that content in their training. But when announcing the AI plan, Albanese exhorted that creators of books, music, art, and news “should retain control of the price and value of their work” when used to train AI, declaring that “anything less is theft.” This is consistent with the Australian government’s position, stated in October 2025, that it would not include a text and data mining exception in its copyright legislation.

The specter that a statutory or compulsory remuneration regime would be created to address the issue took tangible form when a leaked document indicated the government was considering copyright reforms allowing AI companies to train their models on online Australian content without paying its creators, as long as they struck deals with enough rights holders. Understandably, the AI firms are strongly in favor of relaxing Australian copyright laws so that the access to material used in training their models is legalized. But opponents claim the government is planning to “throw creatives under the bus.”

 “Here We Go Again,” anyone?