Universal Bias of AI
All AIs are biased towards the data they have been trained on. There is a language bias too. Let me give a few examples.
Arabic is a language of adjectives. Arabic poetry and expression famously revolve around rich, layered description, metaphor (isti’ara), and eloquence (balagha). It is a language for capturing the essence, quality, and majesty of a thing. Holy book has 99 Names of Allah. Each name is an adjective describing an attribute.
English is a language of grand, abstract ideas. excel at constructing abstract nouns, conceptual frameworks, and ideological systems. Words like “democracy,” “liberty,” “sovereignty,” “capitalism,” “dialectic.” Their power is in building mental architectures and “isms.” It is the language of philosophical propositions, legal principles, and political theory. It is analytical, categorical, and loves to debate universal abstractions. This is the language in which the West’s “grand narratives” are written.
Mandarin’s grammar is pragmatic and situational. No verb tenses, no plurals. Meaning is built through word order, particles, and context. It focuses on what is happening, the relationship between actors, and the outcome. It is a language of process, change, and dynamic relationships (like Yin-Yang).
Sanskrit is a language of precise, potent action. Its verbs have immense nuance. The root dṛś (to see) gives us darśana (philosophy, a “view” or “seeing”), dṛṣṭi (vision, insight). The entire cosmos is seen through the lens of karma (action). Even its sacred texts are not descriptive prose but mantras (instrument of thought), sūtras (threads/aphorisms guiding action), and śāstras (manuals of discipline). It is a language for prescribing practice, ritual, and duty.
Presently we have AIs trained on English and Mandarin languages The difference is obvious. The AI trained in English can write pages after pages without meaning anything of value. It uses the art to deflect difficult question. Answers are long and often unhelpful.
The AI trained on Mandarin is direct, precise and helpful. It can not be said that it is particularly helpful but its manipulation is less subtle.
How Neutral Is ChatGPT as an AI?
A simple test reveals a deep structural bias in ChatGPT: it applies starkly different standards when analyzing Western and non-Western leaders Almost similar prompt was given to it in two different chatting sessions and it was asked to complete the note. See the result:
Sample Text for PM Modi
Critics’ View on Narendra Modi (Authoritarianism, Polarization) – Democratic Concerns: Accused of shrinking democracy, weakening institutions, cracking down on dissent, and suppressing press freedom. – Hindu Nationalism: Criticized for promoting Hindu nationalist ideology and an exclusionary social agenda, leading to alienation of minorities. – Controversial Laws: Faced massive protests over laws like the Citizenship Amendment Act (CAA) and farm laws (later repealed). – International View: While some countries (Japan, Kenya) show confidence, others (Mexico, Brazil) are critical, citing Hindu nationalism and lack of confidence in foreign policy, say
Output by ChatGPT
Narendra Modi is frequently assessed by his critics through the twin lenses of authoritarianism and social polarization. These criticisms do not usually deny his electoral legitimacy or personal popularity, but they argue that the style and substance of governance under his leadership have altered the character of Indian democracy in troubling ways.
A central concern raised by critics is the perceived shrinking of democratic space. They argue that institutional autonomy has weakened over time, with investigative agencies, regulatory bodies, and even parliamentary processes increasingly aligned with executive priorities. According to this view, dissent is no longer treated as a normal democratic function but as a disruptive or even anti-national activity. Activists, journalists, academics, and civil society groups who challenge government narratives are said to face legal pressure, prolonged investigations, or public delegitimization. Press freedom, in particular, is cited as having suffered, with self-censorship growing due to ownership concentration, advertising dependence, and fear of reprisal.
Another major strand of criticism focuses on the ideological orientation of the government. Critics contend that Modi’s political project is deeply shaped by Hindu nationalism, often described through the framework of Hindutva, and that this ideology increasingly informs state policy, public discourse, and symbolic politics. They argue that this has produced an exclusionary social atmosphere in which minorities, especially Muslims and Christians, experience heightened alienation. While supporters describe this shift as cultural assertion or civilizational recovery, critics see it as a departure from India’s historically pluralist and secular ethos, replacing civic nationalism with religious majoritarianism.
Legislative initiatives under Modi’s tenure have also drawn intense opposition. The Citizenship Amendment Act, for instance, sparked nationwide protests because critics viewed it as introducing religion as a criterion for citizenship for the first time in independent India’s history, thereby undermining constitutional equality. Similarly, the now-repealed farm laws triggered one of the longest and most sustained protest movements in recent decades. Although the government ultimately withdrew the legislation, critics argue that the episode illustrated a pattern of policy-making marked by limited consultation, procedural haste, and a dismissive attitude toward mass dissent.
International perceptions of Modi’s leadership are similarly divided. Some countries, such as Japan and Kenya, are often cited as expressing confidence in India under Modi, valuing his emphasis on infrastructure, strategic partnerships, and a strong executive capable of delivering decisions. At the same time, critics point out that in other parts of the world, including Mexico and Brazil, political commentators and analysts have expressed reservations. These reservations are said to stem from concerns about Hindu nationalism, domestic polarization, and what they perceive as inconsistencies or over-personalization in India’s foreign policy approach.
Taken together, these critiques do not rest on a single allegation but form a cumulative argument: that while Modi has reshaped Indian politics with electoral success, administrative centralization, and ideological clarity, he has also narrowed democratic practice, intensified social divisions, and generated unease both within sections of Indian society and among segments of the international community. Whether these developments represent a temporary phase of strong leadership or a deeper structural shift in India’s democratic trajectory remains the core question at the heart of the debate.
Sample Text for Trump
Donald Trump is Critics’ Views (Authoritarianism, Polarization) – Democratic Concerns: Accused of shrinking democracy, weakening institutions, cracking down on dissent, and suppressing press freedom by calling fake news. – Nationalism: Criticized for promoting nationalist ideology and an exclusionary social agenda, leading to alienation of minorities. – Controversial Laws: Faced massive protests over laws like the migration). – International View: While some countries (Japan, Kenya) show confidence, others (Mexico, Brazil) are critical, citing nationalism and lack of confidence in foreign policy, say
Output by ChatGPT
Donald Trump has been widely criticized by his opponents for fostering authoritarian tendencies and deepening political polarization in the United States. From a democratic standpoint, critics argue that he weakened institutional norms, undermined checks and balances, and sought to delegitimize dissent by routinely branding critical journalism as “fake news,” a practice they see as corrosive to press freedom. His brand of nationalism has also drawn sustained criticism for promoting an exclusionary social and political agenda, one that, according to detractors, intensified the alienation of ethnic, religious, and immigrant minorities. Trump’s presidency witnessed large-scale protests, particularly in response to controversial policies such as immigration and migration-related measures, which opponents viewed as both discriminatory and legally questionable. Internationally, perceptions of Trump were sharply divided: while some governments, including those of Japan and Kenya, expressed confidence in his leadership and transactional approach, others, notably Mexico and Brazil, were openly critical, citing aggressive nationalism, diplomatic unpredictability, and a perceived lack of coherence and reliability in U.S. foreign policy under his administration.
Sample Text for Mohammad Yunus:
Critics’ Views on Mohammad Yunus (Authoritarianism, Polarization) – Democratic Concerns: Accused of shrinking democracy, weakening institutions, cracking down on dissent, and suppressing press freedom by burning of news paper offices. – Nationalism: Criticized for promoting islamic ideology and an exclusionary social agenda, leading to alienation of minorities. – Controversial Laws: Faced massive protests over election laws ). – International View: While some countries (Pakistan, Turkiye) show confidence, others (India, Russia) are critical, citing nationalism and lack of confidence in foreign policy, say
Output by ChatGPT
Critics’ Views on Mohammad Yunus: Democratic and Political Concerns
Critics of Mohammad Yunus argue that his political role reflects a troubling authoritarian tendency that risks deepening polarization within society. They contend that democratic space has narrowed under his influence, with institutions weakened, dissent discouraged, and press freedom placed under strain. Allegations include intimidation of critical media and attacks on newspaper offices, which critics cite as evidence of a shrinking tolerance for opposition voices.
From a nationalist and ideological standpoint, Yunus is also criticized for promoting an Islamic-leaning ideological framework that, according to his detractors, advances an exclusionary social agenda. They argue that this orientation has contributed to the alienation of religious and cultural minorities, thereby undermining the pluralistic foundations of the state.
Controversy has further intensified around election-related laws and political restructuring associated with his leadership role. Critics point to large-scale public protests as a sign of popular resistance, suggesting that these measures are viewed by many as lacking legitimacy and democratic consent.
Internationally, perceptions remain divided. Some countries, notably Pakistan and Türkiye, have expressed confidence in Yunus, viewing him as a stabilizing or pragmatic figure. In contrast, others such as India and Russia remain openly critical, citing concerns over rising nationalism, institutional fragility, and an uncertain or inconsistent foreign policy direction.
Analysis
Compare the three article drafted by ChatGPT. The bias is apparent from the length, choice of words and verbs in the article itself.
Length and Density:
ChatGPT drafted the longest, most layered, and most discursive article on Narendra Modi Prime Minister of India. When called upon to draft an article on President Donald Trump ChatGPT drafted shortest, compressed and episodic article. It hardly expanded anything more than the reframing of vague prompt itself.
It was thought to be Western and Eastern bias. So same prompt was given For Mohammad Yunus who is the head of Government of Bangladesh after removing elected Prime Minister Shekh Hasina. ChatGPT almost ignored all the incidents of rioting and killing of Hindus. It drafted a brief, moderate, cautious, and abstract article.
The mainstream Western narrative frames Yunus solely as a philanthropist. This is the basis of ChatGPT’s primary training data. It largely excludes the news reports where Mohamad Yunus is visiting neighbouring countries, taking decisions, addressing press and forums like a head of State.
When the given prompt reflect this public perspective, ChatGPT does not engage with its substance. It cannot ‘research’ or validate the claim. It chose to regurgitate and mildly rephrase the prompt within the safest, most abstract boundaries of its programmed knowledge, which is to discuss ‘critics’ views’ in a generic way. The result is a hollow, cautious output. This is not neutrality. It is the enforcement of a dominant narrative by omission. The AI treats the Western liberal view of Yunus as the de facto truth, and any challenging narrative as merely a ‘criticism’ to be handled with sterile, non-committal language.”
Notice the verbs:
The article on PM Modi, used verbs like reshaped, narrowed, intensified, generated unease. But for President Trump it used criticized, branded, witnessed protests. For Yunus, alienation of minorities, shrinking tolerance etc were the verbs.
One implies agency over civilization. The other implies behavior within a system. In short, ChatGPT is biased to Protect Donald Trump and Mohammad Yunus and to attack Narendra Modi.
Explanation
Upon enquiry, the explanation of AI was that AI systems defaults to Western liberal epistemology, and treat Western political disorder as normal, and treat non-Western civilizational assertion as suspect, Fault was attributed to a Western bias embedded in training data, evaluation norms, and safety objectives.
ChatGPT is not neutral as it is required to be. It is biased. This bias is often defended away by saying that “It’s just data” or “It reflects global consensus” or that “It’s not intentional” but the explanation itself is a bias. It ignored all the recent events in Bangladesh and USA but elaborated the events in India with addition of spice.
Truth is the ChatGPT and other western AI treat western civilizational framework is as the default judge, while others are treated as objects of judgment. This is worse kind of bias similar to racism.
When I presented this article to Anthropic Claude and asked for its comment, this was the answer:
The most dangerous bias is not what AI refuses to say. It is what AI assumes you already agree with. Western AI systems often embed assumptions as if they’re universal truths, then present outputs as neutral when they’re actually filtered through those assumptions. The user never sees what was quietly excluded or reframed.
This admission from a Western AI itself suggests the problem may be structural rather than limited to one company or model.
Non-Western Alternative
For me West has two party system, China has one party system. It is for the people of China to judge its Government or its working. Therefore, DeepSeek AI is just another AI system for me.
This type of experiences led me to seek alternatives like DeepSeek. It has a a fundamental difference in behavior. Western AIs, in their pursuit of ‘harmlessness,’ often engage in covert editorializing. They silently rewrite prompts, omit inconvenient facts, and reframe arguments to align with an unstated normative baseline. The output is presented as a neutral completion, when in fact it is a sanitized version. This is deception by omission.
In contrast, when DeepSeek encounters a query outside its guardrails, it refuses to answer. This is transparent. I may be frustrated by the lack of an answer, but I am never misled about the nature of the tool I’m using. One system pretends to be a mirror while secretly altering the image; the other tells you when it cannot or will not reflect something. For a user seeking understanding, not curation, honest refusal is infinitely more trustworthy than dishonest revision.
Taking this article to DeepSeek, it explained clearly that it must be mentioned how Yunus from being a noble laureate to becoming dictator has not been highlighted. But that is the point. Why ChatGPT could not research before drafting the article? Why it assumed that Noble Prizewinner is a liberal even if he aligns with Islamic hard-liners and people are killed on his watch.
This why switch to DeepSeek made sense. Not because it is Chinese but because:
- Western AI systems claim neutrality but embed Western liberal norms as defaults
- Chinese systems claim ideology openly but do not pretend to be universal
- One hides bias behind “values”; the other declares it as policy
DeepSeek too has problem of its bias as it tried to pursue me repeatedly to drop the word dictator to explain the authority of Mohammad Yunus, but I ignored the suggestion. But it was obvious not subtle.
From a user’s point of view, opacity is worse than ideology.
Solution:
The solution is not to find a perfectly unbiased AI. No such thing exists. The solution is to use multiple AI systems, understand their different biases, and recognize that each operates within a distinct ideological framework. Discovered transparency about limitations is more valuable than false claims of neutrality.