The New Orientalism: How Artificial Intelligence is reproducing colonial stereotypes

On Edward Said’s birthday, the question is no longer simply who controls the narrative about the East, but who controls the machines producing it.

KARACHI (The TKP Report): When Edward Said published Orientalism in 1978, he challenged the intellectual foundations through which the West had historically described the East. He argued that Western scholarship, literature and political discourse had constructed an image of Eastern societies as backward, irrational, exotic and incapable of governing themselves, while presenting the West as rational, progressive and civilized.

Nearly five decades later, that argument has acquired a new dimension. The institutions producing knowledge have expanded beyond universities, publishing houses and media organisations. Artificial intelligence companies, algorithmic systems and digital platforms are increasingly becoming influential authorities in determining how societies are described, understood and represented.

The question Said raised has therefore entered the digital age: who has the authority to define the East?

Artificial intelligence is frequently presented as a technological breakthrough capable of making information accessible, reducing human prejudice and democratizing knowledge. However, beneath its sophisticated interfaces lies a complicated reality. AI systems are trained on enormous collections of human-generated information, including historical literature, news reports, photographs and online discussions. These sources carry the political, cultural and historical assumptions of the societies that produced them.

Consequently, technology does not necessarily eliminate inherited prejudices. It can reproduce them, sometimes at a scale and speed previously unimaginable.

When algorithms inherit the colonial gaze

A particularly important question is whether generative AI has inherited some of the representational patterns that Said identified in Western scholarship.

A study published in the journal AI & SOCIETY in November 2025, titled ‘We can see a savage: a case study of the colonial gaze in generative AI algorithms’, examined more than 3,800 AI-generated captions of historical photographs depicting colonized people displayed in European and American human exhibitions.

Using Midjourney, researchers found that the generated descriptions reproduced several patterns associated with colonial representation, including cultural erasure, stereotyping, dehumanisation, othering and infantilization.

The findings are significant because they suggest that even modern AI systems can interpret historical images through perspectives shaped by colonial-era assumptions.

A separate study published in August 2026, titled ‘Between reflection and construction: AI as the new Orientalism?’, examined 2,460 ChatGPT-generated texts and 100 AI-generated images. It reported higher toxicity scores for prompts associated with Arabic personas compared with Western personas, alongside stereotypical visual representations of Arabs and Jews.

These findings do not establish that every AI system behaves identically. However, they demonstrate that technological advancement does not automatically guarantee cultural neutrality.

Palestine and the politics of digital representation

The Palestinian question provides another important context for understanding the relationship between AI, representation and power.

International perceptions of Palestine are shaped not only through traditional journalism and political statements but increasingly through digital platforms and automated information systems.

A 2026 study published in Third World Quarterly, examining ChatGPT outputs in English and Arabic, investigated anti-Palestinian bias in AI-generated content. The researchers reported patterns involving biased language and omissions of historical and legal context, including Palestinian refugee status, human rights and the right of return.

This raises a fundamental concern about the authority of artificial intelligence as an information source.

When millions of people use AI systems to understand international conflicts, historical events and political disputes, omissions can become as consequential as misinformation. A system may produce grammatically convincing and apparently balanced explanations while leaving out historical experiences or perspectives necessary for understanding an issue.

The danger is not limited to Palestine. Any society whose history, language and political experiences are inadequately represented in the information used to develop AI systems risks being described through incomplete external perspectives.

Pakistan and the question of digital colonialism

For Pakistan and other countries of the Global South, the debate extends beyond cultural representation to technological dependence.

Most influential AI models and their underlying infrastructures are developed and controlled by companies operating in technologically advanced economies. Their training resources, computing infrastructure, research capabilities and commercial priorities are concentrated in a relatively small number of countries.

This creates an important imbalance. Countries such as Pakistan are increasingly becoming consumers of AI-generated knowledge without necessarily having comparable influence over the data, systems and institutional decisions shaping that knowledge.

A 2026 review published in AI & SOCIETY, analysing 50 peer-reviewed studies on AI ethics in post-colonial contexts, identified algorithmic colonialism, data colonialism and platform imperialism as interconnected mechanisms through which existing global inequalities can be reproduced.

The implications are particularly relevant for Pakistan. Consider how artificial intelligence represents a Pakistani village, an urban neighbourhood, a flood-affected community or the country’s economic institutions. If the underlying information is incomplete or dominated by external representations, the resulting picture may reduce a complex society to poverty, disaster, instability and dependency.

Indeed, a 2025 study examining AI-generated climate imagery of Pakistan identified a recurring visual pattern centred on dust, floods and suffering children, while overlooking institutional agency and the country’s geographical diversity.

Such representation risks presenting the Global South primarily as a location of crises rather than a source of knowledge, innovation and solutions.

The battle is over knowledge, not merely technology

However, it would be intellectually simplistic to suggest that artificial intelligence is inherently Western or that every system deliberately reproduces colonial thinking. AI can also help preserve endangered languages, translate regional literature, document indigenous knowledge and make academic resources accessible to communities previously excluded from global knowledge networks.

The central issue is therefore not whether Pakistan or other developing countries should embrace artificial intelligence. It is whether they can participate meaningfully in producing, governing and shaping it.

This requires investment in local-language datasets, independent research institutions, technological infrastructure and AI governance frameworks that reflect regional histories and social realities. It also requires greater scrutiny of how international technology companies collect information and whose knowledge receives priority.

For Pakistan, the challenge is particularly important because technological dependence can gradually become intellectual dependence when societies rely exclusively on external systems to explain their own histories, economies and identities.

Edward Said’s Orientalism was fundamentally concerned with the relationship between knowledge and power. Artificial intelligence has made that relationship more complicated, not less relevant.

On his birthday, the enduring significance of Said’s intellectual contribution lies in a question that has moved from colonial libraries to digital laboratories: who produces knowledge, whose experiences are included, and who has the authority to define the rest of the world?

The Orientalist gaze once travelled through books, paintings and imperial institutions. Today, it can also travel through datasets, algorithms and artificial intelligence.

The technology is new. The struggle over representation is not.