Australia's 2021 News Media Bargaining Code had a design flaw that the new incentive now tries to close: a platform that did not want to pay could simply stop carrying news. The News Bargaining Incentive (NMI), enacted on 26 August 2026 according to KPMG's tax summary, supplements that threat of arbitration with a charge. The goal is legitimate. The mechanism raises harder questions than the government has acknowledged.
What the law does
The charge is 2.75% of a group's relevant Australian digital advertising revenue. It applies to social media services with more than 5 million average monthly active Australian users and search services with more than 10 million, where the group's relevant Australian advertising revenue exceeds A$250 million. Per KPMG, the revenue base is measured from the second-most-recent financial year, so the 2026-27 liability looks back to 2024-25 revenue.
A platform can shrink the bill to nil through "eligible expenditure" on news businesses. Spending must reach at least eight different news business groups. Expenditure on small and medium publishers is uplifted to 200% for offset purposes, and on other publishers to 150%. No single news group's offset can exceed 25% of the total charge payable. Unused expenditure can be carried forward, but not from a year that fails the eight-group test. Returns are due even if offsets reduce the liability to zero.
The money that is actually paid flows into the News Journalism Payments scheme. Treasury's media release records that the package was introduced to Parliament on 13 August 2026, and that ministers changed the distribution scheme to better support smaller and diverse media organisations.
The strongest case for the charge
The government's argument deserves a fair statement. The 2021 code worked through the shadow of designation: platforms signed voluntary deals to avoid being brought under it. When Meta let its deals lapse and reduced news on its services, the code had no remedy. The MediaNama report on the passage of the law notes that Meta's deals, reportedly worth about $70 million, were not renewed. A regime in which exit costs nothing is a regime that pays nothing. A charge with a deal-based offset restores the incentive: sign deals or pay the Treasury. Treasury's consultation page frames the scheme as encouraging platforms "to make or renew commercial deals with news media businesses." That is a coherent policy goal, and the offset model is a more market-oriented design than a flat tax because it lets private contracts, not the state, decide who gets paid.
Where the design is weaker
First, the rate moved up through the process. The April draft proposed 2.25%; consultation ran from 28 April to 18 May 2026, and the enacted figure is 2.75%. Some reports at passage still cite 2.5%, a sign of how quickly the number shifted. A rate that climbs by a quarter between draft and law, with consultation submissions only published on 21 August, the day after Parliament passed the bills according to the Treasury consultation page, invites doubt about how evidence-based the calibration was.
Second, the charge is untethered from news. It applies whether or not a platform carries news, and whether or not news drives its advertising revenue. Meta called the earlier proposal "nothing more than a digital services tax" and argued in June that it discriminated against US companies and could breach the Australia-US Free Trade Agreement, per the same MediaNama account. Those are the platform's claims, not findings, and the trade question is untested. But a levy keyed to advertising revenue and a threshold that few firms can cross does invite the charge of being aimed at specific foreign companies.
Third, the offset rewards structure over value. The eight-group minimum and the 25% cap are anti-concentration rules meant to stop a platform from buying off one large publisher. The intent is sound, but a platform now has reason to spread money across enough groups to meet the test, rather than to pay for the journalism its users actually read. The 200% uplift for small publishers pushes further in the same direction. These are subsidy choices dressed up as bargaining. KPMG also notes a transitional rule under which advertising or marketing consideration in certain deals counts only up to 40% of total consideration, a sign that officials expect platforms to pay in kind rather than in cash.
Fourth, AI scope is drawn at a technical seam. Stand-alone AI or LLM answer services without an underlying internet search sit outside the definition of a search service, while a search engine that uses an LLM to summarise results stays in. Two products that deliver the same answer to a user are treated differently depending on whether a conventional index sits behind them. That creates an incentive to re-architect products around the definition, and it is the kind of line that will be litigated.
A more proportionate path
The sensible reading is that the NMI is better than the code it works alongside, but still more brittle than it needs to be. Four fixes would help, none requiring a repeal.
- Publish the methodology behind the 2.75% rate, and commit to a statutory review within three years that measures whether deals grew, not just whether money moved.
- Require disclosure of aggregate offset spending by recipient category, so Parliament can see whether the 200% uplift reached small publishers in practice.
- Revisit the AI definition on a fixed schedule instead of leaving it to case-by-case disputes.
- Keep the charge's trade exposure under independent review, given the claims already made under the Australia-US Free Trade Agreement.
Press freedom and a viable news sector are worth paying for, and a mechanism that makes platforms bargain is preferable to one that makes them leave. But a levy that mixes a revenue tax, a subsidy scheme and an AI-classification rule needs transparent evidence behind each part. Australia has built a more durable complement to its 2021 code. It has not yet shown that it built an efficient one.