India has changed its recent AI advisory in response to criticism from both local and global entrepreneurs and investors.
The Ministry of Electronics and IT issued an updated AI advisory to industry stakeholders on Friday, removing the requirement for government approval before launching or deploying AI models to users in the South Asian market.
Instead of the previous guidelines, companies are now encouraged to label AI models that are under-tested or unreliable to inform users about potential shortcomings.
This revision comes in response to severe criticism earlier this month, with high-profile individuals like Martin Casado, a partner at venture firm Andreessen Horowitz, denouncing India’s initial move as “a travesty.”
The March 1 advisory marked a departure from India’s previous hands-off approach to AI regulation. Less than a year ago, the ministry had opted not to regulate AI growth, recognizing its importance to India’s strategic interests.
While the new advisory, like its predecessor, has not been published online, TechCrunch has obtained a copy for review.
The ministry clarified earlier this month that while the advisory is not legally binding, it signals the direction of future regulation, and compliance is expected.
The advisory emphasizes that AI models should not be used to disseminate unlawful content under Indian law and should not facilitate bias, discrimination, or threats to the integrity of the electoral process. Intermediaries are also encouraged to utilize “consent popups” or similar mechanisms to explicitly inform users about the unreliability of AI-generated output.
Additionally, the ministry maintains its focus on identifying deepfakes and misinformation, advising intermediaries to label or embed content with unique metadata or identifiers. Notably, the requirement for firms to develop a technique to identify the “originator” of specific messages has been removed.
