AI opacity entrenches corporate power: when unaccountable systems evade democratic scrutiny

Indigenous epistemologies reject AI’s claim to objectivity, framing 'justification' as a relational act requiring consent, reciprocity, and accountability to community needs. Models like the Māori data sovereignty framework (e.g., the 2018 Declaration on Indigenous Data Sovereignty) assert that data cannot be 'mined' without permission, challenging the extractive logic of Silicon Valley’s AI. Traditional knowledge systems (e.g., Andean *ayni* or Ubuntu) prioritise collective benefit over individual profit, offering a blueprint for AI that serves people, not platforms.

The history of 'scientific' racism (e.g., 19th-century craniology) shows how 'objective' systems have long been weaponised to justify oppression, a pattern repeating in modern AI. The eugenics movement’s legacy persists in today’s predictive policing algorithms, which replicate biased historical data to target marginalised groups. The Colorado lawsuit echoes 20th-century cases where corporations (e.g., IBM’s role in apartheid South Africa) used technology to evade accountability, framing discrimination as an unintended side effect.

In Japan, the concept of *wa* (harmony) critiques AI’s adversarial design, where systems are built to 'win' rather than collaborate, exacerbating social fragmentation. African philosophical traditions (e.g., Ubuntu’s 'I am because we are') demand AI that centres communal well-being over individualistic metrics like 'efficiency.' The Chinese principle of *ren* (benevolence) challenges Silicon Valley’s 'move fast and break things' ethos, advocating for AI that repairs rather than disrupts social fabrics.

Peer-reviewed studies (e.g., Buolamwini & Gebru, 2018) demonstrate that commercial facial recognition systems exhibit racial and gender bias, with error rates up to 34.7% for darker-skinned women. The 'black box' problem is not just technical but epistemological: AI’s lack of explainability stems from its reliance on proprietary data and opaque training processes, which prioritise scalability over transparency. Recent advances in interpretable AI (e.g., SHAP values, counterfactual explanations) offer pathways to auditability, but are rarely adopted by dominant firms.

Artists like Trevor Paglen expose AI’s hidden infrastructures through photography, revealing how 'neutral' systems encode violence (e.g., surveillance drones trained on war zones). Spiritual traditions (e.g., Hindu *maya* or Buddhist *dukkha*) frame AI’s illusions as manifestations of collective delusion, where technology obscures interdependence. The Dadaist tradition of absurdity (e.g., Marcel Duchamp’s readymades) offers a lens to critique AI’s pretensions to objectivity, highlighting its arbitrary and often harmful classifications.

Scenario modelling by the AI Now Institute predicts that by 2030, unregulated AI could deepen racial wealth gaps by 20% in the US, while EU-style audits could reduce bias by 40%. Futures where AI is treated as a public utility (e.g., municipal data trusts) contrast with dystopian outcomes where corporations monopolise decision-making, exacerbating authoritarianism. The rise of 'algorithmic sovereignty' movements (e.g., Indigenous data cooperatives) suggests a path where communities reclaim agency over AI systems.

Black feminist scholars (e.g., Safiya Noble) have long documented how AI reproduces misogynoir, from search algorithms to hiring tools, yet their work is sidelined in mainstream tech discourse. Disabled activists argue that 'accessibility' in AI is often an afterthought, with systems designed for abled users first, reinforcing exclusion. Global South researchers (e.g., from the African Centre of Excellence for Information Ethics) highlight how AI’s 'universal' standards erase local contexts, prioritising Western epistemologies over diverse knowledge systems.

Anansi the spider, as trickster, would expose Musk’s lawsuit as a performative act—his companies’ AI (e.g., Tesla’s Autopilot) already discriminates against darker-skinned pedestrians in crash tests. Hermes, the Greek trickster, would note how AI’s 'justification' is a sleight of hand: corporations claim neutrality while their systems are trained on stolen data (e.g., Clearview AI’s facial recognition). Bakhtin’s carnivalesque lens reveals AI’s 'black box' as a grotesque inversion of democracy, where algorithms perform the role of 'neutral' judges while serving corporate interests.