BRICS members agree on very little when it comes to the actual mechanics of AI regulation, despite signing a shared statement on AI governance principles (covered in our companion post). India, China and Brazil — the three BRICS members with the most developed AI policy so far — have each taken a genuinely different route.
India: principle-based guidelines, not a binding law
India's AI Governance Guidelines, released in February 2026 alongside the India AI Impact Summit, are built around what the government calls "seven sutras": trust, people-first governance, innovation over restraint, fairness and equity, accountability, understandability by design, and safety and resilience. It's a techno-legal, principle-driven framework rather than a single binding AI law — it calls for human oversight, transparency and grievance redressal, and creates new institutions (an AI Governance Group, a Technology & Policy Expert Committee, and an AI Safety Institute) rather than a court-enforceable rulebook. The explicit framing is pro-innovation: safeguards, but not restriction as the default.
China: binding, application-specific rules
China's approach is the most legally binding of the three — specific, enforceable regulations targeting individual AI applications (generative AI services, algorithmic recommendation, deepfakes) rather than one overarching framework. Where India is publishing guidance and setting up advisory institutions, China has been issuing rules with real compliance obligations and enforcement behind them, application by application.
Brazil: EU-inspired, rights-based
Brazil's approach leans on the EU AI Act as a model — a rights-based framework organised around risk tiers, closer to Europe's regulatory instinct than to India's or China's. That's a notable choice for a BRICS member, given the bloc's own AI governance declaration pushes back against any single regional model (implicitly, the EU/US one) dominating global AI governance.
The practical takeaway
For an Indian business today, the guidelines mean principle-based expectations (transparency, human oversight, grievance redress) rather than a specific compliance checklist you can be fined against — closer in spirit to early data-protection guidance than to the DPDP Act's more concrete obligations. That's likely to change as the AI Governance Group and AI Safety Institute mature; treat this as the starting framework, not the final word.