Country Overview
India's approach may be shifting. As recently as November 2025, MeitY's India AI Governance Guidelines were explicitly non-binding and light-touch. On 6 July 2026, MeitY Secretary S. Krishnan signaled that the government intends to prepare a draft AI law that could designate banking, finance, health and critical infrastructure as a high-risk tier, using hedged language ('probably the time has come now to look at a separate legislation'). Reporting around the same date indicates the government is weighing emergency shutdown powers for AI systems in that tier, a notably assertive tool few other jurisdictions have proposed — though as of this writing no draft bill, consultation paper, or timeline has been published, so the precise scope of any such law remains unconfirmed.
What is already binding is narrower: the IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules 2026, in force since February 2026, mandate labelling and provenance metadata for synthetic content and a 3-hour takedown window for unlawful AI content (2 hours for non-consensual intimate imagery) — down sharply from a prior 36-hour window. Digital-rights group IFF has reportedly criticised those rules as opaque with overbroad takedown triggers, though this specific attribution has not been independently re-verified against a primary IFF statement.
At a Glance
Overall AI policy direction
Signaling a possible shift toward binding, risk-based legislation
3
Major AI companies
1
Research institutions
3
Live debates
- Latest development
- August 5, 2026
- Regulatory approach
- Officials have signaled intent to draft a binding, risk-based AI law; IT Rules already binding for synthetic content
- Major AI companies
- Krutrim, Sarvam AI, Neysa
- Major cloud providers
- AWS, Microsoft Azure, Google Cloud +2 more
- Leading research institutions
- AI4Bharat (IIT Madras)
- Key ministries
- Ministry of Electronics and Information Technology (MeitY)
- Government involvement
- Rising — from voluntary guidelines toward signaled intent to draft binding legislation, though no formal proposal has been published
- Industrial strategy
- IndiaAI Mission subsidised compute (38,000+ GPUs onboarded) supporting startups pursuing lower-cost ('frugal') AI approaches
- International partnerships
- AI Impact Summit host; engages G20 and UN AI governance processes
- Primary policy focus
- High-risk-sector AI safety (banking, finance, health, critical infrastructure) is among the areas officials have flagged for a possible future law
Current Direction
Tightening
Officials' public rhetoric has moved from 'no separate AI law needed' toward openness to drafting one within eight months, with a senior MeitY official using hedged language about timing and no draft bill yet published. Reporting indicates the government is also weighing emergency shutdown powers for high-risk AI systems, a notably assertive tool most peer jurisdictions haven't proposed — though the specific scope and design of any such power remain unconfirmed.
Recent Developments
The government gave the Lok Sabha a consolidated implementation update (5 August 2026) on the February 2026 IT Intermediary Guidelines amendments covering synthetically generated information: prominent labelling and traceable metadata obligations for AI-generated content, takedown of unlawful content within three hours of government notice (two hours for impersonation or nudity grievances), and a clarification that generative AI services' safe-harbour status depends on their function.
2026-08-05MeitY Secretary S. Krishnan signals intent to prepare a draft binding, risk-based AI law, suggesting banking/finance/health/critical-infrastructure could form a high-risk tier; reporting also indicates the government is weighing emergency shutdown powers — a shift in tone from its November 2025 non-binding stance, though no draft bill has been published.
2026-07-06SourceIAPPIndia's private-sector AI-infrastructure build-out accelerates: Reliance announced (February 2026 AGM) a reported ₹10 lakh crore (~$110B) seven-year AI push anchored by a multi-gigawatt, renewable-powered data centre in Jamnagar (first phase targeting 3GW, with an initial ~120MW slated to come online in late 2026), signed its first AI data-centre deal with Meta in June 2026, and is working with NVIDIA on an India-focused foundation model; Adani has announced comparably large plans, together positioning India as a major AI-infrastructure hub.
Major2026-06IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules 2026 take effect, mandating labelling and provenance metadata for synthetic content and a 3-hour takedown window for unlawful AI content.
Major2026-02-20SourceMondaqDigital-rights group IFF is reported to have criticised the IT Rules amendment as opaque with overbroad takedown triggers; the specific quote/attribution has not been independently re-verified.
2026SourceThe420.inIT Secretary S. Krishnan says government is actively considering a dedicated standalone AI law given deepfake and cyber-risk growth.
2026SourceThe420.inThe IndiaAI Safety Institute, established under the IndiaAI Mission's Safe & Trusted AI pillar, advances: MeitY approved a further tranche of responsible-AI projects (reported as around 13) targeting deepfakes, bias and evaluation tooling, and issued a call for research partnerships, though the institute's full staffing and independent capacity remain unconfirmed.
2026SourceDD NewsMeitY unveils India AI Governance Guidelines under the IndiaAI Mission — non-binding, proposing an AI Governance Group, Technology & Policy Expert Committee, and an IndiaAI Safety Institute.
2025-11SourcePress Information Bureau
Policy and Regulation
Binding risk-based AI law (signaled, not yet drafted)
On 6 July 2026, MeitY Secretary S. Krishnan signaled intent to prepare a binding AI law that could designate banking, finance, health and critical infrastructure as a high-risk tier, and reporting indicates the government is weighing emergency shutdown powers for AI systems in that tier. No draft bill, consultation paper, or timeline has been published as of this writing.
IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules 2026
Mandatory labelling and provenance metadata for synthetic content; 3-hour takedown window for unlawful AI content on government or court order (2 hours for non-consensual intimate imagery), down from 36 hours previously.
India AI Governance Guidelines
Non-binding, light-touch, principle-based guidelines proposing an AI Governance Group, Technology & Policy Expert Committee, and an IndiaAI Safety Institute. Whether these bodies have actually been constituted/staffed since the November 2025 announcement is unconfirmed, and officials have since signaled openness to binding legislation that could eventually supersede parts of this approach.
Research Ecosystem
Applied AI research in India includes AI4Bharat (an IIT Madras initiative building Indian-language AI tooling since 2020) and the IndiaAI Mission's subsidised compute portal, which has onboarded more than 38,000 GPUs for startups and researchers under a ₹10,372 crore (~$1.25B) programme. No dedicated AI-focused think tank was identified in this pass, though several general policy institutes plausibly do AI-relevant work.
Major AI Companies
Krutrim
India's first AI unicorn, founded by Ola founder Bhavish Aggarwal. As of mid-2026 it has pivoted away from its original foundation-model and chip ambitions — pausing model/chip work, pulling its consumer assistant 'Kruti' from app stores (April 2026), and cutting headcount from 550+ to roughly 150-160 — repositioning primarily as an AI cloud-services provider.
Sarvam AI
Bengaluru-based AI company founded by ex-IIT Madras researchers. Released the open Sarvam-105B model (105B-parameter mixture-of-experts, 128K context, 22 Indian languages) alongside Sarvam-30B around the February 2026 AI Impact Summit.
Neysa
Indian AI infrastructure company; raised a Blackstone-led financing of up to $1.2B in February 2026 to scale GPU-based AI cloud infrastructure, targeting 20,000+ GPUs.
Cloud Infrastructure
- Frontier AI
- Emerging / sovereign models
- Compute
- Scaling fast — IndiaAI Mission's 38,000+ subsidised GPUs plus Reliance/Adani multi-GW data-centre plans
- Chip access
- Import-reliant on NVIDIA; no advanced domestic fabrication yet
AWS
Major cloud provider for Indian enterprise AI workloads.
Microsoft Azure
Cloud provider with significant Indian enterprise and government integration.
Google Cloud
Cloud provider used across Indian research and enterprise AI.
Neysa
Indian AI infrastructure company; raised a Blackstone-led financing of up to $1.2B in February 2026 to scale GPU-based AI cloud infrastructure, targeting 20,000+ GPUs.
Reliance Jio (Jio Intelligence)
Building large-scale 'sovereign' AI compute, anchored by a multi-gigawatt renewable-powered data centre in Jamnagar; announced a ~$110B AI push (2026), a June 2026 AI data-centre deal with Meta, and an India-focused foundation-model collaboration with NVIDIA.
Leading Universities
Think Tanks
No think tanks recorded yet.
Government Institutions
Notable Gaps
AI-focused think tanks
No dedicated AI-policy think tank is listed for India. General policy institutes such as ICRIER, Observer Research Foundation, and Carnegie India plausibly do AI-relevant work, but their current AI-specific activity was not verified in this pass.
Key People
Ashwini Vaishnaw
GovernmentMinister of Electronics and Information Technology
S. Krishnan
GovernmentIT Secretary
MeitY
People and Organisations to Follow
S. Krishnan
GovernmentIT Secretary; the clearest official voice on India's shift toward binding AI legislation.
Internet Freedom Foundation
Think tankLeading Indian digital-rights group scrutinising IT Rules enforcement.
AI4Bharat
UniversityIIT Madras initiative building open Indian-language AI tooling.
Current Debates
GovernmentIndia signals a possible shift toward binding AI legislation
Within eight months, official rhetoric shifted from 'no separate AI law needed' toward openness to binding legislation with a high-risk sector tier, though no draft bill has yet been published.
▾
India signals a possible shift toward binding AI legislation
Within eight months, official rhetoric shifted from 'no separate AI law needed' toward openness to binding legislation with a high-risk sector tier, though no draft bill has yet been published.
MeitY's November 2025 guidelines were explicitly non-binding. On 6 July 2026, MeitY Secretary S. Krishnan used hedged language ('probably the time has come now to look at a separate legislation') signaling intent to prepare a binding, risk-based AI law — potentially covering banking, finance, health and critical infrastructure as high-risk. One secondary summary states explicitly that no draft bill, consultation paper, or timeline exists yet. If it materializes, this would mark a genuine change in direction, plausibly driven by mounting deepfake incidents and cyber-risk concerns, though it's not yet clear whether this reflects a broader philosophical shift on innovation-friendliness or a narrower, sector-specific response.
SecurityIT Rules 2026 takedown triggers criticised as overbroad by digital-rights groups
The sharply shortened takedown windows for AI-generated content have reportedly drawn criticism from at least one digital-rights group (IFF) for opaque, overbroad triggers that could sweep in lawful content, though this specific attribution has not been independently re-verified.
▾
IT Rules 2026 takedown triggers criticised as overbroad by digital-rights groups
The sharply shortened takedown windows for AI-generated content have reportedly drawn criticism from at least one digital-rights group (IFF) for opaque, overbroad triggers that could sweep in lawful content, though this specific attribution has not been independently re-verified.
Reporting attributes to the Internet Freedom Foundation, and possibly other digital-rights groups, the argument that the rules' takedown triggers lack clear definitions and due-process safeguards, risking over-removal of legitimate content. The government has defended the shortened windows (down from 36 hours to 2-3 hours) as necessary given the speed at which deepfakes and non-consensual imagery spread. The precise wording and attribution of the IFF criticism have not been independently confirmed against a primary IFF statement.
SecurityEmergency shutdown powers for high-risk AI systems
India is reportedly weighing emergency shutdown powers over AI systems in high-risk sectors — a notably assertive regulatory tool few other jurisdictions have proposed, though the precise scope and design remain unconfirmed.
▾
Emergency shutdown powers for high-risk AI systems
India is reportedly weighing emergency shutdown powers over AI systems in high-risk sectors — a notably assertive regulatory tool few other jurisdictions have proposed, though the precise scope and design remain unconfirmed.
If adopted, emergency shutdown powers would let authorities forcibly halt AI systems operating in banking, finance, health or critical infrastructure in the event of serious malfunction or harm. A 6 July 2026 official source indicates the government is 'weighing' such powers, corroborated by secondary reporting, but no draft bill or consultation paper exists yet, so the specific scope, sectors covered, and legal design remain unconfirmed. Should it proceed as described, this would go further than most peer jurisdictions' AI laws, which generally rely on fines and compliance orders rather than direct operational shutdown authority.
Knowledge Hub
Sources
Deliberately sourced beyond government publications. The bars below show each category’s actual share of this page’s citations — an internal editorial guideline rather than a rigorously derived target, so treat it as a rough steer, not a precise benchmark.
- IT Rules 2026: deepfake regulation and AI labelling obligationsGovernmentMondaq2026-02
- MeitY unveils India AI Governance GuidelinesGovernmentPress Information Bureau2025-11
- IT Secretary on a dedicated AI lawNewsThe420.in2026
- IndiaAI Mission overviewResearchobcrights.org2026
- India's frugal AI approach — Sarvam and KrutrimExpert commentaryRest of World2026
- India signals shift on AI regulationNewsIAPP2026-07-06
- India's first GenAI unicorn shifts to cloud services as AI model ambitions face realityNewsTechCrunch2026-05-05
- Reliance unveils $110B AI investment plan as India ramps up tech ambitionsNewsTechCrunch2026-02
- Meta signs first AI data center deal in India with RelianceNewsTechCrunch2026-06
- IndiaAI Mission's Safe & Trusted AI pillar advances; 13 projects approvedGovernmentDD News2026
- Generative AI Safe Harbour Status Dependent On Function: GovernmentNewsOutlook India2026-08-06
- Govt strengthens framework to curb AI-generated deepfakes; mandates faster takedownNewsIANS2026-08-06