Indian govt removes approval need for new AI models rollout

The Ministry of Electronics and Information Technology (MeitY) has updated its advisory on the deployment of artificial intelligence (AI) by internet companies in India.

Initially, companies were required to obtain government approval before launching AI models deemed "under-tested" or "unreliable." The mandate has been lifted in the latest advisory issued on March 15, which now supersedes the previous directive from March 1.

However, the companies are still expected to comply with the new guidelines immediately, but the process has been simplified.

What are the new requirements?

Under the revised guidelines, AI models, including large language models (LLMs) and generative AI software, must be labelled to indicate the potential unreliability of their outputs.

The move aims to inform users about the fallibility of AI-generated content. Additionally, the platforms are required to implement consent pop-ups to alert users about the possibility of encountering false or unreliable information. This approach seeks to balance innovation with user safety and awareness.

Addressing misinformation and bias

The updated advisory also emphasizes the need for AI models to avoid spreading content that violates Indian laws, including those that could bias, discriminate, or threaten the electoral process.

The government has also instructed platforms to identify deepfakes and misinformation by labelling or embedding content with unique identifiers. The measure is designed to make it easier to trace artificially generated or modified content back to its source.