India Weighs Dedicated Risk-Based AI Regulation
India's IT ministry is weighing a dedicated, risk-based AI law, according to a July 7, 2026 Economic Times report citing government officials. The proposed framework would apply lighter rules to low-risk tools such as chatbots and productivity software, while imposing stricter documentation and oversight duties on high-risk AI used in banking, finance, healthcare, and critical infrastructure. The report follows July 4 Times of India comments from MeitY secretary S. Krishnan, who said the government may prepare draft AI legislation, and builds on PIB's February 2026 India AI Governance Guidelines backgrounder. For AI builders serving Indian users, the practical takeaway is that deployment context and sector risk may soon carry direct compliance weight, not just voluntary guidance.
India is one of the largest AI deployment markets, so a shift from soft guidance toward a dedicated legal framework would change compliance planning for model providers, enterprise software vendors, and data-science teams shipping into regulated sectors. The important signal is not that a bill is final; it is that officials are now discussing a graded, risk-based structure, which would make the use case and sector around an AI system central to its obligations.
What happened
Economic Times reported on July 7, 2026 that officials are considering an AI law that would classify systems and apply stronger requirements to higher-risk uses. The reported framing would keep lighter rules for lower-risk tools such as chatbots, productivity systems, and recommendation engines, while placing stricter obligations on AI used in banking, finance, healthcare, and critical infrastructure.
The report builds on comments covered by Times of India on July 4, where MeitY secretary S. Krishnan said the government may examine a separate legal framework for AI and that the ministry could prepare draft regulation when the timing is right. That marks a sharper posture than India's earlier reliance on existing laws, including the IT Act, privacy law, and intermediary rules, to cover deepfakes, misinformation, and other AI harms.
Regulatory context
PIB's February 2026 India AI Governance Guidelines backgrounder already emphasized trust, accountability, explainability, safety, and institutional capacity through a seven-Sutra, principle-based framework and new bodies such as the AI Governance Group and AI Safety Institute. The new reporting suggests those principles may be translated into more formal, enforceable duties rather than remaining voluntary.
For practitioners
For builders, the likely near-term work is not panic over a final statute, but better risk inventory. Teams serving India should be able to identify whether their AI systems are consumer assistants, workplace productivity features, model APIs, decision-support tools, or high-impact sector deployments. The first controls worth preparing are plain ones: documented intended use, human oversight paths, evaluation records, incident handling, data-protection review, and clear disclosures. If India follows a risk-tiered route, the compliance burden will likely land hardest on AI systems tied to credit, employment, health, public services, critical infrastructure, and security-sensitive workflows, while lower-risk tools may face lighter transparency and safety expectations.
Key Points
- 1India is reportedly considering dedicated AI legislation built around graded obligations for different system risk levels.
- 2The signal matters because high-risk finance, healthcare and infrastructure deployments may face stronger documentation and oversight duties.
- 3AI builders serving India should start mapping use cases, sector exposure, data flows and human-review controls now.
Scoring Rationale
This is a notable policy signal from a major AI deployment market, not a final law. It matters to practitioners because a risk-tiered framework would affect documentation, auditability and sector-specific deployment controls for India-facing AI products.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems