India No Longer Thinks Algorithms Are Neutral

When India's IT Ministry ordered Meta to revamp its algorithms this week, the demand itself was striking. But the legal theory beneath it is the thing that will matter longer. The government is examining whether Meta's use of recommendation systems and paid promotion — the act of deciding what content is shown to whom — disqualifies the company from the intermediary safe harbor that has shielded platforms from liability under Indian law for two decades. The argument is simple and radical: an algorithm that chooses is not a neutral conduit. It is a publisher. And the government has stated it intends to extend this examination to "other major digital platforms operating in India." ^[1] Section 79 of the Information Technology Act has been the quiet architecture of the Indian internet. Passed in 2000, it drew a line between intermediaries — platforms that passively carry third-party content — and publishers, who are liable for what they put out. The line was always a legal fiction, but it was a useful one. The government's new theory dissolves it. If algorithmic recommendation is publication, then every major platform in the country is a publisher, and the safe harbor that made their business models viable in India disappears. This is not a stray regulatory impulse. It is the keystone of a broader project that has been accumulating across domains for the better part of a year — one that recasts the algorithm as something the state has authority to direct, override, and control. The most explicit articulation of this logic sits in the government's approach to compute. A white paper released by the Principal Scientific Adviser in December calls for "treating AI infrastructure as a shared national resource," warning that compute, datasets, and model ecosystems are "concentrated in a few global firms and urban centres." ^[2]

For India, democratising access means treating AI infrastructure as a shared national resource, empowering innovators across regions to build local-language tools, adapt assistive technologies, and create solutions aligned with India's diverse needs. — Office of the Principal Scientific Adviser to the Government of India

The IndiaAI Mission, launched in May, adopted a public-private partnership model explicitly patterned on UPI — the state provides foundational compute and regulatory guardrails, private firms build applications. ^[3] In June, IT Minister Ashwini Vaishnaw inaugurated a Jabil manufacturing plant for AI data-center components, calling domestic production of the hardware layer "crucial." ^[4]

Manufacturing all the major electronic components of AI data centres in India is crucial. — Ashwini Vaishnaw

The government is building compute capacity as a public utility. Separately, it is asserting regulatory authority over content platforms — the three-hour takedown mandate for government-flagged content, the 43 OTT platform blocks, the Meta algorithm-revamp order. ^[5][6][1] The two moves are distinct in their mechanics, but they share a premise: algorithmic capacity is not a private market good the state merely polices from the outside. It is infrastructure the state builds and directs. Two other domains have moved in the same direction, though the evidence points to convergence rather than coordination. In June, the Reserve Bank of India proposed a mandatory "kill switch" for AI models in financial institutions — an emergency-stop mechanism that would let the central bank instantly override or deactivate algorithmic systems. ^[7] The same month, the Supreme Court drafted regulations mandating that "no judicial outcome shall be reached through Algorithmic Decision-Making alone," establishing a three-tier oversight structure and human primacy over AI-assisted legal reasoning. ^[8]

no judicial outcome shall be reached through Algorithmic Decision-Making alone. — Supreme Court of India

In July, the Court voided decisions by two lower tribunals that had relied on AI-hallucinated legal precedents, declaring "zero-tolerance" and insisting that human decision-makers maintain "absolute and total control" over judicial reasoning. ^[9]

It is necessary for Courts to adopt a zero-tolerance mode for producing, citing or using AI-generated precedents without verification. — Supreme Court of India

No public document links the RBI's kill switch or the Supreme Court's AI rules to the IT Ministry's content-moderation agenda. But the logic is the same: the algorithm is not a tool the state defers to. It is a force the state asserts authority over. That is where the clean story ends. Because the judiciary is not merely aligning with the executive's algorithmic sovereignty project. It is contesting it. In March, the Bombay High Court struck down the government's Fact-Check Unit rule — a provision that would have let a state-appointed body declare content false and trigger its removal — as unconstitutional for vagueness and for chilling free speech. The Supreme Court refused to stay that order, insisting on "clear demarcated guidelines." ^[10]

In view of the majority opinion, Rule 3 (1) (V) is declared unconstitutional and is struck down. The petitions are accordingly allowed — Bombay High Court

The same Supreme Court that voided AI-hallucinated judgments and drafted rules for human primacy over algorithms has also been building its own, parallel architecture of algorithmic control. In July, it banned unauthorized social media clips of court proceedings after Solicitor General Tushar Mehta warned that AI voice-cloning could put fabricated words in judges' mouths. ^[11]

What I never said was thrust upon me. — Surya Kant

In February, it restrained WhatsApp from sharing user data with Meta, with the Chief Justice calling the 2021 privacy policy "a decent way of committing theft of private information" and warning the company to "leave India" if it could not follow the Constitution. ^[12]

You can’t play with privacy… we will not allow you to share a single digit of our data. — Supreme Court of India

And in November, it called for an independent, autonomous social media regulator, explicitly saying it would not approve rules used to "gag dissent." ^[13]

Self-styled bodies will not be effective… There has to be some impartial, autonomous body… which will be free from the influence of those who are exploiters of all these, and at the same time, the state also. — Supreme Court of India

The algorithm has been recast as a sovereign object — like spectrum or airspace. The fight is over who exercises that sovereignty. The executive is building the infrastructure and writing the rules. The courts are striking down the rules they find unconstitutional while writing their own. Neither branch is arguing that algorithms should be left alone. The international dimension sharpens what is distinct about India's approach. The European Union is regulating algorithms aggressively — the Digital Services Act imposes binding takedown and transparency rules on very large platforms, and the AI Act phases in risk-based classifications with penalties of up to seven percent of global turnover. ^[14][15] A Dutch court recently ordered xAI to stop generating sexualized deepfakes. ^[16] Some of India's content-side mandates — the deepfake bans, the accelerated takedown timelines — parallel this global regulatory wave. But the sovereignty framing runs deeper. The EU regulates platforms it does not build. India is building compute capacity as a public utility while separately asserting regulatory authority over content platforms, and in February it signed an AI pact with Germany that explicitly prioritizes "technological sovereignty" to create alternatives to US-China dominance. ^[17]

Sovereignty plays a crucial role. — Karsten Wildberger

Former Foreign Secretary Nirupama Rao framed the logic plainly: AI is "an organizing principle of international power," and India must pursue "strategic agency: the ability to shape options, influence standards and make choices that others must take seriously." ^[18]

The objective must be strategic agency: the ability to shape options, influence standards and make choices that others must take seriously. — Nirupama Rao

The EU's model is regulation. India's is something more ambitious — and more contested than either side of that ambition admits, because the two branches building it are also fighting over who holds it.