I. Two Ways to Break the Same Doctrine
Every constitutional democracy inherits the same promise: that the power to make the law, the power to enforce it, and the power to judge it will never sit in the same hands. Montesquieu gave the doctrine its classical form in 1748, arguing that liberty perishes the moment a single body wields legislative and executive power together, and perishes just as surely if the power to judge is not separated from both. The framers of modern constitutions – Australia’s included, however imperfectly – built their institutions on that warning. Parliament writes the law. The executive administers it. The courts interpret it, and stand ready to tell the other two branches when they have exceeded their authority. Each branch is meant to check the others, and none is meant to answer only to itself.
The doctrine is written down. It sits in constitutions, in statutes establishing courts of independent jurisdiction, in the oaths sworn by judges and the conventions that keep ministers from directing prosecutions. All of that is what lawyers call the de jure position – the position “by law,” the position as it exists on paper regardless of what actually happens on the ground. But there is a second, quieter way a government can hold power, and it is the one this essay is really about: the de facto position, the state of affairs “in fact,” in practice, on the ground – which may or may not have anything to do with what the law says should be happening.
The distinction matters because there are two entirely different ways an unscrupulous government can dismantle the separation of powers, and only one of them looks like what most people expect an attack on democracy to look like. The first way is legislative: a government proposes, in public, to change how judicial or quasi-judicial outcomes are reviewed or held to account. This is attention-seeking, contestable, and – because it has to pass through parliament, survive scrutiny, and often clear constitutional challenge – comparatively difficult to pull off outright, and comparatively easy to argue about once it is on the table. The second way is administrative: a government never touches the law at all. It leaves the constitution, the statutes, and the independence of the judiciary formally intact, and instead spends years, sometimes decades, quietly shaping who holds every discretionary post beneath those formal structures. It does not need to instruct a police force or a tribunal to protect its interests if the people running it already understand, without being told, what protecting those interests looks like. Nothing on the statute book changes. What changes is who is inside the building.
This essay sets out both mechanisms. It uses a live, still-unfinished Australian policy proposal as a foil for the de jure route, and one thoroughly documented, independently tested federal case – the Robodebt scheme and the tribunal that was meant to catch it – as the evidentiary core of the de facto route. It does not claim the two mechanisms are equally severe or that they aim at the same destination, only that they are different species of the same underlying failure, and that one of them is considerably better camouflaged than the other.
II. The De jure Attempt: Legislating Around the Judiciary – and Where the Real Line Sits
The de jure route announces itself. In August 2026, Victorian One Nation leader Warren Pickering, appearing alongside Senator Pauline Hanson, said that a priority of any government he led would be to “make… bleeding-heart judges accountable for these second-time offences,” and pledged to end what he called a “tag-and-release” approach to youth bail. The party’s published Victorian justice policy, titled Tackling Crime, proposes tougher bail and sentencing settings for repeat offenders and expanded youth detention capacity, together with a review panel to examine serious offending committed by people already on bail. The policy document itself is specific about the panel’s composition, stating that it “would ideally comprise legal experts (lawyers and ex-judges) as well as genuine community representatives” – language worth quoting directly, because the panel’s design is exactly what determines whether it is a legitimate accountability mechanism or something closer to political pressure on individual cases, and the answer should not be assumed either way.
It would be inaccurate to present this as a proposal to seize control of individual verdicts, and a case built on that description would not survive contact with the actual policy. Nothing in the published platform purports to direct a court how to decide a specific matter, and legislatures routinely – and entirely lawfully – shape sentencing outcomes in advance: mandatory minimum sentences, standard non-parole periods, appellate sentencing guideline judgments, and judicial commissions that publish and monitor sentencing statistics are all long-standing, constitutionally uncontroversial exercises of a parliament’s own, separate function. A government that legislates a mandatory minimum has not abolished judicial independence; it has exercised its own authority to set the law the judiciary is then bound to apply. Judges remaining answerable to the law Parliament writes is not evidence of capture. It is the system working as designed.
The genuinely live question, then, is not whether a legislature may shape sentencing at all – it always has, and reasonable people disagree, often sharply, about how far that shaping should go. The live question is where a specific mechanism stops setting general rules in advance and starts pressuring outcomes in individual, identifiable matters. A panel that includes lay members and examines named offenders’ cases after they have reoffended sits closer to that line than a published sentencing guideline does, because its practical function is to generate visible, case-specific attention on decision-makers considering what happens next to a particular person, rather than to fix a rule of general application ahead of time. Whether this specific panel crosses that line turns on details the policy document does not yet settle: is its remit prospective and general, or retrospective and case-specific? Does it report to Parliament as a policy input, or does it correspond directly with courts or corrections decision-makers about matters still live? Is it insulated from commenting on cases still before a court, in the way sub judice conventions already require of ministers? These are the tests worth applying to the panel once it is fully designed. Whether a press release uses the word “accountable” is not.
What can be said with confidence, whatever this policy becomes, is the structural point this essay is built on: it will have to become something – a bill, introduced, debated, passed or rejected in public – before it can operate, and it will then be available to be tested against the constitutional protections that already constrain legislative interference with judicial process. That requirement to survive daylight before it can bite is what distinguishes a de jure proposal, however aggressive its rhetoric, from the mechanism examined in the rest of this essay.
III. The De facto Method: Capture Through Incumbency
There is a second, quieter way to achieve something like the same result, and it does not require any bill at all. An incumbent government that controls every lever of executive appointment for long enough does not need to legislate its way past judicial or quasi-judicial independence. It can simply outlast the people who would otherwise enforce that independence, and replace them, discretionary appointment by discretionary appointment, with people who do not need to be told what the government wants.
This is the essential feature of de facto separation-of-powers erosion: it requires no instrument. There is no statute to point to, no minister who signs an order directing a prosecution or a tribunal outcome, no clause that formally subordinates a reviewing body to the executive it is meant to check. Every formal structure remains intact. What changes is who sits inside those structures, and what those people understand – without ever being told directly – the government wants from them. A regulator or review body does not need a written instruction to treat a government-favoured party differently from anyone else; it needs members, at every level that matters, who have risen through, or been appointed into, a culture where that difference in treatment is simply how things are done.
I have spent years reporting on a long-running Australian Capital Territory employment and public-interest-disclosure dispute in which a similar pattern appears on the documentary record tendered in later court proceedings – correspondence and an internal departmental email suggesting a whistleblower-protection process was used, within days, against the person who had raised the original complaint. That reporting is offered here only as colour: it illustrates what the mechanism can look like inside a single agency, not as the evidentiary foundation for the argument that follows. The foundation is the case in the next section, which does not depend on this essay’s author having reported on it, having any stake in its outcome, or being believed about anything at all.
IV. The Proof: Robodebt and the Tamed Tribunal
The clearest illustration of de facto capture, tested independently of any single person’s account, is the Administrative Appeals Tribunal’s role in the Robodebt scheme – now a matter of extensive public record following the 2023 Royal Commission into the Robodebt Scheme, chaired by Commissioner Catherine Holmes.
In March 2017, AAT member Terry Carney AO – an administrative and welfare law academic of some four decades’ standing – handed down a decision finding that the income-averaging method underpinning Robodebt’s debt calculations was unlawful, and went on to make several further rulings to the same effect. The Department of Human Services did not appeal any of them to the Tribunal’s General Division, where such an appeal would have been heard more openly and the finding would have become difficult to keep from other debt recipients contesting their own notices. On the Royal Commission’s own findings, the department instead continued the practice Carney had found unlawful and did not report his rulings as the significant legal issue its own obligations required. In September 2017 – six months after his first adverse ruling, and after almost forty years on the Tribunal – Carney’s appointment was not renewed.
Here the caveat matters as much as the fact. Carney has said, giving evidence on oath to the Royal Commission, that he believed the non-renewal followed from his Robodebt rulings; that is his account of his own experience, and it should be read as such. Commissioner Holmes’s final report records the sequence of events and states that it raises the question of whether the government moved to inhibit scrutiny of the scheme by shifting personnel inside the bureaucracy – a question the report poses rather than a finding it makes on that specific causal link. The Royal Commission’s harder, better-supported findings lie elsewhere: that the scheme itself used a method of calculating debts inconsistent with the social security legislation it purported to apply, that departmental officials knew or ought to have known this well before the scheme ended, and that the department did not escalate or publicly report adverse tribunal rulings as its own obligations required. Carney’s removal is properly read as a strand within that pattern, consistent with it, rather than as a fact the Royal Commission itself definitively proved.
The specific figure comes from Carney himself. In a University of Sydney Law School interview published on 13 December 2023, reflecting on the Royal Commission’s findings, Carney said the department had breached its “model litigant” obligations partly by “ignoring and failing to appeal 220 tribunal findings of illegality,” and separately described the government as having “gamed” the AAT “by never appealing to the second tier ‘General Division’ any of the 220 rulings invalidating Robodebt at the first level of the Tribunal… because doing so would have made public the rulings of invalidity, jeopardising the continuation of the scheme.” The figure should be read as Carney’s own count of the tier-one findings he considered clearly on point, not as an independently audited or Royal Commission-certified total. The characterisation he draws from it – that the government kept unfavourable outcomes confined to a forum where they set no public precedent and triggered no reporting obligation – is the substantive point, and it does not turn on the precise figure. When the scheme was finally tested in the Federal Court, the government twice moved to avoid a judgment on the underlying legality: in the first case, by zeroing the specific debt in dispute just as the matter came on for hearing, removing the live controversy; in the second, only after resisting a subsequent challenge in which it eventually conceded the point. Robodebt was found unlawful in November 2019, more than two years after Carney’s original ruling, and only via the Federal Court rather than the tribunal designed to catch exactly this kind of error first.
None of this required amending the Administrative Appeals Tribunal Act, the social security legislation Robodebt purported to apply, or any provision governing appointments to the Tribunal. The Tribunal’s formal jurisdiction, its formal independence from the department whose decisions it reviewed, and its formal power to find a scheme unlawful were never touched by any statute. What changed was who continued to hold appointments to it, which of its findings were allowed to become visible, and which findings a department chose to escalate as required and which it chose not to. That is the mechanism this essay is about – capture of a check through appointment and selective compliance rather than through legislative amendment – documented independently, tested before a Royal Commission with the power to compel evidence under oath, affecting several hundred thousand people.
V. Cataloguing the Mechanisms
The Robodebt case, on its own, displays several distinct mechanisms by which an incumbent government can hollow out a check on its power without amending a single law. Each is transferable well beyond this specific scheme.
Non-renewal of discretionary appointments. A tribunal member’s or statutory officeholder’s contract simply not being renewed requires no reason to be given and no instrument to be signed. It is, formally, the ordinary and lawful expiry of a term. Its capacity to shape a body’s future findings depends entirely on timing and pattern – which is exactly what makes it difficult to challenge in any single instance and easy to see in aggregate.
Selective escalation and reporting. A review body’s formal power to find against the executive is worth little if adverse findings can be kept at a level of the process where they set no public precedent and trigger no obligation to report upward. The mechanism here is not that the Tribunal lacked power; it is that the power was allowed to operate only where its exercise stayed quiet.
Strategic litigation to avoid an adverse precedent. Repeatedly resolving individual disputes – zeroing a debt, settling a claim – just before a court would otherwise be forced to rule on the underlying legality of a broader scheme is a lawful use of ordinary procedural tools that nonetheless functions to keep a systemic question from ever being systemically answered.
Institutional non-compliance dressed as administration. Continuing a practice a tribunal has already found unlawful, without escalating, appealing, or publicly acknowledging that finding, is not a legislative act. It is a sustained administrative choice, repeated long enough to look like ordinary process rather than defiance.
Read together, these mechanisms share the feature this essay keeps returning to: none of them required a bill. Each one, taken alone, is close to unfalsifiable – a non-renewal is just a contract ending, a non-appeal is just a litigation choice, a delay is just administration. It is only in aggregate, and usually only after a body with the power to compel evidence goes looking, that the pattern becomes visible at all.
VI. Why the Quiet Method Deserves More Attention Than It Gets
It is tempting to treat a public pledge to make judges more “accountable” as the more alarming development, because it reads as a headline and photographs well. Section II tried to take that proposal seriously on its own terms rather than caricature it – and having done so, the honest conclusion is that a de jure attempt of this kind, even a poorly designed one, is the version democracy is comparatively well equipped to survive. It has to be introduced. It has to be debated. It can be challenged in exactly the courts it seeks to touch, on exactly the constitutional grounds those courts exist to protect. It must be enacted before it can operate, which means it can be stopped before it does any damage at all.
De facto capture offers none of those points of resistance, because it never asks for permission. No opposition MP can move to disallow a non-renewal. No court can strike down an institutional habit of confining adverse findings to a quiet forum, because no such habit is ever reduced to an instrument capable of being challenged. The people affected – a tribunal member whose contract simply isn’t renewed, several hundred thousand welfare recipients pursued under a scheme a department’s own review process had already found wanting – are told, correctly as a formal matter, that every relevant law was followed. The Tribunal’s independence was never legislatively touched. Each individual step withstands scrutiny in isolation, which is exactly what makes the aggregate pattern so difficult to name and so easy for a government to deny while it is happening.
This is not a claim that the two mechanisms are equally severe, or that a not-yet-enacted review panel and a documented pattern of tribunal capture are the same species of wrong – they plainly are not, and treating them as interchangeable would flatten a distinction this essay has tried to keep sharp throughout. The claim is narrower: whichever mechanism turns out to be more damaging in a given case, it is the quiet one that a democracy’s existing accountability structures – a free press reporting on bills before Parliament, an opposition able to divide the house, courts able to hear a constitutional challenge to a statute – are least designed to catch, because those structures are built to respond to instruments, and de facto capture does not produce an instrument until, sometimes, a Royal Commission is convened years later to go looking for one.
VII. Conclusion: What the Doctrine Requires Now
None of this is an argument that formal, legislative protections for judicial and quasi-judicial independence do not matter – they plainly do, and a genuinely case-specific “accountability” panel of the kind sketched in Section II would deserve the scrutiny, and quite possibly the opposition, it would attract once its actual design became public. But a society that measures the health of its separation of powers only by whether a statute has been amended will miss the erosion that can happen in plain sight, one non-renewed appointment, one un-escalated ruling, one quietly zeroed debt at a time.
The remedy for a de jure attack is public argument and, where necessary, litigation – tools a healthy democracy already has, and applies, before the damage is done. The remedy for de facto capture is harder to build, because it requires exactly the kind of independent oversight – appointment processes insulated from the government of the day, statutory obligations to escalate and report adverse findings that are actually enforced rather than quietly not complied with, Royal Commissions willing to compel evidence on oath when the ordinary channels have failed for years – that an administration well practised at the quiet method has every incentive to resist. Robodebt shows both halves of that problem in one case: the mechanism by which a check was disabled, and the years it took an extraordinary, purpose-built inquiry to make that disabling visible at all. The harder question the Robodebt case leaves behind is not whether any particular government would ever pass a law to discipline a tribunal. It is whether, in a system that only watches for instruments, a government confident enough in its own longevity still needs to.
– – –
Bakchos is the founder of Blak and Black, an Australian media and advocacy platform established in 2010. Bakchos writes from the intersecting perspectives of Wiradjuri heritage, Jewish identity, and humanism.
© Bakchos, September 2026
