Australia Social Media Algorithm Law Explained

Australia has released draft Digital Duty of Care legislation that would require social-media platforms to give users over 16 a durable choice between personalised recommendations and content from accounts they follow. The proposed Australia social media algorithm law would also impose a broader duty on online services to take reasonable steps against foreseeable harm, with maximum penalties of A$109.2 million.

Everyone else is reporting an algorithm off-switch; we are explaining how feed choice becomes a product, compliance and enforcement problem.

What the Australia social media algorithm law proposes

The government’s “My Feed, My Way” release says services would have to send a notification to new and existing users. A person could opt in to a default feed containing content recommended by an algorithm or choose posts from friends and creators they follow. Communications Minister Anika Wells said the duty would require platforms to respect that decision rather than repeatedly steering a user back to recommendations.

That language makes persistence central. Major platforms already offer variants of “following” or chronological feeds, but the route can be buried in menus, reset between sessions or presented as a secondary view. A statutory choice would be meaningful only if it controls what opens by default, remains selected, and is not undermined by prompts or unequal features.

The government also says the law would prohibit personalised algorithms and endless scrolling for users under 16, according to ABC’s account of the draft. The earlier social-media minimum-age regime remains a separate rule. The new proposal focuses on product design and harm prevention across a wider set of digital services.

A duty of care, not only a feed setting

The most consequential part of the proposal may be its general duty. Online services would need to take reasonable steps to prevent foreseeable harms rather than waiting for a regulator to identify each harmful post or feature. The official release names content promoting eating disorders, misogyny, pornography, crime and dangerous stunts, along with design effects linked to mental-health distress and self-esteem.

Coverage would extend to social-media platforms, online games, apps, messaging platforms and artificial-intelligence chatbots. That range matters for product teams because the relevant risk may come from ranking, autoplay, notifications, streaks, infinite scroll, conversational persuasion or combinations of those features. The service would need evidence that its mitigations match the risks it can reasonably foresee.

In practical terms, Australia’s proposal would shift online-safety compliance from reacting to individual pieces of content toward testing how a service’s design repeatedly exposes users to harm. That is a larger governance obligation than adding a toggle because it reaches risk assessment, experiments, default settings, age assurance and records that can be examined by the eSafety Commissioner.

Who would enforce it and what are the penalties?

The independent eSafety Commissioner would oversee compliance and enforcement. The government states that failure to comply could attract penalties up to A$109.2 million. AP and Reuters both reported that ceiling, while the prime minister described fines of more than A$100 million at the 8 September press conference.

A maximum fine is not the same as an automatic penalty. Final legislation would need to define the duty, regulated services, available defences, investigation powers and how courts calculate sanctions. Companies should therefore treat A$109.2 million as the announced ceiling, not the expected cost of every breach.

Proposal element	Current announced position
Users covered by feed choice	Social-media users over 16
Choices	Personalised recommendations or followed accounts
Under-16 design controls	Personalised algorithms and endless scroll switched off
Broader services	Social media, games, apps, messaging and AI chatbots
Maximum announced penalty	A$109.2 million
Status	Draft released for targeted consultation

Why implementation will decide whether choice is real

A feed can be non-personalised without being strictly chronological, and a chronological feed still uses rules for spam, safety and availability. The government release frames the alternative around friends and chosen creators rather than spelling out every ordering rule. The final text will need to distinguish permitted integrity ranking from engagement optimisation.

Persistence is another fault line. If a user must reselect the following feed every time an app opens, the formal choice offers little control. If the alternative feed loses search, notifications, creator discovery or accessibility functions, platforms could make it unattractive without technically removing it. Regulators will need outcome tests as well as interface requirements.

Consent design creates a third issue. A neutral prompt should explain both options in comparable language and avoid preselected boxes, colour bias or extra clicks. Logs must demonstrate that the preference was stored without turning the choice itself into a new source of sensitive profiling.

What platforms and startups should prepare

Large social platforms are the obvious target, but suppliers of recommender systems, age-assurance tools, content moderation and safety analytics may also feel the impact. Product teams should map every place recommendations enter the experience, document the objective function used for ranking, and separate followed-account inventory from suggested content.

Safety teams will need age-sensitive controls and escalation paths. Legal teams will need to connect product tests to a defensible interpretation of “reasonable steps.” Data teams should prepare audit trails showing prompt delivery, user selection, persistence and changes without retaining more personal information than necessary.

For AI chatbots, the feed-choice rule may not apply in the same form, but the wider duty could reach addictive design or harmful conversational outputs involving children. That makes the policy relevant to connected workplace and consumer systems such as ChatGPT Work’s connected-app features. Similarly, device products such as Meta’s AI glasses show how recommendations and assistants are moving beyond a conventional phone feed.

What the proposal does not settle

The release does not resolve which platforms qualify, how small services are treated, whether messaging recommendations fall under the social-feed choice, or how overseas companies must implement Australia-specific defaults. It also does not settle the boundary between harmful content, lawful expression and platform moderation.

Reuters reported that free-speech advocates warned a duty of care could encourage censorship. The policy answer will depend on whether enforcement is tied to transparent risk processes and defined harms rather than pressure to suppress broad categories of lawful speech. The consultation should also test appeal rights and independent scrutiny.

The draft arrives after Australia’s under-16 social-media account restrictions took effect. The government said in June that more than five million accounts had been removed, deactivated or restricted, but that figure comes from the government and does not by itself measure accuracy, circumvention or harm reduction.

India relevance: a design precedent worth watching

Australia’s experiment could influence debates in India because recommendation systems shape news exposure, commerce, entertainment and youth safety across the same global platforms. An Indian policy response would need its own constitutional, competition, privacy and child-safety analysis; Australia’s model cannot simply be copied.

The useful precedent is narrower: regulation can define user control as a persistent product outcome rather than the mere existence of a menu option. Indian startups serving global customers may need architecture that supports country-specific defaults, explainable recommendation modes and verifiable consent. Building those controls early is cheaper than splitting a mature product after enforcement begins.

Frequently asked questions

Has Australia banned social-media algorithms?

No. The proposal would let users over 16 choose whether personalised recommendations appear in their default feed. It would switch off personalised algorithms and endless scrolling for under-16 users under the announced design.

Is the Australia social media algorithm law already in force?

No. The government released draft legislation for targeted consultation on 8 September 2026. Details may change before a bill is introduced, passed and commenced.

What is the proposed maximum fine?

The government says a breach of the Digital Duty of Care could attract a maximum penalty of A$109.2 million. Actual penalties would depend on final legislation and enforcement.

Does the proposal cover AI chatbots?

The broader duty of care is intended to include AI chatbots, apps, games and messaging platforms where children can face harmful content or design features. The social-feed choice is described specifically for social-media users.

Sources: Australian government release; Prime minister’s press conference; Associated Press; Reuters; ABC News.