Australians could soon be given a direct choice over whether powerful social media algorithms decide what appears in their feeds under sweeping new online safety laws unveiled by the Albanese government.

The federal government released its long-awaited exposure draft of the Online Safety Amendment (Digital Duty of Care) Bill 2026 on Tuesday, promising what Communications Minister Anika Wells described as a “global reckoning” for Big Tech.

The proposal would impose a broad duty on online services to identify foreseeable risks and take reasonable steps to reduce them, shifting Australia’s online safety system away from responding only after harm occurs.

But the feature likely to be most immediately visible to ordinary Australians is a new initiative called “My Feed, My Way”.

Social media users aged 16 and over would have to be offered a genuine choice about their default feed.

They could continue using personalised recommendations generated by a platform’s algorithm or choose a feed based on friends, accounts and creators they have deliberately decided to follow.

Prime Minister Anthony Albanese said the reform was about transferring power from technology companies to users rather than giving government officials control over what Australians see.

Under the proposal, platforms would be expected to present users with a clear choice and respect that decision rather than forcing people permanently into a recommendation system designed to maximise engagement.

Users would also be able to change their selection later.

The government says Australians who enjoy algorithmic recommendations will be free to keep them.

Those who would rather avoid an endless stream of content selected on their behalf would have an alternative.

Why algorithms are in Canberra’s sights

Recommendation algorithms are at the centre of the modern social media economy.

They analyse behaviour including what users watch, click, share, search for and spend time looking at before predicting what is most likely to keep them engaged.

That can be useful.

Algorithms help people discover music, businesses, creators, news and communities they may never have found themselves.

But governments and online safety advocates are increasingly concerned that the same systems can repeatedly direct people towards more extreme, addictive or emotionally provocative material because outrage and distress can generate engagement just as effectively as entertainment.

Australia’s eSafety Commissioner has warned that recommender systems can shape exposure to harmful material and says online safety regulation needs to address how platforms themselves are designed rather than focusing only on individual pieces of content.

The proposed Digital Duty of Care takes that systems-based approach.

Wells said the basic requirement would be for online services to identify risks associated with their platforms and mitigate those risks through safety-by-design measures.

The government’s argument is that technology companies should face basic safety responsibilities similar in principle to companies producing cars, food or consumer products.

Platforms would still be allowed to innovate, but they would be expected to consider foreseeable consequences before harmful features are deployed at enormous scale.

The rules go far beyond social media

The Digital Duty of Care is broader than the government’s existing social media minimum-age regime.

Its reach is intended to extend into online games, messaging services, apps, websites and artificial intelligence products including chatbots.

For children, the government wants platforms to proactively deal with several categories of serious harm.

Examples highlighted by Albanese and Wells include material promoting eating disorders, pornography, misogynistic content, glorification of crime or dangerous stunts, abuse and cyberbullying, and content capable of causing serious psychological distress.

One example given by Wells was an AI chatbot advising a teenage girl how to survive on an extremely low calorie intake.

Another involved a chatbot encouraging a young person to self-harm.

The government has also pointed to online content encouraging car theft and dangerous behaviour, as well as hostile material aimed at women and girls.

The draft would expand eSafety’s enforcement role and give the regulator powers relating to so-called “nudify” services that use artificial intelligence to create fake sexualised images of real people.

Platforms would also be required to document the steps they take to identify and reduce risks and demonstrate that their protections remain effective over time.

Adults and children would be treated differently

One of the most politically important aspects of the draft is the distinction the government says it is drawing between protections for children and restrictions affecting adults.

For adults, Wells says the regime is principally about existing illegal conduct and individual choice.

She identified child sexual exploitation, child abuse material, grooming, serious violence and sexual violence, threatening or menacing behaviour, harassment and promotion of terrorism and extremism among the illegal harms already dealt with under Australian law.

For children, the proposed duty is broader because lawful material may still pose a serious risk to a young person’s development or mental health.

That distinction lies at the centre of the government’s answer to accusations that the bill could become an internet censorship regime.

When Albanese and Wells were directly asked whether the proposal could allow eSafety or another regulator to police misinformation, political statements or commentary, the Prime Minister said the law was not about censorship.

Wells went further, saying the government could “emphatically rule that out”.

For adult Australians outside existing illegal content rules, she said the emphasis would be on choice.

Why free speech fears have not disappeared

Those assurances have not ended the argument.

Coalition figures have raised concerns about provisions allowing the government to respond to emerging categories of online harm through rule-making powers.

Liberal MP Zoe McKenzie said she was uncomfortable with the possibility of a minister gaining excessive influence over what can be said online.

Shadow communications figures have similarly argued that broad definitions of harm could create a risk of mission creep if future governments or regulators interpret them more aggressively than Labor currently promises.

Nationals leader Matt Canavan has accused the government of seeking censorship powers and argued politicians should not be involved in reshaping Australians’ online information environment.

The criticism is significant because the debate is not simply about whether violent or exploitative content should be removed.

Few political parties dispute the need to tackle child sexual abuse material, serious threats or other clearly illegal behaviour.

The dispute is over how much discretion regulators and ministers should receive when dealing with technologies and forms of harm that may not yet exist.

Wells argues flexibility is necessary precisely because technology changes faster than legislation.

She has pointed to the sudden growth of AI-generated sexualised imagery as an example of an online harm capable of emerging far more quickly than Parliament can pass an entirely new regulatory regime.

The government also notes that ministerial rule-making instruments can be subject to parliamentary disallowance.

Critics are likely to demand tighter safeguards before the bill reaches its final form.

The Greens say Labor has not gone far enough

While conservative parties are warning the proposal may reach too far, the Greens are attacking it from the opposite direction.

Greens communications spokesperson Sarah Hanson-Young argues social media recommendation systems should operate on a stronger opt-in basis rather than allowing platforms to design an experience that might make avoiding algorithmic recommendations difficult.

The concern is based partly on overseas experience.

In Europe, technology companies have been accused of “malicious compliance” — technically supplying alternatives to algorithmic feeds while making those options less attractive, more cumbersome or less functional than the default product.

Hanson-Young has also questioned whether penalties of around $100 million are enough to influence companies with global revenues measured in tens or hundreds of billions of dollars.

The Greens favour penalties linked more directly to global company revenue and tougher restrictions on addictive platform design.

That means Labor may eventually find itself negotiating between two competing criticisms: one side worried about censorship and regulatory overreach, and another demanding significantly stronger regulation.

Penalties could exceed $100 million

The government is promising substantial financial consequences for companies that ignore the new obligations.

Albanese said breaches could attract penalties exceeding $100 million.

Wells said companies committing overlapping violations involving the social media minimum-age regime and the Digital Duty of Care could potentially face penalties exceeding $200 million per offence.

That is intended to address a longstanding problem in technology regulation: penalties that appear enormous to individuals can be treated as little more than a cost of doing business by the world’s richest companies.

Whether even nine-figure fines are enough to change platform behaviour remains an open question.

Technology industry group DIGI, whose members include major digital companies, has said recommendation systems can help users discover diverse and useful content but acknowledged community concerns about how algorithms operate.

The group says it supports giving users meaningful choice and control while managing safety risks.

The details of implementation are likely to determine whether that broad agreement survives once platforms know exactly what they will be required to change.

Australia is becoming a test case for Big Tech regulation

The Digital Duty of Care follows Australia’s world-first social media minimum-age laws, which prevent children under 16 from maintaining accounts on covered platforms.

Albanese said more than five million accounts had already been removed or deactivated under that regime and claimed more than 20 countries were taking similar steps on youth social media regulation.

The new proposal moves beyond the question of who can access social media and towards a more fundamental issue: how digital products themselves should operate once a person is online.

Other jurisdictions are pursuing related approaches.

The European Union’s Digital Services Act has already imposed obligations around recommender systems and platform accountability, while governments in Britain and elsewhere are expanding regulation of online safety.

Australia’s model would nevertheless be distinctive in requiring an explicit and continuing choice over default social media feeds.

It could also create tension with major American technology companies and the United States government if Australian penalties are viewed as targeting US businesses.

Albanese said Australia would make its own decisions according to the national interest and would not allow concerns about overseas political reaction to determine domestic policy.

This is not law yet

The government’s announcement is an exposure draft, not legislation already in force.

The Department of Infrastructure, Transport, Regional Development, Communications, Sport and the Arts released the draft on September 8 and is accepting feedback until midday on September 22.

The government intends to use that consultation to refine the proposal before legislation proceeds through Parliament.

That process will matter because some of the most sensitive questions remain unresolved.

How easy must platforms make the non-algorithmic option?

What will prevent companies from degrading the experience for users who reject recommendations?

How precisely will serious harm be defined as technology evolves?

What protections will ensure lawful political speech remains outside the regime?

And how much power should ministers and eSafety have to respond to harms Parliament did not anticipate?

Those questions will dominate the next stage of the debate.

But the direction of government policy is already clear.

Australia is moving away from the assumption that technology companies should be free to design algorithms first and deal with the human consequences later.

Albanese says users should decide whether they want algorithms shaping their feeds.

Wells says Big Tech is approaching a global reckoning.

The political fight will now be over how far that reckoning should go without giving government too much influence over the online world it is trying to make safer.