India’s approach to artificial intelligence governance combines innovation, access and risk management. The government argues that AI should support development and be accessible to countries in the Global South. At the same time, AI can reproduce bias, threaten privacy, generate deceptive content and concentrate power. A credible policy has to address both the opportunity and the harms through transparent rules, capable regulators and meaningful accountability.
At the 2026 AI Impact Summit in New Delhi, India presented its “MANAV” vision: moral and ethical systems, accountable governance, national sovereignty, accessibility and inclusion, and valid and legitimate AI. The summit brought together governments, researchers, companies and civil society. Its declarations and discussions can create international momentum, but voluntary principles do not automatically protect people; domestic laws and enforcement remain decisive. (Prime Minister’s address at AI Impact Summit 2026; India AI: Responsible AI framework)
India’s scale and linguistic diversity make inclusion a practical governance issue. Systems trained mainly on English or urban data may perform poorly for regional-language speakers, rural users and communities underrepresented in datasets. Public services using AI need testing across populations, clear routes to appeal consequential decisions and human review where errors can affect rights or access to benefits. Data sovereignty can strengthen national control, but it must coexist with privacy protections and secure cross-border research where appropriate.
The G20 New Delhi Declaration in 2023 endorsed responsible AI development that protects human rights, safety, transparency, fairness, accountability, privacy and data protection. In 2026, India established an intergovernmental AI Governance and Economic Group to develop and oversee the country’s position and strategy. These initiatives indicate growing attention to governance, but the public will need clear regulatory responsibilities, risk assessment requirements and remedies for people harmed by AI systems. (G20 New Delhi Leaders’ Declaration; MeitY: constitution of AI Governance and Economic Group)
There is a balance to strike. Overly broad rules can burden small developers and public-interest research; weak rules can leave individuals with little recourse when systems discriminate or expose data. Risk-based requirements, independent audits, incident reporting and accessible appeals can focus oversight where the stakes are high. The government should publish impact assessments for public-sector deployments and ensure that procurement contracts do not prevent scrutiny.
An inclusive AI agenda must address the infrastructure needed to participate. Compute, reliable electricity, high-quality datasets, research funding and skilled workers are unevenly distributed across countries and within India. Access initiatives can widen participation, but they need transparent allocation rules and safeguards against concentrating public resources in a few firms. Regional-language datasets and evaluation methods require investment by universities, communities and public institutions. Inclusion means more than allowing people to use finished products; it includes a role in shaping and auditing them.
Government use of AI deserves especially strong controls because public agencies can affect welfare, education, health, employment and law enforcement. Before deployment, an agency should identify the system’s purpose, data sources, error rates, affected groups and alternatives. People should be told when AI materially influences a decision and have a way to obtain human review. Procurement contracts should preserve access for auditors and regulators. A public register of high-impact systems would help citizens understand where automated tools are being used.
The risk-based approach needs clear definitions and enforceable duties. Developers and deployers should know which systems require impact assessments, testing, incident reporting and independent audit. Small enterprises may need templates and shared testing resources, but exemptions should not remove protections where the consequences for individuals are severe. Rules should be updated as capabilities change, with consultation that includes researchers, workers, affected communities and civil society as well as industry and government.
International cooperation can expand access while preserving safeguards. India can work with developing countries on safety evaluation, language resources, skills and affordable compute, and can engage major technology powers on interoperable standards. Global rules should not be written only by countries that control the largest models and cloud infrastructure. At the same time, inclusion should not be used to excuse opaque systems or weak remedies. A useful multilateral framework shares expertise and sets expectations that apply across borders.
AI governance must remain open to evidence. Independent researchers need safe access to systems and information sufficient to assess reliability, bias and security. Public reporting on incidents and corrective action can identify recurring problems. Citizens should have routes to complain and seek remedy, particularly when an automated output affects an essential service. India’s global leadership will depend on whether its governance institutions can act on evidence and enforce rules consistently, not only on whether the country convenes a high-profile summit.
Workers affected by AI also need a voice. Automated tools can change hiring, workplace monitoring and task allocation, yet employees may not know how decisions are made or how to challenge errors. Clear notice, consultation and review channels can help identify impacts before deployment. Skills programmes should support workers whose roles change, while employment rules remain enforceable. An inclusive governance model considers both users of AI and people whose work is reshaped by it.
India can contribute to global AI governance by advocating wider access to compute, data and skills while supporting interoperable safety standards. International cooperation should include developing countries in rulemaking, but inclusiveness must not dilute basic safeguards. The measure of India’s approach will be how it protects people in real deployments, especially those with limited power to challenge automated decisions.