Digital terra nullius? Artificial intelligence experimentation and sovereignty in Australia

Abstract

The rapid expansion of artificial intelligence (AI), data infrastructures, and digital experimentation in Australia raises a pressing question: are we witnessing the emergence of a new form of terra nullius, rearticulated through digital systems rather than land? The notion of ‘digital terra nullius’ although provocative, risks being either too metaphorical or too determinist. Instead, I propose that the expression can be productive if it is carefully situated in place and time. In doing so, I ask: to what extent do contemporary AI infrastructures reproduce colonial spatial logics and where might they diverge?

Australia has long functioned as a zone of experimentation, from penal colonies to nuclear testing at Maralinga (Arnold and Smith, 2006). These projects relied on the perception of remoteness and political marginality. Contemporary AI experimentation appears to inherit this spatial imaginary. Yet is Australia uniquely positioned as a testing ground, or is this part of a broader global pattern where peripheral regions are incorporated into technological experimentation? The answer likely lies in both. While Australia is not exceptional its colonial history renders these interactions particularly legible. In fact, the proliferation of AI testbeds in Australia, from autonomous mining in the Pilbara to drone delivery trials in Canberra and in Queensland, illustrates how experimentation has moved from controlled environments into everyday life. These systems depend on continuous data flows, environmental sensing and iterative optimisation.

Phan's contributions also beg the question, is the ‘wild’ the landscape, the population, or the regulatory environment? Framing everyday environments as experimental spaces risks naturalising the enrolment of citizens into technological systems. At stake here is the constitution of territory into a site of continuous experimentation. Moreover, is technological experimentation inherently extractive and colonial, or can it be reconfigured as participatory and sovereign? This question is one of the significant issues of Australia's emerging digital political economy, the subject of Phan's (2025) journal article where the ‘test bed nation’ morphs temporally from whiteness to multi-ethnicity and automated drone delivery over the past 240 years. In the pages that follow, my commentary attempts to engage with some of Phan's arguments, with a particular focus on conception of Australia as terra nullius.

The doctrine of terra nullius functioned as a legal fiction that enabled British colonisation by declaring Australia an empty land (Reynolds, 2006). As Wolfe (2006) famously argued, settler colonialism operates through a ‘logic of elimination’ through overt violence, the erasure of Indigenous presence and sovereignty. Extending this framework to digital infrastructures, however, requires caution and conceptual precision.

We need to unpack what exactly is being rendered ‘empty’ in digital contexts. It is not land in the traditional sense, but social relations, cultural meanings and forms of knowledge that are abstracted into data. I suggest that this process of abstraction can be understood across three interrelated dimensions. First, an epistemic dimension, where complex forms of local and Indigenous knowledge are reduced to data points that can be processed and monetised (Kukutai and Taylor, 2016; Smith, 2012). Second, a legal dimension, where regulatory gaps or ‘sandboxes’ enable experimentation without full accountability (Allen, 2019; Brummer and Yadav, 2018; Pasquale, 2015). Third, a social dimension, where everyday life is reframed as neutral ‘user behaviour’, obscuring its embeddedness in culture, history and inequality. In this sense, digital terra nullius, a term first outlined by Kereopa-Yorke (2024) in the context of LLM induced ‘corporate cognitive colonialism’, resonates with what Couldry and Mejias (2019) conceptualise as data colonialism; the appropriation of human life through data extraction, albeit in a specifically settler-colonial context where land dispossession continues to shape infrastructural development.

However, scholars have shown it is simplistic to interpret all AI experimentation as coercive or exploitative. Some think tanks and policy centres (Productivity Commission (2024); CSIRO (2025) emphasise that such infrastructures generate economic opportunities, improve efficiency and position Australia within global innovation networks. Autonomous mining, for instance, has increased productivity and safety in hazardous environments. Amoore (2020) suggests that algorithmic systems operate through continuous data capture, transforming everyday actions into computational resources. This suggests that ‘testing-in-the-wild’ is as Phan has compellingly claimed a form of experimentation, but equally importantly a mode of governance that operates through ongoing optimisation rather than regulatory interventions.

The recent expansion of data centres in Australia ¹ driven by ‘AI hyperscaler’ firms such as Amazon (AWS), Microsoft and Google, underscores the material foundations of digital economies. Far from being ‘cloud-based’ in any abstract sense, AI systems rely on energy-intensive infrastructures and cooling systems embedded in specific geographies. Although these developments are often framed within narratives of progress, innovation and national competitiveness, it is debatable who benefits from such innovations.

Data centres, apart from storing information, enable the extraction and monetisation of data generated through social and environmental interactions. As Kitchin (2019) argues, algorithms are socio-technical systems embedded within power relations. The localisation of data infrastructure does not necessarily equate to local control over data. Hence, while Australia hosts data centres, the ownership and governance of these infrastructures often remain largely in the hands of multinational corporations. This raises the possibility that national territories function as nodes within global data extraction networks rather than as sites of sovereign control. This is an extension of what Zuboff's (2019) calls ‘surveillance capitalism’, where behavioural data is extracted, processed and monetised as predictive products, raising concerns about autonomy, consent and asymmetrical power. Yet, as pointed out earlier, hosting data centres may also enhance national digital capacity and support domestic industries. So, it is necessary to examine how their governance structures distribute power and value.

In Australia, debates around Indigenous data sovereignty provide a crucial lens through which to evaluate these workings. Kukutai and Taylor (2016) articulate Indigenous data sovereignty as the right of Indigenous peoples to govern data derived from their lands, communities and resources. This framework challenges the assumption that data is a freely available resource. Moreton-Robinson's (2015) concept of the ‘white possessive’ further illuminates how settler colonial governance extends into contemporary institutions. If land was historically appropriated through legal fictions, it is possible that data can now be appropriated through technological infrastructures. However, as scholars have importantly noted (Kukutai and Taylor (2016); Carroll et al. (2019), indigenous innovation in data governance, environmental monitoring, and digital storytelling is a testament to the fact that Indigenous communities are active agents in shaping digital futures via cultural and environmental production and land relations. More recent developments in Indigenous data governance further emphasise principles such as Collective benefit, Authority to control, Responsibility, Ethics (CARE), which seek to reframe data practices in line with Indigenous values and rights (Carroll et al. (2019).

The question then is also to explore ways in which digital infrastructures are being reconfigured to support Indigenous sovereignty. Initiatives in Indigenous data governance suggest that alternative models are possible, but they require structural changes in how data is owned, accessed and used. Watego (2021) reminds us that technological governance cannot be separated from broader struggles for justice. Without addressing underlying inequalities, digital inclusion risks becoming another form of incorporation into unequal systems.

Here, Phan's analysis situates technological experimentation within the broader framework of innovation nationalism, where states compete to attract investment and technological capacity is significant. Australia's investments in AI, defence technologies and digital infrastructure reflect this strategic positioning. Yet there is often a paradox as AI models and systems are developed, trained, deployed and governed through transnational networks that often exceed national control and innovation agendas. Weiss (2014), Couture and Toupin (2019) suggest that nation-states retain significant regulatory power and can shape technological development through policy frameworks. Australia's debates around data sovereignty, privacy and AI ethics, like many other countries, indicate an awareness of these challenges. However, the effectiveness of such interventions remains uncertain. The concentration of technological power in a small number of firms limits the capacity of individual states to assert control and reflects what Srnicek (2017) identifies as ‘big tech’ platform capitalism. Only with the global concentration and the oligarchic domination of three or four AI models, the domination is set to be repeated on a much more lopsided scale.

The concept of ‘digital terra nullius’ captures the possibility that contemporary technological systems may reproduce logics of availability, extraction and experimentation. In that sense, the analogy with terra nullius should be understood as heuristic device rather than literal. A way of highlighting continuities without erasing differences. Moreover, focusing on colonial analogies risk overlooking other dimensions of digital power, such as capitalism, platform monopolies and geopolitical competition. While colonial frameworks are essential for understanding Australia's context, they should be integrated with broader analyses of digital capitalism (Srnicek 2017).

The key issue is how these colonial logics are reconfigured through data, platforms and the intertwined experimentation of AI models and applications. Most of these AI applications remain in the future at this point. Can AI infrastructures be governed in ways that support sovereignty and equitable participation centreing Indigenous data sovereignty. These questions do not admit simple answers. They do, however, point towards the need for a critical politics of digital infrastructure.

Just as terra nullius was a fiction that enabled dispossession, scholars need to carefully interrogate new fictions of digital emptiness. This is critical given the current ominous undertones and uncertainties of AI's impact on societies and particularly Indigenous people. We need to insist on the rights and agency embedded within every data infrastructure and algorithmic system.