Free beta access. 50 free AI tokens each week. Optional top-ups available. See what is included
Opening Word document…

The securitization and regulation of AI creates an algorithmic settler colonialism---relegating AI to elitist legal discourse disavows policymakers from the violent and exploitative effects of AI securitization

Stark et al 20, Luke Stark---Assistant Professor in the Faculty of Information and Media Studies at the University of Western Ontario, Daniel Greene--- Assistant Professor of Information Studies at the University of Maryland, Anna Lauren Hoffman---Assistant Professor with The Information School at the University of Washington (“Critical Perspectives on Governance Mechanisms for AI/ML Systems,” in The Cultural Life of Machine Learning: An Incursion into Critical AI Studies, Palgrave Macmillan, 2020, pp. 269-271)

However, this emphasis on AI through the lens of already existing cybersecurity discourses and as a national security problem risks further entrenching AI policy in the hands of the few. Securitization theory argues that discourses around “security” deploy particular “grammars of securitization” to connect “referent objects, threats, and securitizing actors together” to both depoliticize and control particular sociotechnical arenas (Hansen & Nissenbaum, 2009). AI’s securitization shares similar traits with those of cybersecurity more broadly (which itself draws strongly from Cold War discourse around nuclear arms). AI’s “grammar of securitization” both reinforces AI’s status as an elite technical discourse and distracts from the effects of that same discourse on AI policymaking across the board. Hansen and Nissenbaum (2009) identify three “securitization modalities” which they suggest are particularly powerful in broader cybersecurity discourse: hyper-securitization or “a tendency both to exaggerate threats and to resort to excessive countermeasures”; everyday security practices, i.e., “situations in which private organizations and businesses mobilize individual experiences” to “make hyper-securitization scenarios more plausible by linking elements of the disaster scenario to experiences familiar from everyday life”; and technification, “a particular constitution of epistemic authority and political legitimacy” whereby some subjects are depoliticized into the realm of the “expert” (p. 1164). The concept of technification aligns closely with our observations in past work (Greene et al., 2019) and above: that most AI vision statements or sets of ethical principles seek to restrict debates around the societal impacts of AI to a coterie of technical experts whose positions are posited, chiefly by themselves, as technocratic and thus apolitical. Securitization rhetoric thus allows both industry and governments free reign to set the parameters for AI governance on their own terms. In its simultaneous insistence on the inevitability of AI technologies and that experts’ hold responsibility for ethical AI governance, much AI discourse also performs a flavor of hyper-securitization, hyping the disruption being produced by AI systems while disavowing the role of policy choices in accepting the deployment of problematic technologies. And in line with other surveillance technologies such as CCTV cameras, institutional actors are rapidly securitizing AI as a threat in itself if deployed by the wrong hands, while also advocating its use as a means to identify threats to the national polity (Cassiano, 2019; Chen & Cheung, 2017). The securitization of AI discourses and technologies also further exacerbates existing patterns of digital inequality and power asymmetries between the global North and South. Rumman Chowdhury and Abeba Birhane argue the extractive data practices of many AI firms constitute “algorithmic colonialism,” a digital analogue to the exploitative material extraction of natural resources and human capital to which the Global South has been subjected for centuries (Chowdhury, 2019). Algorithmic colonialism entails both the depredations of neocolonial powers (in particular the United States and China) and those of local elites supported by broader networks of global capital (Couldry & Mejias, 2019); it also includes algorithmic settler colonialism, through which settler colonial states look inward to exploit and dispossess the data heritage of indigenous populations (TallBear, 2013). Finally, it relies on globalized digital infrastructures to outsource and occlude the immense amounts of human labor of which “AI” services often actually consist: workers around the world are paid meagerly to clean and tag data, train models, and facilitate other purportedly automated services, while the more privileged developers of these systems reap the lion’s share of the financial reward (Gray & Suri, 2019; Poster, 2019a, 2019b). Securitization discourse, particularly around technification, thus seeks to push AI/ML systems out of the hands of citizens in both the global North and South and to install them as yet another weapon in the game of neoliberal great power politics. More broadly, securitization supports and underpins the various regulatory, technical, and discursive proposals outlined above designed to narrow the range of people and institutions able to have a say in the governance of AI/ML—to varying degrees, the proposed governance mechanisms for these technologies ensure their fate lies with a particular subset of more or less privileged individuals.

Sign in to recut, update, or save this card.