India’s silent AI law — Written not in Parliament, but in spam complaints

India’s spam crackdown has quietly become something far bigger than a fight over unwanted phone calls. It has become the country’s first real stress test of how — and whether — to regulate Artificial Intelligence itself. And judging by the clash erupting between the Telecom Regulatory Authority of India (TRAI), the nation’s biggest banks, its loudest fintechs, and its most powerful telcos, the government is deliberately keeping its legislative cards close to the chest.

The unspoken pilot project
Officially, TRAI’s draft amendments to the 2018 Telecom Commercial Communications rules are about curbing 2.5 million annual spam complaints and the 13 scam messages the average Indian now receives daily. Unofficially, they are a live-fire experiment in AI governance. Empower an algorithm to penalize senders on the basis of just three complaints in ten days, and you have written — in all but name — India’s first enforceable AI law. New Delhi appears to understand this. Which is precisely why it is moving through sectoral regulators like TRAI rather than through Parliament. Keep the options open. Test the mechanics. Let the market push back. Legislate later.

Banks become reluctant AI ethicists
In a role reversal nobody predicted, it is now HDFC Bank, PhonePe, and DHL Express — not civil society or academics — leading the charge on AI accountability. HDFC’s data privacy officer warned TRAI on April 13 that AI detection “may result in false positives, particularly for high-volume legitimate banking communication.” PhonePe called the thresholds “too harsh.” The subtext is unmistakable: when your OTP is one server hiccup away from being flagged as spam, the abstract debate about “algorithmic fairness” becomes a very concrete boardroom crisis.

The fintech industry, through FACE, has even tried to rescue a single word — “inquiry” — from TRAI’s edit pen, arguing that a user who abandons a loan application midway has not forfeited the right to a follow-up call. One deleted word, 380 companies lobbying, millions of half-completed digital journeys in the balance. If this is what pre-legislation looks like, imagine the fight when an actual AI Act lands.

The telco paradox
Bharti Airtel and Reliance Jio — who would most benefit from an AI-first enforcement regime — are also asking for restraint. Airtel wants penalties tied to “high-certainty proof.” Jio wants “corroborated evidence.” Both are quietly acknowledging what the government will not say on record: today’s AI is not good enough to be trusted as judge and jury, even when it’s built by and for the telecom industry itself. That admission, buried in submissions to TRAI, may be the most important sentence in India’s AI policy debate this year.

The 50-paise Trojan horse
Meanwhile, the real battle is commercial. Telcos want 50 paise per minute on bulk voice calls, 5 paise per call attempt, and full inclusion of 1400/1600 series numbers under deterrent charges. IAMAI says this would push costs onto consumers and should be left to private contract. TRAI’s draft caps termination charges at 5 paise per minute. Behind the math is a deeper question the government seems keen to defer: who pays the price of AI governance — the platforms, the senders, or the citizens?

Keeping the legislative powder dry
By routing this debate through TRAI’s consultation process rather than a standalone AI statute, the government gains three things. It gets a real-world sandbox where banks, telcos, and fintechs battle-test AI enforcement at scale. It collects written positions from every major digital stakeholder — HDFC, Jio, Airtel, PhonePe, DHL, IAMAI, FACE — that will become the foundation of any future national AI bill. And it preserves political flexibility: if AI enforcement works, expand it. If it misfires, scale it back. No legislation to repeal. No Parliament to face.

The bigger picture
India has long said it will pursue “light-touch” AI regulation, unlike the EU’s AI Act or China’s algorithm filing regime. The spam crackdown shows what “light-touch” looks like in practice: sectoral, iterative, consultative, and — crucially — legally reversible. The government is not writing an AI law. It is watching one write itself, through the friction between regulators and the regulated.

To conclude, the spam war of 2026 was never really about spam. It is India’s first public negotiation over who gets to deploy AI at scale, who bears the cost of its mistakes, and how much algorithmic authority the state is willing to delegate. By keeping options open on future AI legislation, New Delhi is doing something unusual for a government: admitting that it does not yet know the answer — and letting the market, the courts of consultation, and the algorithms themselves reveal it.

For now, the spammers send, the AI flags, the banks protest, the telcos invoice, the consumers complain, and the legislators watch. Everyone has a role. Nobody has a law. And that, increasingly, looks like the plan.

CT Bureau