On September 22, 2026 — the closing day of public consultation — the US Embassy in Canberra filed a formal submission attacking Australia's exposure draft of the Online Safety Amendment (Digital Duty of Care) Bill 2026. The embassy warned the bill would enable "viewpoint-based censorship," objected that mandatory algorithm changes could reshape "what users see, say, and hear not just in or from Australia, but globally," and asked Canberra to clarify how it will define "harm" and "risk" before pressing ahead — while separately pushing for US platforms to be carved out of the regime entirely.
Prime Minister Anthony Albanese rejected the exemption request the next day, telling reporters the reform "is not about government control, it's about giving people back control over what they receive on their devices," and pledged the government would "engage constructively" with industry before introducing the bill to Parliament.
What the bill actually does
The exposure draft, released September 8, 2026, would repeal the Online Safety Act 2021's Basic Online Safety Expectations and the co-regulatory industry codes and standards, replacing them with a single statutory duty: providers must ensure, "so far as is reasonably practicable," a safe online environment. The duty rests on three pillars — annual written risk assessments kept for six years, mitigation measures against identified risks, and design obligations including the government's "My Feed, My Way" initiative, which requires platforms to let users over 16 choose a chronological feed over an algorithmically personalised one. Civil penalties for breach reach roughly A$109.2 million, enforced by the eSafety Commissioner — the same regulator that currently administers the BOSE determinations and the social media age-restriction rules under Part 4A of the Act.
The case for the bill
The strongest argument for a duty of care is that Australia's current model — codes negotiated with industry, backed by BOSE expectations with no real teeth — has been slow and reactive. A statutory duty forces platforms to build safety into design decisions before harm occurs, rather than responding to individual pieces of content after the fact. That's a meaningfully different (and arguably more proportionate) intervention than the content-takedown regimes most safety laws rely on: giving a 17-year-old the option to turn off an engagement-optimised feed is a structural fix, not a speech restriction. Australia isn't inventing this approach in isolation either — it sits alongside the EU's Digital Services Act and the UK's Online Safety Act 2023 in treating algorithmic amplification, not just individual posts, as the regulatory target.
Where the objection has merit
The US Embassy's core complaint — that "reasonably practicable" and "foreseeable harm" are undefined — is a real drafting problem, not a talking point. A platform facing a $109 million penalty for guessing wrong on an undefined harm threshold will rationally over-remove and over-filter to stay safe, which is exactly the chilling dynamic the embassy describes. That risk doesn't require bad faith from the eSafety Commissioner; it's a predictable response to penalty exposure paired with regulatory ambiguity, and Australia's own Online Safety Act already shows what loosely bounded takedown powers can produce in practice. The extraterritoriality point is also not manufactured: because platforms build features at global scale rather than per-jurisdiction, a design mandate calibrated for 27 million Australian users can plausibly reshape defaults for users who never fall under Australian law — the same dynamic that made Australia's under-16 social media ban a template other governments are now examining.
Where it doesn't
The request to exempt US platforms specifically is a different claim, and a weaker one. A duty of care that applies to Meta, X and Google but not to TikTok or a domestic Australian service isn't more speech-protective — it's a carve-out for the largest incumbents, precisely the companies with the most algorithmic reach to regulate. Equal treatment under a badly drafted law is a reason to fix the drafting, not a reason to exempt the biggest players from it. And Washington's own domestic record on this question — including ongoing federal and state fights over the Kids Online Safety Act and algorithmic-feed regulation for minors — makes the blanket claim that any harm-based design duty is inherently censorship difficult to sustain evenhandedly.
The proportionate path
The algorithm opt-out itself is close to the right instrument: it changes a default rather than mandating removal of any content, which is the least speech-restrictive way to address amplification harms. What Parliament should fix before introduction is the vagueness the embassy correctly flagged — bounded, published definitions of "foreseeable harm" and "reasonably practicable," ideally with a safe-harbor process so platforms aren't penalized for good-faith judgment calls under an undefined standard. That would blunt both the over-removal risk and the extraterritorial spillover, without carving out an exemption for whichever platforms happen to be headquartered in the country objecting loudest.