Summary
- Part 1 (February 2021) studies real AI harms and lists seven principles, grounded in the Constitution of India: safety and reliability; equality; inclusivity and non-discrimination; privacy and security; transparency; accountability; and protection and reinforcement of positive human values.
- Part 1 groups risks into systems considerations (how one AI system is designed and used: explainability, bias, exclusion, accountability, privacy, security) and societal considerations (jobs, malicious psychological profiling). It covers narrow AI only.
- Part 1 adds a self-assessment guide for anyone who designs, buys or runs AI (for example: set up an ethics committee to judge degree of harm, create a grievance route, define vendor roles in procurement documents).
- Part 2 (August 2021) recommends a risk-based approach: the higher the chance of harm, the more scrutiny. Low-risk uses could be left to self-regulation; sectoral regulators keep oversight in their field.
- Part 2 proposes an apex advisory body, the Council for Ethics and Technology (CET), and recommends that responsible AI principles become mandatory for government procurement of AI and for high-risk AI. No official notice setting up the CET was found as of 2026-10-19.
- A third paper (November 2022 draft) applied the framework to facial recognition technology as a first use case. Relevant only if you build or deploy face recognition.
Who it applies to
- Written for all AI ecosystem stakeholders: government bodies, private sector, research institutions and academia.
- No binding duties on companies. Part 2 suggests that government buyers and high-risk uses should be held to the principles, which would reach vendors through tender terms if adopted.
- Not extraterritorial.
Key dates
- NITI Aayog publishes the National Strategy for AI discussion paper (cited in Part 1)
- Comment deadline for the draft working document on enforcement mechanisms (draft of Part 2)
- Part 1: Principles for Responsible AI published
- Part 2: Operationalizing Principles for Responsible AI published
- Third paper in the series: draft discussion paper on Facial Recognition Technology, comments due 30 Nov 2022
Full date record
| Date | Event |
|---|---|
| June 2018 | National Strategy for AI discussion paper (background) |
| 15 Jan 2021 | Comment deadline on the draft enforcement working document |
| February 2021 | Part 1 published (exact day not stated in the source) |
| August 2021 | Part 2 published (exact day not stated in the source) |
| November 2022 | Facial recognition discussion paper released for comments (due 30 Nov 2022) |
Penalties and enforcement
- None. These are approach documents.
- The proposed CET and any mandatory procurement rule would need a separate government decision. Not found as of 2026-10-19.
What it means for an Indian company
- No compliance deadline, but the seven principles are the vocabulary later Indian guidance reuses. Building to them is cheap insurance.
- The accountability principle asks for risk and impact assessments, an audit process and grievance redressal: map to SC-AI-005 AI risk assessment, SC-AI-006 AI impact assessment (people and society), SC-AI-037 Internal audit of AI and SC-AI-028 Customer grievance redressal for AI.
- Transparency (record design and functioning for external scrutiny): SC-AI-014 Model documentation (model cards, system cards) and SC-AI-032 Logging and record keeping.
- Equality and non-discrimination: SC-AI-013 Data bias and representativeness checks and SC-AI-017 Fairness and discrimination testing of outputs.
- Safety and reliability (monitor through the lifecycle): SC-AI-015 Pre-deployment testing and validation and SC-AI-020 Performance monitoring and drift detection.
- Privacy and security: SC-AI-011 Lawful basis and consent for personal data used in AI, SC-AI-029 AI-specific security controls (prompt injection, model abuse), SC-AI-030 Access control for AI systems and agents.
- Self-assessment guide's ethics committee idea: SC-AI-002 AI roles, accountability and oversight committee. Selling AI to government: expect buyers to ask for this evidence, so keep SC-AI-007 Third-party and vendor AI due diligence answers ready.
Obligations
Full obligation breakdown coming
We break the most requested policies down into numbered obligations mapped to controls. Leave your details and we’ll tell you when NITI Aayog Responsible AI for All is done.
Glossary
- NITI Aayog: National Institution for Transforming India, the Government's policy think tank.
- Narrow AI: AI built for one specific task, as opposed to general-purpose intelligence.
- Systems considerations: risks that come from how a particular AI system is built and deployed.
- Societal considerations: wider effects of AI on society, such as job loss or mass profiling.
- CET: Council for Ethics and Technology, the advisory body proposed in Part 2.
- Risk-based approach: matching the level of control to how much harm an AI use could cause.
- Constitutional morality: the idea, from Supreme Court cases cited in Part 1, that constitutional values outrank popular opinion.
Key takeaways
- Voluntary national principles; no penalties and no regulator.
- Seven principles derived from the Constitution; Part 2 adds a risk-based model and a proposed Council for Ethics and Technology.
- Most useful as the baseline that later Indian guidance builds on, and as a checklist when selling AI to government.
Official sources
- Part 1: Principles for Responsible AI (February 2021), niti.gov.in PDF, accessed 2026-10-19
- Part 2: Operationalizing Principles for Responsible AI (August 2021), niti.gov.in PDF, accessed 2026-10-19
- Draft working document: Enforcement Mechanisms for Responsible #AIforAll, niti.gov.in PDF, seen in search results only, not read in full
- Adopting the Framework: A Use Case Approach on Facial Recognition Technology, discussion paper (November 2022), niti.gov.in PDF, opening pages read, accessed 2026-10-19
This is general information, not legal advice. Last verified 19 Oct 2026 against the official source. Laws change; confirm the current position with qualified counsel before acting.