What the bill does
Communications Minister Anika Wells has brought the Digital Duty of Care bill to Parliament, revised from the September exposure draft. It would cover social media, search engines, AI chatbots, dating apps, gaming, messaging and pornography sites (The Nightly). Penalties for non-compliance can exceed A$100 million, and the eSafety Commissioner gets wider powers. If the bill passes, companies would get a 12-month transition period (afaqs).
The bill follows the independent review of the Online Safety Act, led by Delia Rickard and delivered in October 2024. That review made 67 recommendations and put a technology-neutral digital duty of care at the centre (Minister's release on the review). The government responded on 14 April 2026, saying it would implement, partly implement or further consider 64 of the 67 (Government response). So this is the end of a two-year process, not a snap decision.
The strongest case for it
The case for the bill is a serious one. Australia's current rules mostly work after the fact. A victim complains, a notice goes out, and the content may already have spread. A duty of care moves the burden onto the service, which is the party that designs the recommender system and knows what it is optimised for. Wells has said the bill sets "basic safety standards" and shifts the burden of harm from victims to companies (The Nightly, 8 September). Her closest description of the harm is content that "causes serious distress", including cyberbullying and a chatbot coaching a teenager on self-harm (Cyber Daily). Nobody should defend that outcome.
What the revision fixes
The September draft let the minister decide which services must comply and which features count as harmful, by legislative instrument. Coalition shadow minister Sarah Henderson said this gave Wells "unfettered powers to make regulations" (The Nightly, 8 September). The revised bill removes the provisions that let ministers decide what counts as "serious harm", and Parliament now decides (The Nightly).
This matters for speech. A definition of harm that shifts with whoever holds the portfolio is a standing risk to political expression. Under the new design, the contested lines are drawn in public, by elected legislators, in a law that can be challenged. It is the most important change in the revision, and a reasonable one.
'My Feed, My Way' is the best part
The feed provision requires a prominent opt-out from algorithmic recommendation. The choice is offered each time the app opens, platforms may not make the app harder to use after a user opts out, and the opt-out feed shows only accounts the user chose to follow (afaqs).
This is regulation by user choice rather than by content censorship. It does not ask any platform to judge whether a post is harmful. It also stops short of the non-algorithmic default that the Greens and some safety advocates wanted (The Nightly, 8 September). A default would have overridden what most users seem to prefer. An opt-out leaves that decision with them. If other jurisdictions want a model for tackling engagement-maximising design without policing speech, this is it.
Where the risk remains
The harder problem is the general duty itself. Seven very different service types fall under one framework: a search engine, a dating app, a game lobby, a private messenger and an AI chatbot have little in common in how harm arises or how it can be reduced. A duty to prevent harm that is only vaguely defined invites over-removal. A platform facing a fine above A$100 million has every reason to remove lawful but controversial speech, restrict features and end up with a worse product. Henderson called the bill "dangerously vague" and a threat to political speech (The Nightly). The Coalition's motives are partisan, but the vagueness charge does not depend on them.
Messaging and AI are the sharpest cases. Duties applied to private messaging raise questions about whether platforms would need to weaken or scan encrypted communications. AI chatbot developers, including Australian startups, would face compliance costs that incumbents can absorb more easily than new entrants.
The Greens are pushing in the opposite direction. They want tougher regulation, fines tied to a percentage of global revenue, and a private right of action letting individuals sue platforms (The Nightly). Revenue-based fines can be a fair deterrent for the largest firms. But a private right of action attached to an undefined duty would turn litigation into the main enforcement tool, and legal uncertainty would fall hardest on smaller services.
What the Senate inquiry should test
The bill is expected to go to a roughly five-week Senate inquiry, and its fate probably depends on the Greens (The Nightly). Four questions deserve answers on the record:
- Proportionality: Will obligations scale with a service's size and risk, or will one standard apply to all seven sectors?
- Definition: How narrowly will Parliament define "serious harm", and will it include an explicit protection for lawful political speech?
- Oversight: What appeal and transparency rights will apply to eSafety's expanded powers?
- Encryption: Does the duty create any obligation to scan or weaken private messages?
The revision shows the government will move when pushed on ministerial discretion. The same scrutiny should now reach the duty's breadth. A narrower bill, with feed choice at its centre, clearer definitions and proportionate penalties, would protect users better and cost innovation less.