Canberra Gets Big Tech To Do Their Censorship Dirty Work

The most sophisticated thing about the Albanese Government's proposed Digital Duty of Care is that Canberra doesn't need to censor you itself. It can make censorship somebody else's problem.

That's the part of this legislation Australians need to understand, because the debate isn't simply about government ordering Facebook to delete particular posts. The machinery works further upstream.

Under the proposed law, online services would be legally responsible for providing a “safe online environment”. They would have to identify foreseeable risks, assess them, document what they're doing about them, and take measures to mitigate them. Then, substantial penalties sit behind those obligations.

Think about what that does inside a corporation. Proposed section 26A requires providers to conduct written risk assessments. These assessments must identify all reasonably foreseeable risks posed by the service, including content giving rise to those risks, design features, other systems or processes, and the people who may be affected.

The company then has to consider the likelihood and potential severity of the risk. It must identify measures it has implemented or proposes to implement to address the problem and assess how effective those measures are expected to be. This isn't a one-off exercise. The assessments have to be reviewed, and the eSafety Commissioner can impose additional requirements concerning how they're done.

Now imagine you're running compliance for a giant technology company. You aren't merely asking whether a particular Facebook post is illegal. You're asking whether your entire service is adequately identifying and mitigating foreseeable risks under a statutory duty enforced by a powerful regulator.

Your lawyers are involved. Your engineers are involved. Your trust-and-safety people are involved. Your executives know that getting this wrong can expose the company to enormous penalties.

Then somebody brings you a controversial piece of lawful content sitting somewhere near the edge of a regulatory category. Perhaps it's fine. Perhaps the regulator eventually sees things differently. What does the corporate lawyer say?

Why take the risk?

That's the mechanism. You don't need Anthony Albanese sending Mark Zuckerberg a secret spreadsheet of conservatives to silence. You don't need a smoke-filled room. You don't need some grand conspiracy.

The government creates the regulatory incentive to censor. Big Tech responds rationally to that incentive. And your speech becomes somebody else's compliance risk.

This matters because censorship in the digital age doesn't require deleting anything. That's the old model. Imagine that I publish a political argument to my 140,000 followers. Facebook doesn't delete it. Technically, I'm completely free to speak. However, the recommendation system stops recommending the post. It disappears from suggested content. Its distribution is throttled. The account's reach falls through the floor.

The post remains sitting there on my profile, so everybody can deny that censorship occurred. The Government says it never ordered the post removed. The regulator says it never banned my political opinion. The platform says its independent systems determined distribution.

Wonderful. Everybody's hands are clean. The post survived. But the audience didn't. It's kind of like letting you speak your mind from the middle of the town square, as long as it's done within a soundproof booth.

This is why the bill's provisions dealing with the design of online services matter so much. Proposed section 25F expressly addresses features such as recommender systems, logged-in features, endless feeds, feedback features, and time-limited features.

The bill defines a recommender feature as one that selects material using information associated with a user's account and displays it to the user. In ordinary English, we're talking about the machinery deciding what appears in your feed.

Section 25G treats various features, including recommender features and endless feeds, as having negative behavioural impacts for the purposes of protections involving children. Once again, the Minister is given power to determine additional design features by legislative instrument.

Now consider how profound this is. Most Australians don't manually choose everything they encounter online. They open an app. Then the algorithm takes over.

It decides which political story appears first. Which video gets recommended. Which controversy explodes. Which argument reaches millions. Which account suddenly finds a massive audience. And which one quietly disappears.

Algorithms have become the invisible editors of modern society. The ability to influence those systems is therefore the ability to influence attention, and attention is political power.

The Government will argue that its regulation of these systems is about safety rather than politics. But the distinction becomes much less comforting when political speech itself is increasingly discussed in terms of harm, extremism, radicalisation, division, and social cohesion.

A controversial political movement doesn't need to be formally outlawed to be crippled online. It merely needs to become risky: risky to recommend, risky to promote, risky to expose children to, risky to leave unrestricted when a regulator may later ask why the platform's mitigation systems didn't act. Then the incentives do the rest.

Section 26 goes further by allowing the Minister, through a legislative instrument, to require specified online services to provide specified “user empowerment tools”. The legislation expressly contemplates those tools affecting “the kind of content that is recommended for the user”.

This will be marketed as consumer choice. But there's a much bigger principle hiding underneath it. Canberra is claiming a statutory role inside the architecture of private recommendation systems.

Once that principle is established, the future argument is no longer about whether government belongs there. It becomes an argument about degree. Which systems? Which recommendations? Which categories? Which risks? Which controls? How far should the government go?

And all of this sits inside a wider duty requiring companies to ensure, so far as reasonably practicable, a safe online environment. Then come the penalties.

Proposed section 26D allows the eSafety Commissioner to issue a written remedial direction where the Commissioner reasonably believes a provider has failed, or is failing, to comply with the digital duty. The Commissioner can require specified action directed towards ensuring future compliance. Failure to comply with a remedial direction carries a civil penalty of 60,000 penalty units. That's a regulatory weapon that will be in the hands of the censorious eSafety Commissioner.

Now put yourself back in that corporate boardroom. On one side sits free expression. On the other sits regulatory exposure. If you under-moderate, the government can punish you. But if you over-moderate, what does it really cost the company? A post gets buried. An account loses reach. A controversial argument disappears from people's feeds. The real cost is borne by the person whose speech has been restricted, while the company has protected itself from regulatory risk.

The incentives aren't balanced. They lean towards restriction. That's why the danger of this legislation isn't confined to whatever the Government explicitly intends to suppress. The greater danger is what platforms will suppress because they don't want to find out whether Canberra considers it a problem.

And that is how modern censorship can become extraordinarily difficult to see. Nobody bans your political opinion. Nobody prosecutes you. Nobody knocks on your door. You can still type whatever you want. Your account remains online. Your post remains online. You remain gloriously “free”.

It's just that fewer and fewer people ever encounter what you said. That is the beauty of algorithmic suppression from the censor's perspective. It achieves much of the practical effect of censorship without producing the political spectacle of an actual ban, and it gives everyone plausible deniability.

That's why Australians should stop asking only whether this bill explicitly prohibits political speech. That's the wrong question. Ask instead what incentives the legislation creates around political speech.

Ask what happens when a controversial opinion intersects with a category of harm. Ask what happens when regulators believe a platform isn't adequately mitigating a risk. Ask what happens when a company has to choose between maximum freedom and minimum regulatory exposure. Then remember who has the money at stake.

The Government tells Big Tech what risks it must control. Big Tech knows the consequences if Canberra decides it hasn't done enough. So Big Tech protects itself.

Big Tech does the Government's dirty work. And you get restricted.

That's the censorship mechanism Australians should be worried about. Not the red pen of some Minister of Truth. Rather, it's the algorithm.

“The remedy to be applied is more speech, not enforced silence.”
― Justice Louis Brandeis, Whitney v. California (1927).