India needs an AI safety framework designed for India, rather than simply replicating America's White House Accord on Super Intelligence. Its enormous population, linguistic diversity, digital scale and application-led AI ecosystem create risks and requirements fundamentally different from frontier-model-centric markets.
The US accord asks leading AI companies to establish internal controls, oversight teams, independent external evaluations and board-level supervision. These principles offer useful lessons, but India's AI ecosystem extends far beyond companies developing foundational models.
India is primarily becoming a massive AI deployment economy. AI is entering banking, healthcare, government services, education, telecom and commerce. Safety must therefore address not only frontier-model risks but also deepfakes, impersonation, cyberattacks, privacy breaches, misinformation and harmful automated decisions.
Nasscom Chairman Srikanth Velamakanni argues that safety must remain ahead of AI capability as increasingly agentic systems gain greater autonomy and access to critical systems. India therefore needs participation from government, enterprises, AI developers, academia, independent researchers, users and civil society.
A major bottleneck will be testing at India's scale. AI systems must work safely across multiple languages, dialects, cultural contexts and demographic groups. Building representative datasets, multilingual benchmarks, red-teaming capabilities, independent auditors and technically competent evaluation institutions will require substantial investment and specialised talent.
Another challenge is avoiding regulatory fragmentation and compliance overload. Banking, healthcare, telecom, government and defence carry different levels of risk. Excessively rigid pre-release certification could slow innovation, while purely voluntary self-regulation could leave accountability gaps. India therefore needs a hybrid, risk-based approach combining mandatory safeguards for critical applications with flexible standards elsewhere.
The real challenge is balancing innovation, sovereignty, safety and inclusion. India's framework should establish clear accountability, independent audits, incident reporting, deepfake detection, data protection and human oversight without creating barriers that favour only large technology companies. India's opportunity is not to copy America, Europe or China, but to create an India-scale AI trust framework capable of becoming a model for the Global South.
