India’s AI governance strategy is becoming easier to misread because it does not look like Brussels.
There is no single horizontal AI law doing the dramatic work. There is no giant compliance code with a clean launch date. The real machinery is forming somewhere less theatrical: inside IndiaAI program offices, inter-ministerial coordination, standards work, public compute allocation and sector pilots.
That makes the governance layer more operational than legalistic. India is building the rails before it writes the full statute.
The clearest signal arrived in April, when MeitY constituted the AI Governance and Economic Group. AIGEG is not another discussion forum with a logo. The government describes it as the central institutional mechanism for AI governance policy development and coordination. Its terms include coordinating policy across ministries, departments and sectoral regulators; overseeing cross-sectoral governance issues; reviewing existing mechanisms; promoting responsible deployment in key sectors; studying regulatory gaps; and classifying AI use cases into “deploy”, “pilot” and “defer” categories based on readiness.
That last verb matters. Classify.
Governance becomes real when someone can say this system may be deployed, this one belongs in a sandbox, and this one should wait because the data, legal basis or labour-adjustment capacity is not ready.
The Mission Is Already Spending And Allocating
The AIGEG announcement matters because IndiaAI is no longer only a strategy document.
The IndiaAI Mission was approved with a budgetary outlay of Rs. 10,371.92 crore, spanning compute capacity, datasets, application development, future skills, startup financing and other pillars. MeitY told Parliament in March that more than 38,000 GPUs had been onboarded through the AI compute portal and that 190 projects had been approved under the mission.
That is the important threshold. Once compute, funding and approved projects exist, governance stops being a future consultation question. It becomes a resource-allocation question.
Who gets subsidized compute? Which projects qualify as public-interest AI? Which model builders receive support? Which sectors get pilots first? Those decisions create practical oversight long before a courtroom sees a test case.
IndiaAI’s foundation-model program shows the same pattern. The government says it is supporting twelve organisations and consortia to develop sovereign foundation models, with financial assistance provided in calibrated form to cover compute usage costs. The same update says ten startups were selected for the IndiaAI Startups Global program, and that IndiaAI Data and AI Labs are being expanded through NIELIT sites and approved ITIs and polytechnics.
This is not advisory governance. It is procurement-adjacent governance. The state is shaping the market by deciding what capacity is cheap, what projects are recognized, what training pipelines are funded and what standards those beneficiaries will be expected to meet.
The Regulator Map Is Federated, Not Empty
India’s approach is sometimes described as light-touch because it avoids a single sweeping statute. That misses the design.
The India AI Governance Guidelines recommend a whole-of-government model in which ministries, sectoral regulators, standards bodies and advisory institutions develop and implement governance frameworks together. The action plan names MeitY, MHA, MEA and DoT among government agencies; RBI, SEBI, TRAI and CCI among sectoral regulators; NITI Aayog and the Office of the Principal Scientific Adviser among advisory bodies; and BIS and TEC among standards bodies.
The April AIGEG office memorandum is narrower in membership. It seats senior officials including the Electronics and IT ministers, the Principal Scientific Adviser, the Chief Economic Adviser, NITI Aayog’s CEO, the secretaries of Telecommunications, Economic Affairs, Science and Technology, a National Security Council Secretariat representative and MeitY’s secretary as member convener. Sector regulators are not the core table.
But they are explicitly inside the operating perimeter. AIGEG’s terms require coordination across sectoral regulators. The guidelines go further: sectoral agencies should monitor harms, issue domain-specific guidance, enforce applicable rules, supervise compliance and handle grievances in their domains. In finance, the guidelines point to the Ministry of Finance and the Reserve Bank of India as the implementation owners.
That is a federated model. MeitY coordinates. AIGEG prioritizes and interprets cross-sector questions. Sector regulators keep control over lending, markets, telecom, competition, insurance, health and other regulated domains.
The benefit is speed. India can move AI governance into existing supervisory channels without waiting for a clean statutory monument. The cost is ambiguity. Companies will need to read not only MeitY notices, but also regulator-specific expectations as they emerge. A model used in a chatbot, a hospital workflow and a credit decision will not live under one operational rulebook.
Credit Is The First Hard Test
Small-business lending is where this architecture stops being abstract.
The Finance Ministry said public-sector banks launched a digital-footprint-based Credit Assessment Model for MSMEs in 2025. Between April 1 and Dec. 31, 2025, banks sanctioned more than 3.96 lakh MSME loan applications amounting to more than Rs. 52,300 crore under digital credit-underwriting programs.
That is not a small lab demo. It is production lending volume.
The model-governance question is whether this remains a bank modernization program or becomes the template for supervised AI deployment. The answer is still mixed. The government is clear that digital public infrastructure and AI can strengthen credit access. A May PIB note on AI-led financial inclusion says the Account Aggregator framework is strengthening digital credit infrastructure and improving AI-based credit models. It also says seventeen companies have RBI registration to operate as Account Aggregators.
But lending governance is less forgiving than a startup accelerator. Credit models affect access to capital, pricing, rejection reasons, bias, explainability and grievance redressal. If India wants sector pilots to become durable governance practice, the next step is not another slogan about inclusion. It is a repeatable supervisory pattern: data provenance, model validation, audit logs, adverse-action explanations, human escalation and clear responsibility when a model-based decision harms a borrower.
The IndiaAI governance framework points in that direction. The guidelines call for India-specific risk assessment and classification, common standards, high-risk sandboxes, incident reporting and practical guidance for regulators. AIGEG can coordinate that policy. RBI and the Finance Ministry will have to make it real in lending.
The Statute Is No Longer The Starting Gun
India’s bet is that AI governance should begin where deployment already happens.
That means compute portals, funding approvals, sector pilots, standards bodies, regulator guidance and public-sector procurement. It is less elegant than a single AI Act. It may also be more adaptive in a country where AI will move through banking, agriculture, telecom, health, education, welfare delivery and language infrastructure at different speeds.
The risk is fragmentation. A federated model can become a maze if ministries and regulators interpret AI risk differently. It can also underperform if AIGEG becomes a convening layer without enforcement pull.
The upside is that India is putting governance next to the levers that matter. Subsidized GPUs, foundation-model funding, data platforms, account-aggregator rails and MSME credit workflows are not policy abstractions. They are where firms and public agencies decide what AI systems get built and shipped.
That is why AIGEG matters. Not because it is the final AI regulator. It is not. It matters because it gives India a place to convert broad principles into deployment choices before statute catches up.
For companies building in India, the message is direct. Do not wait for the AI law. Watch the programs. Watch the standards. Watch the regulator that owns your sector.
That is where the rules are already starting to bite.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.