India AI Governance Guidelines, 2025

The Guidelines set seven principles and give advice in six areas. They say India does not need a separate AI law right now, because existing laws already cover many risks.

LawThe Guidelines add no new duty. The laws you already follow still apply to your AI, such as the IT Act and the DPDP Act.

Good practiceThe Guidelines ask businesses to follow all Indian laws, adopt voluntary frameworks, publish transparency reports, offer a way to complain, and use technical tools to reduce risk.

Good practiceWatch for changes. The Guidelines suggest changing the IT Act and copyright law. Those changes would need new law.

A drafting committee set up by MeitY in July 2025 wrote the Guidelines. The seven principles (the document calls them sutras) are adapted from RBI's FREE-AI committee report. The Guidelines name all seven in the preface (PDF page 4) and credit the RBI FREE-AI report in Part 1 (PDF page 7, endnote ii). They propose the AIGG and TPEC as new bodies (PDF page 68). They say RBI's committee published its principles in August 2025, and that they were "suitably adapted" to work across sectors.

The seven principles

1. Trust is the Foundation — People will not use AI they cannot trust.
2. People First — Design AI around people, with humans in control.
3. Innovation over Restraint — When all else is equal, favour careful innovation over holding back.
4. Fairness & Equity — AI should include everyone and not discriminate.
5. Accountability — Someone must be clearly responsible, and rules must be enforced.
6. Understandable by Design — Users and regulators should be able to understand what the AI does.
7. Safety, Resilience & Sustainability — Systems should be safe, hard to break and kind to the environment.

Advice in six areas

• Infrastructure — Wider access to data and computing power, and use of India's digital public platforms.
• Capacity building — Education, skills and training about AI's risks and uses.
• Policy & regulation — Review current laws, find gaps, and fix them with targeted changes.
• Risk mitigation — An India-specific way to judge risk. Mostly voluntary steps, with extra duties in sensitive uses.
• Accountability — Liability graded by role, risk and whether you took due care. Enforce the laws that already apply.
• Institutions — Ministries and regulators work together. Sector regulators keep their enforcement powers.

Three proposed bodies

• AI Governance Group (AIGG) — a proposed group to coordinate AI policy across government.
• Technology & Policy Expert Committee (TPEC) — a proposed group of experts to advise the AIGG.
• AI Safety Institute (AISI) — already set up. The Guidelines' preface says the IndiaAI Mission "has instituted" the AISI (PDF page 4). They say it should get resources for research, testing and safety advice.

What is advice, and what would need law

• Advice only: the principles, voluntary frameworks, risk frameworks and transparency reports.
• Needs government action: setting up the AIGG and TPEC, and funding the AISI.
• Needs new law: the suggested changes to the IT Act (how AI systems are classed and who is liable) and to copyright law (training AI on large amounts of data).

The action plan lists short-, medium- and long-term steps. Long-term, it includes drafting new laws for new risks. It gives no dates.

• We have not found an official notice setting up the AIGG or the TPEC.
• No bill to change the IT Act or copyright law for AI has been found by us.

Sector regulators can make their own rules. For example, SEBI's Regulation 16C is binding. RBI's FREE-AI report is advice, not a rule.