Work

Governing AI Without Governing Speech: The Ontopolitics of the Onopticon, Human Rights, and the Global South

Year: 2026
Type: article
Abstract: Artificial intelligence is moving from a tool that transmits speech to an infrastructure that ranks, predicts and pre-configures the field in which speech and belief can appear. This article argues that AI governance must control the power of technology without becoming a mechanism that governs what people think, say or believe. It traces the theoretical shift from the Panopticon of Bentham and Foucault, through Deleuze's societies of control and Zuboff's surveillance capitalism, to Hakkı Tan's Onopticon: a regime of infrastructural imposition whose operating formula is esse est verificari—to be is to be verifiable—and whose characteristic subject is the Homovictimus. In that setting, freedom of expression is threatened less by overt deletion than by shadowbanning, algorithmic downranking and generative interference with the forum internum protected by Article 19(1) of the International Covenant on Civil and Political Rights. The article then sets out a human-rights architecture that treats platform and model operators as information fiduciaries, binds content moderation to the Santa Clara Principles and the Manila Principles, and insists on linguistic equity. Published tokenization audits show that, on a controlled technical-tutoring corpus, Bengali text requires about 1.56 times as many tokens as equivalent English under a leading commercial tokenizer, and about 4.5 times as many under widely used open-weight tokenizers, imposing cost, latency and context-window penalties on one of the world's largest language communities. The closing sections locate those penalties within data colonialism and argue for digital sovereignty, digital public goods and South–South standard-setting, so that technology is made accountable to people rather than people to technology.
Language: English
Open Access status: green
OpenAlex ID: W7233613747
FWCI:
Cites:
Cited by:
Related to: