India Changes AI Policy: Government Approval Now Required for Model Launches

India has entered the global AI discourse by issuing a new advisory that mandates major tech companies to obtain government approval before launching new AI models. On Friday, India’s Ministry of Electronics and IT released this advisory, which, while not publicly published, has been reviewed by us. The advisory calls for tech companies to ensure their products and services "do not allow for bias or discrimination and do not jeopardize the integrity of the electoral process."

Although the ministry acknowledges that the advisory is not legally binding, IT Deputy Minister Rajeev Chandrasekhar emphasized that it serves as a signal for future regulatory expectations. He stated, “We are issuing this as an advisory today and urging compliance.”

In a tweet earlier this week, Chandrasekhar clarified that the advisory is aimed at “untested AI platforms operating on the Indian internet” and does not target startups.

The ministry references its authority under the IT Act of 2000 and the IT Rules of 2021 in this advisory. It demands compliance “immediately” and requests that tech firms submit an “Action Taken-cum-Status Report” to the ministry within 15 days.

This advisory, which also requires tech companies to clearly label the “possible and inherent fallibility or unreliability” of their AI outputs, represents a significant shift from India’s previous hands-off stance on AI regulation. Less than a year ago, the ministry had opted against regulating the burgeoning AI sector, recognizing it as crucial to India's strategic interests.

The new directive has caught many industry leaders off guard. Several Indian startups and venture capitalists express concern that this regulation might hinder India's competitiveness in the global AI arena, where it is already facing challenges.

“I was naive to think I could introduce GenAI to Indian agriculture from San Francisco,” shared Pratik Desai, founder of Kisan AI. “We were excited about training an affordable multimodal model for pest and disease detection. This is disheartening after investing four years to bring AI into this field in India.”

Silicon Valley leaders are also voicing their discontent with India’s regulatory shift. Aravind Srinivas, co-founder and CEO of the AI startup Perplexity AI, criticized the advisory as a “regrettable decision by India.” Martin Casado, a partner at the venture capital firm Andreessen Horowitz, added, “What a travesty.”

This advisory follows a recent incident where Chandrasekhar expressed dissatisfaction with the response from Google's AI model, Gemini. When a user questioned whether India’s Prime Minister Narendra Modi was a fascist, Gemini stated, referencing unnamed experts, that Modi was accused of implementing policies characterized as fascist. Chandrasekhar responded by warning Google that this sort of answer constituted “direct violations” of the IT Rules, 2021, and various Criminal Code provisions.

The advisory also notes that failure to comply with the IT Act and IT Rules could lead to “potential penal consequences” for intermediaries, platforms, or their users when identified.