2.1 | Infrastructure: enable adoption through access to data, compute, evaluation datasets and Digital Public Infrastructure | 0 |
2.2 | Capacity Building: awareness campaigns, training for officials, regulators and law enforcement, and reach into smaller cities | 0 |
2.3a | Policy and Regulation: agile, principle-based frameworks, a review of existing law, targeted amendments and common standards | 0 |
2.3b | Policy and Regulation: content authentication standards, regulatory sandboxes, diplomacy and foresight | 0 |
2.4a | Risk Mitigation: an India-specific risk assessment and classification framework and a national AI incidents mechanism | 0 |
2.4b | Risk Mitigation: voluntary frameworks, secure-by-design systems, techno-legal measures and safeguards against loss of control | 0 |
2.5a | Accountability: clarify value-chain roles, impose proportionate obligations, enforce laws and provide compliance guidance | 0 |
2.5b | Accountability: mandatory grievance redressal with feedback loops, voluntary accountability mechanisms and value-chain transparency | 0 |
2.6 | Institutions: a whole-of-government structure with the AI Governance Group, the Technology and Policy Expert Committee and a resourced AI Safety Institute | 0 |
3 | Action plan and institutional framework | 0 |
4.I.1 | Industry: comply with all applicable Indian laws, including those protecting women, children and vulnerable groups | 0 |
4.I.2 | Industry: demonstrate compliance when called upon by agencies or sectoral regulators | 0 |
4.I.3 | Industry: adopt voluntary measures on privacy, security, fairness, inclusivity, non-discrimination and transparency | 0 |
4.I.4 | Industry: create a grievance redressal mechanism with resolution in a reasonable timeframe | 0 |
4.I.5 | Industry: publish transparency reports on risk of harm in the Indian context, sharing sensitive content confidentially with regulators | 0 |
4.I.6 | Industry: explore techno-legal solutions such as privacy-enhancing technologies, machine unlearning, algorithmic auditing and automated bias detection | 0 |
4.R | Regulators: pursue the twin goals, stay agile, prioritise real harm, avoid compliance-heavy requirements, choose the least burdensome instrument and use techno-legal approaches | 0 |
S1 | Trust is the Foundation: embed trust across the AI value chain | 0 |
S2 | People First: human-centric design, human final control and oversight | 0 |
S3 | Innovation over Restraint: innovate responsibly, maximising benefit and reducing harm | 0 |
S4 | Fairness and Equity: design and test for fair, unbiased, non-discriminatory outcomes | 0 |
S5 | Accountability: developers and deployers remain visible and accountable in proportion to function and risk | 0 |
S6 | Understandable by Design: clear explanations and disclosures as a core design feature | 0 |
S7 | Safety, Resilience and Sustainability: safeguards, anomaly detection, early warning and resource-efficient models | 0 |
gloss | Glossary and annexures | 0 |
pref | Purpose and structure of the Guidelines | 0 |