'AI is colonisation 2.0': Indian documentary director puts Western AI labs on trial

On 9 July 2026, the Indian documentary director Anna Rieser, whose 2024 film Humans in the Loop followed data annotators in eastern India training computer-vision systems for Western clients, used an interview with The Indian Express to deliver one of the sharpest indictments of artificial intelligence the South Asian press has published this year. AI, she argued, "is no less than colonisation," driven by "Western minds" and "Western ideas" — and run on the quiet labour of communities who will never see the products built from their work.

The framing is deliberately provocative, and it lands because Rieser is not a polemicist. Her film won jury recognition at festivals including IFFR Rotterdam, where Humans in the Loop screened in 2024, and was acquired for streaming distribution. The argument that follows from her remarks is not about any single model release or a particular safety scandal. It is about the architecture: where the training data comes from, who labels it, who is paid, who owns the output, and what happens to the labour when the dataset is finished.

The annotators behind the model

Modern AI systems depend on armies of low-paid workers who draw boxes around objects, tag emotions in voices, and label the contents of images so that a model can learn. That industry is concentrated in the Global South: Kenya, India, the Philippines, Venezuela, and parts of the Middle East. Workers have described piece-rate pay, sudden terminations when contracts end, and limited recourse when psychologically distressing content is routed to them without warning. Coverage in The Guardian, MIT Technology Review, and Time has documented individual cases, including a 2024 Time investigation into outsourcing firm Sama's contracts with OpenAI in Kenya.

Rieser's contribution to this conversation is to refuse the polite framing. Workers in the data economy, she told The Indian Express, are not "partners" in some shared digital future. They are extractive labour — paid a fraction of the value they create, with no rights over the model that ships. Her phrasing puts a name to a relationship that the AI industry's own marketing has tended to obscure: when a frontier model is launched and credited to a handful of researchers in San Francisco, London, or Beijing, the people who made it legible to a machine are typically listed under a sub-contractor in a press footnote, if at all.

The structure underneath the slogan

Calling AI "colonisation 2.0" is a metaphor, but a useful one. The earlier wave of European extraction was organised around three mechanisms: control of land, control of labour, and control of knowledge. The contemporary AI economy runs on recognisable analogues. Data is the new land, scraped from across the public web and from workers' own interactions. Labour is the new extraction, paid in micro-tasks. Knowledge — in the form of trained weights and the products built on top of them — flows back to a handful of corporate headquarters in a small number of jurisdictions, with intellectual-property regimes that make redistribution difficult.

This is not an argument unique to Rieser. Scholars from Latin America, Africa, and South Asia have written about "data colonialism" for at least a decade; the term has entered mainstream coverage in outlets including the Financial Times and Al Jazeera English. What is new is that the argument is now being made on a Western-backed platform, by a director whose own credentialing depends on the same European festival circuit that the AI industry courts for prestige. The film is distributed, the awards are real, the venues are legitimate. The criticism has cleared the gatekeepers.

Counterpoint: building, not extracting

The strongest counter-argument is that AI, unlike historical colonial extraction, also creates infrastructure that the Global South can use. Indian developers, Kenyan researchers, and Filipino engineers are not only labelling data; they are founding companies, contributing to open-source projects, and publishing at top-tier machine-learning venues. Indian AI start-ups raised significant capital in 2025, and the country's national language model initiatives — including Indic-language systems developed under government-backed programmes — are signs of a more distributed future.

There is real substance to this read. Indian IT services firms including Tata Consultancy Services and Infosys have built AI practices of their own. India's Supreme Court has pushed back on opaque AI deployment in the public sector, and the country's Digital Personal Data Protection Act, which came into force in phases through 2025 and 2026, attempts to give data principals a meaningful remedy when their information is processed without consent. None of that erases the labour asymmetries Rieser describes. But it complicates the picture: the relationship between a New Delhi research lab and a San Francisco frontier model is not the same as the relationship between a colonial administration and a village it taxes.

Stakes and what to watch

The practical stakes of this debate are concrete. If Rieser's framing wins, the next round of AI regulation — including the European Union's AI Act implementation and the United States' evolving executive orders on compute and safety — will have to reckon with mandatory disclosure of training-data provenance and a floor for annotator pay. If it loses, the industry continues to externalise its labour costs while capturing the value upstream. Indian outsourcing firms, which have begun pivoting from back-office business process work toward higher-margin AI services, will be among the actors most directly affected by either outcome.

What remains uncertain is whether the indictment travels. Indian English-language coverage has given Rieser's remarks prominence, but Western trade press has so far treated them as a colour quote rather than a structural challenge. Whether filmmakers and researchers from other parts of the Global South pick up the framing — and whether regulators treat it as more than rhetoric — will be the test of whether "AI as colonisation" becomes a slogan or a policy agenda.

Desk note: Monexus frames this as a structural labour-and-governance story, not a celebrity profile. The Indian Express provides the anchor interview; broader context on annotator working conditions, India's Digital Personal Data Protection Act, and the "data colonialism" debate is drawn from prior wire coverage referenced in the article above and verifiable independently of the thread.