India’s AI Governance Guidelines Explained

Verified against the primary source — MeitY’s "India AI Governance Guidelines: Enabling Safe and Trusted AI Innovation" PDF, published via the Press Information Bureau — and cross-checked against MediaNama’s reporting; last checked September 2026. On November 5, 2025, India’s Ministry of Electronics and Information Technology (MeitY) released the final version of its India AI Governance Guidelines. The headline framing, in the words of MeitY Secretary S. Krishnan, was that the government has taken "a conscious and deliberate approach of not leading with regulation." That is not spin: the document itself, in its own preface, describes India’s approach as "a techno-legal framework supported by voluntary measures and Digital Public Infrastructure (DPI)" — not a law, and not a regulator with penalty power.

What MeitY Actually Released

The Guidelines are the output of a drafting committee MeitY convened in July 2025, working from a mandate to recommend a governance framework that "promotes innovation and adoption of AI in India while mitigating the risks to individuals and society." That committee did not start from a blank page: MeitY had already published an earlier draft report, AI Governance Guidelines Development, in January 2025, and opened it for public comment. According to MediaNama’s reporting, the government received over 650 inputs on that January draft, which the July 2025 committee then reviewed before producing the final document. (Separately, an earlier round of consultation on a 2023-era sub-committee report drew a much larger 2,500-plus submissions — the 650-plus figure is specific to feedback on the January 2025 draft that immediately preceded this final release.)

The final Guidelines’ preface is signed by Prof. Ajay Kumar Sood, Principal Scientific Adviser to the Government of India, who also chairs the high-level advisory group under which the drafting committee was constituted. The document runs to four parts: seven guiding principles, six pillars of recommendations, an action plan with short/medium/long-term steps, and practical implementation guidance for industry and regulators.

Voluntary by Design, Not by Accident

MeitY’s "not regulating AI" framing is easy to read as a dodge. It is not: it is the document’s actual operating premise, stated plainly and repeatedly. The Guidelines do not create new offenses, new penalties, or a new AI regulator. Instead, they recommend amending existing law where gaps show up — the guidelines specifically flag the two-decade-old Information Technology Act, 2000 as needing updates to how it classifies AI developers, deployers, and users, and how liability under Section 79’s intermediary-immunity provisions should apply to systems that generate or modify content rather than just transmit it. Enforcement, where it exists at all, is left to sector regulators already in place — the Reserve Bank of India, SEBI, TRAI, and the Competition Commission of India among them — applying their existing powers to AI-specific facts, not a new AI-specific statute.

The Seven Sutras

Part 1 of the Guidelines sets out seven guiding principles — the document calls them "sutras" — explicitly adapted from the Reserve Bank of India’s FREE-AI Committee report (the RBI’s own August 2025 "Framework for Responsible and Ethical Enablement of AI," produced by a separate committee). MeitY’s document generalizes those principles from the financial sector to AI governance across all sectors:

• Trust is the Foundation — without trust, innovation and adoption stagnate.
• People First — human-centric design, human oversight, human empowerment.
• Innovation over Restraint — all else equal, responsible innovation is prioritized over cautionary restraint.
• Fairness & Equity — promote inclusive development and avoid discrimination.
• Accountability — clear allocation of responsibility and enforcement of regulations.
• Understandable by Design — disclosures and explanations the intended user and regulators can actually understand.
• Safety, Resilience & Sustainability — safe, secure, robust systems able to withstand systemic shocks and remain environmentally sustainable.

Six Pillars, Three Domains

Part 2’s recommendations are organized around six pillars, which the document itself groups into three domains: enablement (infrastructure, capacity building), regulation (policy & regulation, risk mitigation), and oversight (accountability, institutions).

• Infrastructure — expand access to compute and data, attract investment, and use Digital Public Infrastructure for scale and inclusion.
• Capacity Building — education, skilling, and training programs to build public trust and AI literacy.
• Policy & Regulation — balanced, agile frameworks; review existing law for AI-specific gaps and address them with targeted amendments rather than a new statute.
• Risk Mitigation — an India-specific risk-assessment framework grounded in real-world evidence of harm, with additional obligations reserved for sensitive applications or vulnerable groups.
• Accountability — a graded liability system based on function, risk level, and due diligence, with existing laws enforced and guidelines helping organizations meet their obligations.
• Institutions — a whole-of-government approach coordinated through the three bodies below.

Three Institutions — Two Proposed, One Already Running

The Institutions pillar is where the Guidelines get concrete about who is actually supposed to do this work.

The AI Governance Group (AIGG) is a proposed new inter-ministerial body, chaired by the Principal Scientific Adviser, with a mandate to coordinate AI policy across ministries and sectoral regulators, oversee national AI-governance initiatives, and study emerging risks and regulatory gaps. Its suggested composition spans MeitY, the Ministry of Home Affairs, the Ministry of External Affairs, DST, DoT, TRAI, CCI, the Data Protection Board, NITI Aayog, and sectoral regulators such as RBI, SEBI, ICMR, and UGC.

The Technology & Policy Expert Committee (TPEC) is a proposed advisory body MeitY would convene to support the AIGG — a small panel of experts in frontier-technology R&D, machine learning and data science, law and public policy, public administration, and national security, briefing the AIGG on emerging AI capabilities, regulatory gaps, and India’s international AI-policy engagements.

The AI Safety Institute (AISI) is different from the other two in one important respect: it is not a proposal. It already exists, instituted under the India AI Mission, and the Guidelines describe it as providing "the critical technical expertise needed to conduct research, develop draft standards, and perform safety testing." Under the Institutions pillar, AISI is the body earmarked to supply the AIGG and TPEC with technical grounding on trust-and-safety questions, while sector regulators keep their existing enforcement powers. Worth noting: India’s AISI does not currently appear among the ten member institutes of the international coordination network now called NAAIMES (Australia, Canada, the EU, France, Japan, Kenya, South Korea, Singapore, the UK, and the US) — it is a separate, purely domestic body for now.

No Separate AI Law — and the Digital India Act Stays Stalled

The Guidelines are not a substitute for a comprehensive AI statute, and as of MediaNama’s most recent reporting (December 17, 2025), India is not planning one. MeitY Secretary S. Krishnan put it directly: the government intends to "avoid putting in a new law, a new regulation, unless you absolutely have to," relying instead on amendments to existing statutes — the IT Act and the Digital Personal Data Protection Act chief among them. The long-promised Digital India Act, which has been discussed as a possible vehicle for comprehensive AI rules since 2023, remains unintroduced in Parliament and has had no public draft released; per MediaNama, references to it have largely disappeared from official government communications as policymakers pivot toward this voluntary-guidelines-plus-existing-law approach instead. Practically, that means there is currently no binding, AI-specific statute in India — only these Guidelines, sector-regulator advisories, and whatever amendments eventually get made to the IT Act and the DPDP Act.

How CASRAI’s NIKOLAI Tracks This

CASRAI’s own NIKOLAI project is an independent, unendorsed dictionary of frontier-AI-safety elements, organized into ten tracks (N1–N10). The most relevant track here is N1, Actors, models and scope, which is where NIKOLAI catalogues the national and organizational actors that frontier-AI-safety frameworks name — developers, model identifiers, and the bodies that set or enforce scope.

To be precise about what does and doesn’t exist today: we checked NIKOLAI’s N1 element pages directly, including Coverage scope threshold — N1’s element for the if-then test that decides whether a framework applies to a given developer or model — and neither India’s AI Safety Institute nor the proposed AI Governance Group appears in that element’s crosswalk table, or in any other NIKOLAI crosswalk we could find. That crosswalk currently runs Anthropic, OpenAI, Google DeepMind, Meta, xAI, the EU, California, and a few others; no Indian body is on it. This is not an oversight to correct quietly — it’s a genuine gap. AISI and AIGG are exactly the kind of national institutional actors N1 exists to catalogue, and they are strong candidates for a future N1 crosswalk row once NIKOLAI’s coverage extends to national-institution mapping. Until CASRAI adds and verifies that row, treat this as a candidate, not an existing NIKOLAI mapping.

Separately, India’s voluntary, no-penalty-regime design is a useful additional data point for the binding-vs-voluntary axis used in CASRAI’s jurisdiction-map comparison of AI regulatory approaches: it sits closer to Japan’s no-penalty AI Promotion Act than to the EU AI Act’s binding, fine-backed regime, though India’s approach differs from Japan’s in relying on existing sectoral regulators’ enforcement powers rather than a single new law with no penalty clause at all.

Frequently Asked Questions

Is India’s AI Governance Guidelines a law?

No. It is a voluntary, non-binding policy framework released by MeitY. It creates no new offenses or penalties and does not establish a new AI regulator. Where the Guidelines identify legal gaps, they recommend amending existing statutes — principally the IT Act, 2000 — rather than passing new AI-specific legislation.

What are the seven sutras?

Trust is the Foundation, People First, Innovation over Restraint, Fairness & Equity, Accountability, Understandable by Design, and Safety, Resilience & Sustainability. They are adapted from the Reserve Bank of India’s FREE-AI Committee report and generalized from the financial sector to AI governance across all sectors.

What is the AI Governance Group (AIGG)?

A proposed permanent inter-ministerial body, chaired by India’s Principal Scientific Adviser, meant to coordinate AI policy across ministries and sectoral regulators. As of the Guidelines’ November 2025 release, it has not yet been formally constituted — it is a recommendation, supported by a proposed Technology & Policy Expert Committee (TPEC).

Is India’s AI Safety Institute (AISI) the same kind of body as the UK’s or the US’s?

It serves a similar technical-expertise role — research, draft standards, and safety testing — and, like those institutes, it already exists (instituted under the India AI Mission, rather than proposed). Unlike the UK AI Security Institute and the US CAISI, it is not currently listed among the ten members of the international coordination network now called NAAIMES.

Does NIKOLAI have a crosswalk entry for India’s AI Governance Guidelines, AISI, or AIGG?

No. We checked NIKOLAI’s N1 track and its Coverage scope threshold element directly, and no Indian institution appears in any current crosswalk. AISI and AIGG are candidates for a future N1 row, not an existing NIKOLAI mapping — NIKOLAI is CASRAI’s own independent, unendorsed reference project, not a body that certifies or represents any government’s guidelines.

Is India planning a separate, binding AI law like the EU AI Act?

Not currently. Per MediaNama’s reporting as of December 2025, MeitY has said it intends to avoid new AI-specific legislation where existing law can be amended instead, and the previously discussed Digital India Act — a possible vehicle for comprehensive AI rules — remains stalled, with no draft introduced in Parliament since 2023.