India’s Techno-Legal Vision for AI Governance
India is laying the groundwork for how artificial intelligence (AI) should be developed, deployed, and regulated in the country—with a clear objective: protecting citizens while encouraging innovation. The Office of the Principal Scientific Adviser (OPSA) to the Government of India has released a White Paper titled “Strengthening AI Governance Through a Techno-Legal Framework”, outlining India’s vision for building an AI ecosystem that is trusted, transparent, accountable, and innovation-friendly.
The Techno-Legal Approach
At the core of the paper is the concept of a “techno-legal” approach to AI governance. This means that AI regulation, including legal safeguards, technical controls, and institutional oversight, must be embedded directly into AI systems from the design and development stage itself. Principal Scientific Adviser Professor Ajay Kumar Sood emphasized that this approach offers a practical way forward by embedding safety, accountability, and compliance into AI systems by design.
Regulatory Framework and Compliance
The Government of India is tightening rules to curb deepfakes and other AI-generated content under amended IT regulations, requiring platforms to clearly label AI-generated or synthetic media and remove harmful content within three hours of it being flagged by authorities. The OPSA White Paper advocates managing AI risks without suppressing innovation or progress. It breaks AI governance into clear and practical areas, including:
- What a techno-legal framework for AI actually means
- Ensuring AI is safe and trustworthy across its entire lifecycle—from design to deployment
- The role of technical tools in enforcing governance
- How India can implement AI governance across sectors
- Building compliance, monitoring, and accountability mechanisms
Citizen Rights and Transparency
According to Abhishek Agarwal, President of Judge India & Global Delivery at The Judge Group, a credible AI governance framework must begin with two basic rights for citizens: notice and remedy. “People should know when AI affects important decisions like credit, jobs, or benefits, and they should have a simple way to appeal. This means human review, a clear explanation, and a defined timeline,” Agarwal stated.
India’s Unique Approach
This White Paper is the second in a series by OPSA on Emerging Policy Priorities for India’s AI Ecosystem. The first paper, released in December 2025, focused on “Democratising Access to AI Infrastructure.” It argued that AI infrastructure—such as high-quality datasets, affordable computing power, and Digital Public Infrastructure (DPI)—should be treated as a shared national resource rather than being concentrated among a few large players. Together, these papers aim to spark informed discussion and guide policymaking as India’s AI ecosystem continues to evolve.
Comparative Models of AI Regulation
India’s techno-legal approach differs significantly from AI regulation models being adopted in the EU, US, or China. Agarwal noted that India’s approach is framed as a hybrid model. Instead of starting with “one big AI law,” India is building a techno-legal stack where guardrails are designed into systems and supported by existing legal duties and sector-specific rules. In contrast, the EU has adopted a clearly risk-tiered model under the AI Act, the US remains largely sector- and agency-led, while China emphasizes platform responsibility, content controls, and algorithm governance.
Embedding Governance in AI Systems
Governing bodies must consider how governance can be “built into AI systems by design,” particularly for startups and smaller developers. This involves using simple defaults that teams can realistically maintain. Essential elements include:
- Clear model cards
- Basic documentation of training data sources
- Light red-teaming to catch obvious risks
- Logs that explain model outputs
Startups can begin small by defining what the model must not do, adding safety filters, involving humans in sensitive use cases, and ensuring updates are traceable. Recent government guidance reinforces this approach, emphasizing due diligence, user disclosures, and misuse prevention as minimum expectations.
Addressing Rapidly Scaling Risks
Agarwal pointed out that the most urgent AI risks in India are those that scale rapidly, such as deepfakes, fraud, and harmful content, alongside privacy risks and bias in significant areas like credit, employment, and welfare. Recent policy actions have focused heavily on curbing deception and synthetic media that can mislead or impersonate citizens. A sensible way forward is a harm-first approach: the greater the risk to rights, safety, and trust, the stronger the safeguards and accountabilities must be.
Sector-Specific Regulations
AI governance works best when guided by shared core principles that regulators adapt into domain-specific rules. For instance, healthcare can focus on patient safety and clinical responsibility, finance can prioritize fairness and audit, and telecom can emphasize consumer protection and prevention of misuse. A central coordinating mechanism can set baseline standards—documentation, disclosures, and incident reporting—while regulators enforce them within their jurisdictions. India’s techno-legal approach fits this model: one common framework supported by sector-specific playbooks.
Enforcement and Accountability
Technical tools such as audits, model documentation, or algorithmic transparency are essential for legal enforcement. Rules can state “be fair” or “be safe,” but tools like model cards and independent audits demonstrate how those principles are applied in practice. These tools also allow regulators to act proportionately—lighter checks for low-risk systems and deeper scrutiny for high-impact deployments. India’s approach is to embed these safeguards into the AI lifecycle, creating a technical evidence bridge between policy and accountability.
Generative AI Considerations
On generative AI, Agarwal stressed the importance of a risk-based approach without being innovation-blind. Generative AI evolves too quickly for rigid licensing of everything. However, clear disclosures, safety testing for high-reach deployments, and strong controls where harm is likely—such as deepfakes or automated decisions affecting rights—are realistic and necessary. India’s advisories on synthetic content reflect this thinking: prevent misuse while allowing legitimate creativity and experimentation.
Responsibilities of Deployers and Developers
Accountability primarily lies with the deployer—the entity using AI in real-world decisions. Developers must be accountable for unsafe design, known limitations, and misleading claims. Data providers bear responsibility for consent and provenance violations, while platforms cannot ignore misuse at scale. India is likely to adopt a similar approach to the EU, mapping duties to those best placed to manage risk.
A Mixed Approach to Governance
Effective AI governance must mix high-impact use cases facing mandatory duties with broader best practices encouraged through incentives—such as faster approvals or recognition for audited compliance. For startups, this staged approach reduces uncertainty and ensures predictability, allowing companies to understand expectations at each risk level.
A Model for the Global South
India’s influence on global AI norms, particularly for developing and Global South countries, is noteworthy. Its model is appealing because it is implementable with safeguards. Many developing nations seek protection but also need space to innovate and build local AI capacity. The techno-legal approach—embedding guardrails, utilizing sector regulators, and focusing on harms—is easier to replicate than a complex AI statute. India’s pro-innovation framing aligns closely with the needs of emerging markets.
Capacity Building and Open-Source AI
Capacity-building must be practical, not academic. Regulators need training in AI lifecycles, audit reports, and incident investigations. Courts require clear concepts of causality, negligence, and duty of care in algorithmic systems. Enforcement teams need technical support—labs, auditors, and standard checklists—to ensure effective implementation. Additionally, open-source AI should not be treated as “no responsibility.” Those deploying models at scale must uphold duties around disclosure, safety testing, and misuse prevention.
This comprehensive study highlights India’s innovative approach to AI governance, focusing on balancing regulatory safeguards with the need for innovation, making it a potential model for other nations.
