India AI Governance Guidelines

MeitY has released the India AI Governance Guidelines under the IndiaAI Mission with ‘Do No Harm’ as the core principle for the AI Governance framework(here).

Applicability: The guidelines propose a governance framework applicable to all stakeholders in the AI ecosystem.
Four Key Components: (i) Guiding principles on ethical and responsible AI (sutras); (ii) recommendations; (iii) short, medium and long-term action plans and (iv) practical guidelines for industry, developers and regulators.
Key Principles (Sutras): (i) Trust, (ii) People first, (iii) Innovation over restraint, (iv) Fairness and equity, (v) Accountability, (vi) Disclosures and explanations, and (vii) Safety, resilience and sustainability.
Six Main Categories of Risk: (i) Malicious use, (ii) Bias and discrimination, (iii) Transparency failures, (iv) Systemic risks due to disruptions, geopolitical instability, (v) Loss of control over AI and (vi) National security.
Graded Liability: Recommends liability for AI systems based on (i) Function performed, (ii) Level of anticipated risk, and (iii) Degree of due diligence.
Content Authentication: Recommends setting standards for content provenance and authentication; review by the proposed AI Governance Group.
Governance Framework
- AI Governance Group (AIGG): Chair – Principal Scientific Adviser (PSA); Agencies – MeitY, MHA, MEA, DST, DoT; Regulators – TRAI, CCI, DPB, SEBI, RBI; Advisory bodies – NITI Aayog, Office of PSA.
- Technology & Policy Expert Committee (TPEC): To be set up by MeitY, consisting of experts to brief AIGG on matters of national importance in relation to AI policy and governance.
Legislative Amendments: Recommends amendment of the IT Act to clarify the status of AI systems as intermediaries and their obligations and liabilities.
Data Privacy: Identifies the scope of legitimate use exception for AI development under the DPDP Act as a key issue.
Intellectual Property: Recommends Text and Data Mining (TDM) exemption to be considered by the DPIIT committee examining copyright issues.
Reporting & Infrastructure: Recommends a national-level centralised database for incident reporting, voluntary frameworks, and a techno-legal approach such as Data Empowerment and Protection Architecture (DEPA).
Standards: Reports that standards for validation, verification, and benchmarking of AI systems are under development.