By Elise Bant
The recent decision of the National Anti Corruption Commission not to refer charges against any individual named in the Robodebt Royal Commission Final Report may have shocked the public, but cannot have surprised any legal observer versed in organisational misconduct.
As is well-known, modern government bureaucracies commonly operate as and through complex and devolved structures, including through outsourcing to private corporate actors. Additionally, governments have increasingly turned to automated and algorithmic processes to carry out core public functions and services: see Exhibit 1 – ‘Robodebt’.
In this brave new world of diffused and AI-enhanced operational structures, the law’s siloed focus on individual ministers’ or officials’ subjective purposes, knowledge and good faith is no longer ‘fit for purpose’ to promote either good governance or proper organisational accountability.
The result is a woeful lack of public accountability, endangering fundamental requirements of equality and the rule of law.
The truth is that public juristic persons suffer from the same structural barriers to accountability that have long frustrated corporations law. Indeed, the individualistic attribution rules that apply to public entities were adopted from corporate law. However, unlike in the public sphere, here, the law has recently undergone profound changes, as explained previously in the Brief article, Corporate Conduct and the Systems Intentionality Model.
In Productivity Partners Pty Ltd v Australian Competition and Consumer Commission [2024] HCA 27; (2024) 98 ALJR 1021, members of the High Court threw their support behind a holistic approach to corporate responsibility, which recognises that corporations, as Gordon J put it (at [108]), ‘“think” and act through systems of conduct’. More fulsomely, corporations manifest (in the dual senses of reveal and instantiate) their states of mind through their real-life systems of conduct, policies and practices.
Courts (and directors, regulators, and members of the public) may readily discern these mental states by engaging in an objective process of interpreting or construing the system of conduct: its positive features, default settings, audit and remedial aspects. This interpretive process is apt to reveal a great deal about corporate states of mind relevant to their responsibility. In the words again of Justice Gordon (at [108]):
Systems are inherently purposive. Systems are objectively designed to achieve certain ends; they coordinate and connect steps and processes to an end. Knowledge of certain matters is often implicit in a system – for example, that a predatory business model will only be profitable if a certain class of vulnerable customer exists and is successfully exploited.
In addition to the sort of specific knowledge alluded to by Gordon J, I have argued elsewhere that corporations must enjoy a certain baseline level of organisational knowledge and understanding of the existence and patent features of what is, after all, the corporation’s own system of conduct.
Otherwise, successful, repeated deployment of a system of conduct is a matter of accident or good luck: unlikely in an organisation of any size. Any claims of ‘systems errors’ or accidental deployment therefore fall to be substantiated by the corporation, including in light of how they responded to its repeated deployment: Brent Fisse’s powerful idea of ‘reactive corporate fault’. The longer a system operates without correction, the more likely it is that the system is operating according to its terms: as intended. None of this is surprising, and all heavily (if incidentally) supported by Australia’s rich jurisprudence on unconscionable systems of conduct.
No doubt, much remains to be seen about how this model of corporate responsibility will shape Australian law and practice. But just suppose, for a moment, that this model was applied to Robodebt. What would it say about the Commonwealth’s intentions and knowledge throughout that sorry saga, assessed independently of the individuals through whom it acted? In my 2025 article in the UWA Law Review, I set out the accepted, core features of Robodebt:
- it was an automated scheme for calculating, raising and asserting debts;
- premised on ‘income averaging’, which divided a person’s annual income by 26 to produce a notional fortnightly income;
- addressing a class of ‘social security recipients’;
- members of which would, by definition, have insecure income.
Objectively construed, the inevitable and patently obvious consequence of this system design was that the asserted fortnightly income would be wrong for many, perhaps most, recipients. This was a significant problem, given that welfare recipients’ entitlements (and hence any ‘overpayments’), were required by law to be based on their actual income.
It follows that, functioning according to these very basic terms, the Robodebt system would patently (objectively, on its face) and inevitably impose false ‘debts’ on members of the target class. Remembering that systems are inherently purposive, and recalling that a corporation must know the key features of its own, systemic conduct, through this systemic, organisational lens, Robodebt was far from accidental, and not mistaken.
Underscoring the point, the system was deployed over years and in the face of mounting evidence of its harmful effects. This was intended and knowing conduct. Further, the Robodebt scheme (system, plan, strategy) was itself supported by a broad and powerful range of supporting sub-systems of enforcement, all of which worked in a coordinated way to ensure that the Scheme raised (cf ‘recovered’), and retained, as much cash as possible, generally by removing existing legal and procedural protections that protected the vulnerable class of welfare recipients.
As I observe in my article, in a commercial setting, this would plausibly constitute a quintessential example of statutory unconscionable conduct: a deliberate advantage-taking of the known circumstances of special vulnerability of welfare recipients as a class. Indeed, Gordon J’s observations on ‘predatory’ business models appear an apt way to describe the then-government’s ‘savings’ (alternatively, revenue-raising) aspirations manifested through the Robodebt scheme.
Now, governments are not subject to fair trading laws in the same way as are commercial corporations. And our existing liability mechanisms (such as the ancient tort of ‘misfeasance in public office’) remain wedded to individualistic conceptions of abuse of public power, which require proof of individual dishonesty, recklessness and knowledge. So they largely fail, for the reasons already articulated.
In my view, it is time to re-imagine what public responsibility looks like. And this will be my final contribution, pending retirement, to the Royal Commission research ecosystem with which all this started.
