The next AI race is not about models. It is about rules
The country that builds the cleverest machine will not inherit the next era of finance. The country that writes the rules those machines must obey will. India has been handed a rare seat at that table. It may be the last to understand what it is sitting at.
In February 2022, seven Russian banks discovered that the most dangerous thing aimed at them was not a missile. It was a messaging standard. SWIFT, a cooperative based in Belgium that does nothing more dramatic than move secure financial messages between banks, removed them from its network, and in doing so removed them from the world. No money was seized in that act. A standard simply stopped carrying their instructions, and a G20 economy found a part of itself amputated from the global financial body. Iran had learned the same lesson a decade earlier. The instrument was never the technology. It was the rule about who the technology would serve.
This is the fact the conversation about artificial intelligence keeps walking past. We are transfixed by capability, by which laboratory has the largest model, and that contest is real. But it is the visible half, and the less important half. The half that will still be deciding outcomes in twenty years is being settled now, quietly, in standards bodies and regulatory committees that draw no headlines, over the rules that will govern how AI is audited, explained, trusted, and admitted into the financial system. Capability is loud and brief. Standards are silent and permanent.
History is almost monotonous on this point, which is perhaps why it is so easy to ignore. The technology that wins a race rarely keeps the power afterward. Power drains out of the invention and settles into the standard everyone is eventually compelled to adopt, because a standard, once widely held, becomes ruinously expensive to leave. SWIFT did not invent the cross-border payment; it set the grammar in which payments are spoken, and that grammar became leverage no army could replicate. International accounting standards do not record value; they decide what counts as value, which means they decide what a company is worth before any investor forms an opinion. A powerful model enriches whoever owns it; a widely adopted standard governs everyone who comes near it, including those who never chose it. That asymmetry is the entire contest, and beneath the engineering vocabulary the fight over AI standards is a fight over who reaches the market, whose approval confers legitimacy, where compliance costs fall, and on what terms trust is granted or withheld.
In finance this stops being abstract, because finance is the one domain where AI ceases to be a convenience and becomes a question of systemic trust. A model that writes advertising can be wrong cheaply. A model that prices someone's credit, freezes a transaction on suspicion of financial crime, or shapes a supervisor's judgment is wrong expensively, in money and in lives. The rules now being drafted will define what counts as an adequate explanation when a machine refuses a loan, what auditability demands in practice rather than on paper, and what a board must attest before it deploys a system it cannot fully see inside. They will determine how capital moves, how risk is priced, and how trust is manufactured in a financial system increasingly run by systems their own operators do not entirely understand.
Three answers to who writes these rules have taken shape, and each deserves to be taken seriously rather than mocked, because each rests on a coherent theory of where trust comes from. The European Union has chosen comprehensive, risk-based law. Its AI Act, the first binding statute of its kind, sorts systems by the harm they could do and regulates the dangerous ones with something near the severity reserved for medicine or aviation, betting that legitimacy grounded in rights is itself a competitive advantage. The United States, since the start of 2025, has bet the other way; a new executive order swept aside the prior administration's safety-first posture to clear obstacles from the path of innovation, trusting markets and existing regulators to settle trust faster than any central code. China treats the question as an instrument of state strategy, coordinated from the center and bound to national aims, wagering that direction and speed defeat improvisation. Three philosophies, three answers to who decides what a machine may do, all three now being exported as briskly as the technology itself.
What keeps this contest open, rather than already lost, is that finance cannot tolerate three contradictory rulebooks forever. Global banks already pay dearly to straddle the fracture, and the pull toward a common reference layer is real. Whoever helps shape that layer, the thing everyone benchmarks to even when their laws diverge, will be writing rules long after today's models are obsolete.
Which brings the argument to India, and to a claim the familiar narrative gets exactly backward. India is usually cast as the fast follower, sprinting to catch a race led by others. That framing misses what India holds. India is one of a tiny number of countries that has governed digital systems at the scale of a civilization and survived the experience. India Stack, Aadhaar, and the Unified Payments Interface are not talking points. They are working proof that a state can build digital public infrastructure for more than a billion people while holding inclusion, cost, and oversight in tension at once, in public, with the failures visible and answered for. Almost every other nation drafting AI rules is reasoning about scale from the outside. India has had to make scale survive contact with reality. That is not a disadvantage to overcome. It is authority that cannot be manufactured in a committee.
India's recent conduct suggests it half-understands this. Rather than importing Europe's framework or copying America's retreat, it has taken a distinctive line: national AI governance guidelines, set out in late 2025, that rest on a few foundational principles and push operational detail down to the regulators who already know their sectors. In finance the Reserve Bank of India's committee on responsible AI reported in 2025 with concrete expectations rather than slogans, board-approved AI policies, hardened audit frameworks, incident reporting built to expose systemic weakness before it spreads. The Bureau of Indian Standards has adopted ISO and IEC 42001, the first credible international standard for managing AI systems, as its own. And in February 2026 the global AI summit met in New Delhi, the first convened by a nation of the Global South. These are the motions of a country positioning itself to write.
And yet positioning is not authorship, and this is where the unease begins. The window in which standards can still be shaped is narrow and closing, because standards harden fast once enough large institutions have built atop them, after which the cost of changing them climbs beyond reach. India can speak with an authority no theory can match, and offer the world a model of governance that is innovation-friendly, sector-specific, and rooted in infrastructure no rival possesses. The danger is not that India will be excluded. It is subtler, and more characteristic of how these windows are missed: that India will mistake presence for power, will attend the summits and host the gatherings and contribute the principles, and discover only later that hosting the conversation is not the same as authoring its conclusions.
For the deeper truth runs past any one country. The nations that hold power in the age of financial AI will not be the ones that built the most capable machines. They will be the ones that defined the standards through which every machine must pass to be trusted, governed, and let into the global economy. A standard looks like a technical detail right until the morning it becomes the unalterable condition under which everyone else must operate. The countries treating AI rule-making as something to handle later, once the models mature, are repeating the error their forebears made with payments, with protocols, with the definition of profit itself. They will grasp the importance of standards at the exact moment they realize the standards have already been written, by someone else, and that the only choice left is whether to comply.
(Dr Aditya Vikram Kashyap is AI Researcher and an AI Expert based in New York. Kashyap is an award-winning technology leader. His core competencies focus on enterprise-scale AI, digital transformation, and building ethical innovation cultures. Views expressed are strictly his own and do not reflect any entity or affiliations, past or present.)
In February 2022, seven Russian banks discovered that the most dangerous thing aimed at them was not a missile. It was a messaging standard. SWIFT, a cooperative based in Belgium that does nothing more dramatic than move secure financial messages between banks, removed them from its network, and in doing so removed them from the world. No money was seized in that act. A standard simply stopped carrying their instructions, and a G20 economy found a part of itself amputated from the global financial body. Iran had learned the same lesson a decade earlier. The instrument was never the technology. It was the rule about who the technology would serve.
This is the fact the conversation about artificial intelligence keeps walking past. We are transfixed by capability, by which laboratory has the largest model, and that contest is real. But it is the visible half, and the less important half. The half that will still be deciding outcomes in twenty years is being settled now, quietly, in standards bodies and regulatory committees that draw no headlines, over the rules that will govern how AI is audited, explained, trusted, and admitted into the financial system. Capability is loud and brief. Standards are silent and permanent.
History is almost monotonous on this point, which is perhaps why it is so easy to ignore. The technology that wins a race rarely keeps the power afterward. Power drains out of the invention and settles into the standard everyone is eventually compelled to adopt, because a standard, once widely held, becomes ruinously expensive to leave. SWIFT did not invent the cross-border payment; it set the grammar in which payments are spoken, and that grammar became leverage no army could replicate. International accounting standards do not record value; they decide what counts as value, which means they decide what a company is worth before any investor forms an opinion. A powerful model enriches whoever owns it; a widely adopted standard governs everyone who comes near it, including those who never chose it. That asymmetry is the entire contest, and beneath the engineering vocabulary the fight over AI standards is a fight over who reaches the market, whose approval confers legitimacy, where compliance costs fall, and on what terms trust is granted or withheld.
In finance this stops being abstract, because finance is the one domain where AI ceases to be a convenience and becomes a question of systemic trust. A model that writes advertising can be wrong cheaply. A model that prices someone's credit, freezes a transaction on suspicion of financial crime, or shapes a supervisor's judgment is wrong expensively, in money and in lives. The rules now being drafted will define what counts as an adequate explanation when a machine refuses a loan, what auditability demands in practice rather than on paper, and what a board must attest before it deploys a system it cannot fully see inside. They will determine how capital moves, how risk is priced, and how trust is manufactured in a financial system increasingly run by systems their own operators do not entirely understand.
Three answers to who writes these rules have taken shape, and each deserves to be taken seriously rather than mocked, because each rests on a coherent theory of where trust comes from. The European Union has chosen comprehensive, risk-based law. Its AI Act, the first binding statute of its kind, sorts systems by the harm they could do and regulates the dangerous ones with something near the severity reserved for medicine or aviation, betting that legitimacy grounded in rights is itself a competitive advantage. The United States, since the start of 2025, has bet the other way; a new executive order swept aside the prior administration's safety-first posture to clear obstacles from the path of innovation, trusting markets and existing regulators to settle trust faster than any central code. China treats the question as an instrument of state strategy, coordinated from the center and bound to national aims, wagering that direction and speed defeat improvisation. Three philosophies, three answers to who decides what a machine may do, all three now being exported as briskly as the technology itself.
What keeps this contest open, rather than already lost, is that finance cannot tolerate three contradictory rulebooks forever. Global banks already pay dearly to straddle the fracture, and the pull toward a common reference layer is real. Whoever helps shape that layer, the thing everyone benchmarks to even when their laws diverge, will be writing rules long after today's models are obsolete.
Which brings the argument to India, and to a claim the familiar narrative gets exactly backward. India is usually cast as the fast follower, sprinting to catch a race led by others. That framing misses what India holds. India is one of a tiny number of countries that has governed digital systems at the scale of a civilization and survived the experience. India Stack, Aadhaar, and the Unified Payments Interface are not talking points. They are working proof that a state can build digital public infrastructure for more than a billion people while holding inclusion, cost, and oversight in tension at once, in public, with the failures visible and answered for. Almost every other nation drafting AI rules is reasoning about scale from the outside. India has had to make scale survive contact with reality. That is not a disadvantage to overcome. It is authority that cannot be manufactured in a committee.
India's recent conduct suggests it half-understands this. Rather than importing Europe's framework or copying America's retreat, it has taken a distinctive line: national AI governance guidelines, set out in late 2025, that rest on a few foundational principles and push operational detail down to the regulators who already know their sectors. In finance the Reserve Bank of India's committee on responsible AI reported in 2025 with concrete expectations rather than slogans, board-approved AI policies, hardened audit frameworks, incident reporting built to expose systemic weakness before it spreads. The Bureau of Indian Standards has adopted ISO and IEC 42001, the first credible international standard for managing AI systems, as its own. And in February 2026 the global AI summit met in New Delhi, the first convened by a nation of the Global South. These are the motions of a country positioning itself to write.
And yet positioning is not authorship, and this is where the unease begins. The window in which standards can still be shaped is narrow and closing, because standards harden fast once enough large institutions have built atop them, after which the cost of changing them climbs beyond reach. India can speak with an authority no theory can match, and offer the world a model of governance that is innovation-friendly, sector-specific, and rooted in infrastructure no rival possesses. The danger is not that India will be excluded. It is subtler, and more characteristic of how these windows are missed: that India will mistake presence for power, will attend the summits and host the gatherings and contribute the principles, and discover only later that hosting the conversation is not the same as authoring its conclusions.
For the deeper truth runs past any one country. The nations that hold power in the age of financial AI will not be the ones that built the most capable machines. They will be the ones that defined the standards through which every machine must pass to be trusted, governed, and let into the global economy. A standard looks like a technical detail right until the morning it becomes the unalterable condition under which everyone else must operate. The countries treating AI rule-making as something to handle later, once the models mature, are repeating the error their forebears made with payments, with protocols, with the definition of profit itself. They will grasp the importance of standards at the exact moment they realize the standards have already been written, by someone else, and that the only choice left is whether to comply.
(Dr Aditya Vikram Kashyap is AI Researcher and an AI Expert based in New York. Kashyap is an award-winning technology leader. His core competencies focus on enterprise-scale AI, digital transformation, and building ethical innovation cultures. Views expressed are strictly his own and do not reflect any entity or affiliations, past or present.)
