India's artificial intelligence ambitions are moving beyond model building to a critical question: can the country build trust while scaling AI? As reported by Business Standard, unlike some global peers pursuing dedicated AI legislation, India has chosen a more flexible approach, relying on existing data protection rules, IT regulations, sector-specific oversight and voluntary governance frameworks. Rishi Agrawal, CEO of Teamlease Regtech, emphasized that the real challenge is not whether India has an AI Act, but whether every AI deployment has clear accountability, documented decision-making, continuous monitoring, audit trails and mechanisms to demonstrate compliance throughout the AI lifecycle. Bruce Keith, CEO of InvestorAi, drew parallels with India's earlier digital growth model, noting that UPI and Aadhaar grew under patchwork regulations, but adoption and safe deployment are different questions.
India's AI ambitions face a critical challenge as the country has nearly two dozen official languages and more than 100 dialects, creating significant barriers to technological adoption. According to reports from Business Standard, if AI cannot close this linguistic gap, it will become another technology that divides the English-speaking elite from the rest of the population. The key to both inclusion and business success depends on whether models can comprehend Bengali voice notes, Gujarati payment queries, and code-switched Hindi-English business calls - all the messy spoken words that drive daily commerce and public life. Despite India's importance as a growth market, AI performance remains limited by language barriers, with GPT 5 achieving only about 45% accuracy on a human-curated benchmark covering 11 Indic languages, including Gujarati.
As AI moves into decision-making systems, responsibility becomes a critical concern, especially in public-sector deployment where decisions can affect welfare delivery, finance, healthcare and citizen rights. Rishi Agrawal stated that AI should never become a mechanism for diffusing accountability, with responsibility continuing to rest with public authorities and officials making final decisions, not with algorithms. He emphasized that public-sector AI requires governance mechanisms similar to financial controls, including defined ownership, independent validation, continuous monitoring, explainability, incident reporting and periodic audits. Bruce Keith warned that this accountability gap is becoming visible globally, with vendors often trying to limit liability or disclaim responsibility for model behavior, which he called unacceptable.
The governance debate is becoming more urgent as deepfakes move from isolated incidents to a broader challenge affecting fraud prevention, information integrity and public trust. Rishi Agrawal suggested focusing on technical safeguards such as provenance, watermarking, content labelling, traceability standards and platform accountability, noting that restrictive regulation alone may not solve the problem. Rahul Agarwalla cautioned against overregulation, emphasizing that flagrant violations are already covered under multiple regulations, while the challenge lies in finding the right balance in regulation. Experts believe that procurement standards may become the next phase of AI governance, with government purchasing decisions determining how AI systems are designed, tested and monitored throughout their lifecycle.
Indian companies are actively addressing the language challenge as part of their broader AI strategy. According to Business Standard, Sarvam AI launched a new funding round, with co-founder Pratyush Kumar stating that building AI that works 'at India's scale' presents a massive opportunity. Rahul Agarwalla noted that sovereign AI remains a strategic necessity, with current policy efforts supporting both startup and enterprise adoption. However, experts acknowledge that complete self-reliance may not be realistic, as even countries investing heavily in domestic AI still rely on global supply chains for chips, cloud infrastructure and research ecosystems. The focus should be on reducing critical dependencies in sensitive areas like public-sector data, critical government workloads, indigenous language models and trusted infrastructure.
