The algorithm switch is the easy part to defend
What if you could simply switch off the social-media algorithm?
That is the most appealing part of the Albanese Government’s latest online-safety package.
Under its My Feed, My Way proposal, Australians would be given greater control over whether social-media platforms use personalised recommendation algorithms to decide what appears in their feeds (Australian Government, 2026a).
On its own, that is hard to object to.
If people want to see content primarily from accounts they have deliberately chosen to follow, rather than material selected for them by an engagement algorithm, they should be able to.
Platforms such as X (formerly Twitter) already demonstrate the basic concept through separate For You and Following feeds. Users have also long been able to block content containing keywords of their choosing.
Giving users more control over algorithms is not the problem.
The problem is that the algorithm switch is only one part of a much broader package.
On 8 September 2026, the Government announced My Feed, My Way alongside draft Digital Duty of Care legislation.
And that is the part that deserves much closer scrutiny.
Because while the algorithm switch is about giving users more control, the Digital Duty of Care is about government-defined rules for what platforms must identify, suppress or prevent as harmful.
That is a much bigger proposition.
The Digital Duty of Care is the bigger change
The Government describes the Digital Duty of Care as a shift toward requiring online services to take reasonable steps to prevent foreseeable harm.
The Government’s May 2026 framework says covered services would be required to:
“assess their risks and determine what preventative responses, systems and processes they can and ought to develop and implement”
(Australian Government, 2026b, p. 3).
This is much broader than changing which TikTok or Instagram posts appear in someone’s feed.
The proposed duty would apply across services already captured by the Online Safety Act, including social-media platforms, messaging applications, online games, dating services, websites, search engines, hosting providers, app stores, internet service providers and generative-AI systems such as chatbots and image generators (Australian Government, 2026b).
Services would be expected to identify potential risks, implement systems intended to prevent or mitigate reasonably foreseeable serious harms and continually assess whether those measures are effective.
The framework specifically contemplates measures including content classification and moderation systems, account and recommendation settings, parental controls, and the design of algorithmic recommendation systems themselves (Australian Government, 2026b).
There are legitimate reasons for such an approach.
Children can be exposed to pornography, self-harm material, eating-disorder promotion, grooming and other genuinely harmful content.
Platforms should not be immune from reasonable safety obligations simply because harmful interactions occur online rather than offline.
But this is also a significant change in regulatory philosophy.
The question increasingly becomes not merely:
What material is illegal?
but:
What harms should platforms be required to anticipate and prevent?
And once that becomes the question, another follows immediately:
Who defines harm?
Before getting to that, however, it is worth considering something the Communications Minister herself has said about what she believes algorithms are doing to Australian society.
What does the Government think the problem actually is?
Communications Minister Anika Wells has connected the growing power of online algorithms with declining public trust in Australia’s institutions.
In a September 2026 interview with The Saturday Paper, Wells said:
“I’ve always found until this year even that the way to solve that malaise and that sense of distrust in the establishment was to go and knock on people’s doors and look them in the eye.”
“What I’m noticing now is that they don’t. The power of online is so strong, the power of the algorithm is so strong that it is overwhelming our public institutions and the decline of trust is problematic for everybody.”
(Barlow, 2026).
The significance of those comments became even clearer three days later.
At the Government’s September 8th press conference announcing My Feed, My Way, a journalist explicitly raised Wells’ comments about declining political trust and algorithms.
The journalist asked whether distrust could also arise from government behaviour, including governments walking back political commitments, rather than from social-media algorithms.
Wells responded that people were arriving at her mobile offices after reading things online and then refusing to believe her even when she told them what the Government itself was legislating (Australian Government, 2026c).
That does not prove that My Feed, My Way or the Digital Duty of Care is designed to suppress political criticism.
But Wells’ comments do reveal something important about how the Communications Minister conceptualises the problem.
The concern is not confined to children encountering pornography or self-harm material.
She also sees algorithmically distributed information as powerful enough to overwhelm public institutions and contribute to declining trust.
That deserves scrutiny because these are not the same policy problem.
Protecting children from sexual exploitation, suicide-promoting material or predatory behaviour is one thing.
Responding to declining public confidence in governments and institutions is another.
And Australia’s recent history shows why this matters.
Australia’s Labor government has already tested the boundaries of information regulation
In 2023, the Albanese Government released an exposure draft of the Communications Legislation Amendment (Combatting Misinformation and Disinformation) Bill.
A revised bill was introduced in 2024.
The legislation aimed to require digital communications platforms to manage risks created by misinformation and disinformation, with ACMA receiving additional regulatory powers (Parliamentary Library, 2024).
The 2023 exposure draft defined misinformation around information that was false, misleading or deceptive and reasonably likely to cause or contribute to serious harm.
Most controversially, the original proposal excluded government-authorised content from the misinformation regime.
That did not mean government statements would legally be declared true. But it would have created a regulatory exemption for government-authorised content unavailable to other speakers.
Following substantial criticism, that exemption was removed before the revised 2024 bill was introduced.
The revised bill contained stronger safeguards but retained a broad concept of “serious harm”.
Its categories included:
- harm to elections and referendums;
- harm to public health, including preventative health measures;
- vilification; and
- harm to the Australian economy.
The economic category specifically included:
“harm to public confidence in the banking system or financial markets”
(Parliamentary Library, 2024).
There are understandable reasons government might be concerned about those scenarios.
A claim that a major bank is insolvent could contribute to a bank run.
Deliberately misleading voters about polling arrangements could interfere with an election.
False health information can cause genuine harm.
The difficult question is where harmful falsehood ends and legitimate disagreement, criticism, uncertainty or unpopular opinion begins.
The Parliamentary Library raised uncertainty around the breadth of content potentially captured and whether aspects of the bill would operate compatibly with Australia’s international freedom-of-expression obligations (Parliamentary Library, 2024).
The Australian Human Rights Commission was also concerned.
It warned that claims of misinformation and disinformation can themselves be used to delegitimise competing viewpoints and restrict discussion on matters of public importance.
The Commission ultimately concluded that the 2024 bill did not adequately protect freedom of expression in its existing form (Australian Human Rights Commission, 2024).
The legislation ultimately failed to secure sufficient Senate support and did not proceed.
The Government has now returned with a different regulatory mechanism.
Rather than constructing a regime specifically around misinformation and disinformation, the Digital Duty of Care would require platforms to identify and mitigate foreseeable online harms, with the framework presented primarily as a protective measure aimed at reducing harm, particularly to children and other vulnerable users.
The continuity in regulatory philosophy matters.
→ Identify the risk
→ Mitigate the risk
→ Regulator oversees compliance
→ Government defines the framework
→ The framework can expand
That brings us back to the most important question in the current proposal.
Who defines “harm”?
The May 2026 Digital Duty of Care framework contains an important provision that deserves much more public attention.
Under the proposed framework, the Communications Minister, or the eSafety Commissioner acting at the Minister’s direction, would be empowered to make binding rules:
“designating additional categories and types of harm”
and:
“specifying compliance requirements for service providers”
(Australian Government, 2026b, p. 3).
That does not mean a Minister could secretly invent new categories of prohibited speech overnight.
There are safeguards.
The Government’s framework says these rules would be made as disallowable legislative instruments, meaning they would be subject to parliamentary scrutiny.
They would also be required to include a statement addressing compatibility with Australia’s human-rights obligations, including freedom of expression (Australian Government, 2026b).
Those protections should be acknowledged. However, Labor and the Greens currently hold a combined majority in the Senate. If both supported an expansion, that would ordinarily be enough to defeat a motion to disallow it, making the safeguard less protective in practice than it may first appear.
The underlying principle remains significant.
Parliament would establish the broader framework, while additional categories of harm could subsequently be designated through delegated legislation rather than requiring an entirely new Act of Parliament.
The framework itself states:
“The Minister for Communications would have the ability to determine, by disallowable legislative instrument, additional harms as being within the scope of the Duty.”
(Australian Government, 2026b, p. 5).
This matters because the Digital Duty of Care is explicitly preventative.
Platforms would be expected to assess the potential for serious harms to emerge and put systems in place to prevent or reduce those risks.
The Government’s framework includes content moderation among the systems that providers may use to comply with the duty (Australian Government, 2026b).
That creates another potential problem.
The government does not need to delete a post itself
When people hear the word “censorship”, they often imagine a government official personally deciding which newspaper article, social-media post or political opinion must be removed.
Modern regulatory systems do not necessarily operate that way.
A more realistic risk is regulatory over-compliance.
If a platform faces substantial penalties for inadequately identifying or preventing categories of harm, its rational commercial response may be to moderate borderline material more aggressively than is strictly necessary.
The incentive becomes:
When uncertain, remove, suppress or reduce the reach of the material rather than risk regulatory action.
That outcome does not require a Minister to personally direct that a particular political opinion disappear. It can emerge from the incentives created by the regulatory framework itself.
This is why the breadth of “harm”, the process for expanding it and the safeguards surrounding enforcement matter so much.
And the international debate shows that questions about government influence over the information environment do not necessarily end with deciding what platforms should remove or suppress.
The international debate is already moving toward which sources should be amplified
In June 2026, the UK Labour Government released its Watch This Space Media Green Paper.
Its proposals move the debate beyond removing harmful information.
The British Government says it will explore legislation requiring social-media companies to make news from public-service media providers prominent and easily discoverable (UK Department for Culture, Media and Sport, 2026a).
The accompanying announcement said designated publishers could become more likely to appear prominently when people search for news on social-media platforms (UK Department for Culture, Media and Sport, 2026b).
A subsequent parliamentary answer confirmed that platforms might achieve such prominence through recommendation algorithms or other mechanisms (UK Parliament, 2026).
The UK proposal is also not currently Australian policy and should not be represented as though it is.
But the principle being debated internationally is significant:
Should governments merely regulate what platforms remove, or should regulation eventually extend to which sources platforms make more visible?
That represents another important shift.
The information-policy debate can move from:
→ What must platforms remove?
to:
→ What must platforms prevent?
and eventually:
→ What should platforms promote?
Information regulation deserves a higher threshold
Australians are generally quick to recognise the danger when authoritarian governments control access to information.
In North Korea, the state runs all domestic media outlets and ordinary citizens cannot access the global internet; unsanctioned media can attract severe punishment (Freedom House, 2025a; Human Rights Watch, 2025).
Iran maintains extensive internet censorship, with major social-media platforms restricted and authorities pushing users toward a more controllable domestic internet (Freedom House, 2025b).
Russia has progressively tightened online information controls through website blocking, platform restrictions, limits on independent media and laws targeting “extremist”, foreign-agent and “fake news” content (Freedom House, 2025c; 2025d).
Australia is plainly not comparable to those systems, and the current proposals are nowhere near that level of control.
But that is exactly why the principle matters.
The lesson from more authoritarian systems is not that every information-regulation law inevitably leads to repression. It is that government influence over what information citizens can access, what sources are prioritised, and what speech is treated as harmful is a category of power that deserves an unusually high threshold of scrutiny.
Democratic safeguards matter precisely because governments change, definitions evolve and powers created for one purpose can later be used for another.
Would you give the same powers to the other side?
Perhaps the simplest test of any government power is to imagine it being exercised by the political party you trust least.
Suppose a future One Nation government inherited the same regulatory architecture.
- The same Digital Duty of Care.
- The same regulatory powers.
- The same ability to designate additional categories of harm.
- The same ability to specify compliance requirements for digital platforms.
Would Labor and Greens supporters remain equally comfortable with those powers?
If you replaced Anthony Albanese and Anika Wells at the press conferences of the past few days with Pauline Hanson and Barnaby Joyce (or, for American readers, Donald Trump and JD Vance) announcing the same powers over algorithms, online harms and platform obligations, many of the people currently treating these proposals as benign would almost certainly view them very differently and be sounding the alarm.
That is the test.
Good institutional design should remain acceptable when political power changes hands.
The safeguards around government power should not depend on whether you happen to trust or support the politicians currently holding it.
The relevant question should never simply be:
“Do I trust this government?”
It should be:
“Would I trust any government with the same power?”
What do you think?
Rather than guessing where Australians draw that line, we’re asking them directly.
The short survey below explores views on algorithm choice, online safety, government power, freedom of expression and how comfortable people are with these powers expanding over time.
References
Australian Government 2026a, My Feed, My Way, Prime Minister of Australia and Minister for Communications, 8 September 2026.
Australian Government 2026b, A Digital Duty of Care for Australia: Developing a duty of care framework for online services used by Australians, Department of Infrastructure, Transport, Regional Development, Communications, Sport and the Arts, May 2026.
Australian Government 2026c, Press conference — Parliament House, Canberra, Prime Minister of Australia, 8 September 2026.
Australian Human Rights Commission 2024, Combatting Misinformation and Disinformation Bill 2024 — Submission, Australian Human Rights Commission, 10 October 2024.
Barlow, K 2026, ‘The Anika Wells interview’, The Saturday Paper, no. 616, 5–11 September 2026.
Department of Infrastructure, Transport, Regional Development, Communications and the Arts 2023, Communications Legislation Amendment (Combatting Misinformation and Disinformation) Bill 2023 — Exposure Draft, Commonwealth of Australia.
Freedom House 2025a, North Korea: Freedom in the World 2025, Freedom House, Washington, DC.
Freedom House 2025b, Iran: Freedom on the Net 2025, Freedom House, Washington, DC.
Freedom House 2025c, Russia: Freedom on the Net 2025, Freedom House, Washington, DC.
Freedom House 2025d, Russia: Freedom in the World 2025, Freedom House, Washington, DC.
Human Rights Watch 2025, World Report 2025: North Korea, Human Rights Watch, New York.
Parliament of Australia 2024, Communications Legislation Amendment (Combatting Misinformation and Disinformation) Bill 2024, Parliament of Australia, Canberra, status: Not Proceeding.
Parliamentary Library 2024, Bills Digest: Communications Legislation Amendment (Combatting Misinformation and Disinformation) Bill 2024, Parliamentary Library, Parliament of Australia, Canberra.
UK Department for Culture, Media and Sport 2026a, Watch This Space: A New Strategic Direction for UK Media — Green Paper and Public Consultation, UK Government, 23 June 2026.
UK Department for Culture, Media and Sport 2026b, Plans for prominence of trusted news sources on social media alongside measures to reform Public Service Media in the UK, UK Government, 23 June 2026.
UK Parliament 2026, Written question 17870: Public Service Media prominence, UK Parliament, answered 16 July 2026.
