Break western bias in AI, India must go cultural: EY India

New Delhi: As India accelerates its quest for sovereign AI, the nation must prioritise the release of government-held data to counter the western standpoint of current global models, EY’s Rajnish Gupta said.

Ahead of the Union Budget, Gupta urged the government to make vast amounts of public data available to local developers to ensure India builds AI that reflects its own identity. Gupta, Partner, Tax and Economic Policy Group, EY India, warned that today’s most popular Large Language Models (LLMs) are fundamentally biased because they are trained primarily on US and European data. To ensure India builds the next generation of AI that is culturally and linguistically relevant, he argued the government must act as a primary data provider. Gupta noted that India’s unique cultural nuances, diverse languages, and religious complexities are often misunderstood or ignored by global models. “One of the big things which we could do a little more in India is make a lot of data available publicly. (Most of) the answers that you get are very western. They’re very much from a US or a European standpoint. “...We have our own culture. We have our own languages, our own nuances… A lot of this data can be generated by the government and made available to the people who are developing LLMs,” he said in an interview with PTI. Comparing AI development to India’s success with Aadhaar and UPI, Gupta said India can democratise AI by treating data and compute as a public good. He suggested that while public investment shouldn’t necessarily dictate innovation, it is essential for creating the foundation. Despite the need for government-provided data, Gupta urged a hands-off approach to regulation. He praised the government for not mimicking the EU’s AI laws, arguing that a lack of interference is what will allow Indian entrepreneurs to experiment with these new, culturally aware models.