The case for the bill
The strongest argument for the Online Safety Amendment (Digital Duty of Care) Bill 2026 is that notice-and-takedown has never kept pace with scale. A regime that waits for harm to appear and then removes it will always be late. Australia's own Statutory Review of the Online Safety Act, led by Delia Rickard, reasoned that providers control their systems and must answer for the safety of their services. A duty that makes platforms assess and reduce foreseeable risk, rather than react to individual posts, is a defensible reform. Australia's eSafety Commissioner has lodged a submission to the Senate Environment and Communications Legislation Committee backing the bill's policy intent and setting out proposed enforcement pathways. We could not retrieve that document for this analysis, so we do not characterise its contents beyond that summary.
Where the critics converge
The notable feature of this inquiry is how many supporters of the goal worry about the drafting. The UTS Centre for Media Transition, which backs a single statutory duty of care, called the bill "a positive step" in its 30 October 2026 submission. It then identified the central flaw: a "lack of clarity over whether the duty is aimed at platform systems or at protection from harmful content."
That distinction is not academic. The Centre argues that a duty to ensure a safe online environment, defined by classes of harmful material, "could incentivise platforms to proactively block or filter content, and to block or filter more than is required by the bill in order to assure compliance." It adds that the bill "does not rule out general monitoring" and contains no explicit protections for privacy and freedom of expression. Its recommendations are modest: direct the duty at the contribution of platform systems, require rights impacts to be weighed when assessing what measures are appropriate, and add appeals and an external complaints body.
The Information Technology and Innovation Foundation (ITIF) reaches a similar conclusion from a different starting point. Its 30 October comments say the bill's list of seriously harmful material is specific, but that the category of material "harmful to children" is "far vaguer," reaching any material that "may cause serious harm to children." In ITIF's words, that "leaves companies as the arbiters of what material and conduct is appropriate for children instead of parents." ITIF also warns that services "may overcorrect" and make helpful content on mental health or sexuality harder to reach.
The bill has improved on the exposure draft. ITIF credits the carve-outs for professional news, public-interest material, research and public-health information, and the requirement that Parliament vote to expand the harm definitions. That is real progress, and it shows consultation worked. It does not settle the question.
What the hearing exposed
SBS reported on 9 November that Liberal senator Sarah Henderson told the inquiry: "It seems this is more about shutting down free speech than tackling harm." She also objected that mandatory codes and standards are being replaced with eSafety guidelines that are "not enforceable," and said the compliance load on small businesses is "completely unjustified." We do not need to accept her framing of motive to take the drafting point seriously.
Industry testimony was sharper on certainty. TikTok's Jessika Lofstedt said the bill replaced "certainty with ambiguity and with subjectivity." SBS paraphrased Google and Meta as asking for clearer definitions of regulated recommender systems and of the Commissioner's new powers. The wording under most pressure covers material "harmful to children," content that "encourages or promotes hostile attitudes towards women or gender equality," and content that "encourages dangerous stunts." Each is a reasonable concern in the abstract. None is a legal test a compliance team can apply consistently across billions of items.
Why vagueness is the real cost
A penalty regime changes how vague language gets read. SBS reports fines of up to $109 million for non-compliance. When the downside of under-removal is that large and the standard is subjective, rational operators remove more than the law requires. Lawful speech about health, sexuality, gender and risky-but-legal activity is the first to go. Pinsent Masons notes that the bill moves the standard from "reasonably practicable" to "appropriate and effective", which its commentator says could raise compliance burdens. A less forgiving standard on top of loose definitions compounds the problem.
There is also a competition angle. Large platforms can absorb risk assessments and over-moderate at scale. Smaller services, games and app developers cannot, which is the substance of Henderson's small-business objection. A proportionate duty should not make incumbency a regulatory advantage.
A narrower, stronger bill
The fix is not to abandon the duty of care. It is to make the bill do what its own explanatory materials say it does. Three changes would do most of the work:
- Aim the duty at systems. Define the obligation by reference to the design and governance choices that amplify harm, such as recommender systems and reporting tools, not by a list of content types.
- Write rights into the test. Require providers and the regulator to weigh privacy and freedom of expression when judging whether a measure is appropriate and effective, as the UTS Centre proposes.
- Give users a way back. Provide appeals for wrongful removals and an external complaints body, so over-removal has a cost the system can see.
The Senate committee was due to report by 20 November 2026. Whatever it recommends, the test for Parliament is simple. A bill meant to make platforms safer should be precise enough that a lawful post about a hard subject does not become a compliance risk. On the evidence in the submissions, it is not there yet.