India requires AI beta versions to be pre-approved by the government

India's Ministry of Electronics and Information Technology (MeitY) has issued guidelines stating that artificial intelligence technologies in the development stage should get government approval before releasing to the general Indian audience.

Here's What We Know

According to the document, the deployment of "untested/unreliable" AI models, language models and generative AI in the Indian internet segment should be authorised by the authorities. It also recommends the use of pop-ups and labelling to inform users about possible errors and inaccuracies in generative AI content.

In addition, platforms and service providers must ensure that AI-based products do not bias, discriminate or threaten the electoral process. Compliance is mandated within 15 days.

According to media reports, developers may be required to demonstrate their AI product to officials or conduct stress testing after applying for authorisation.

Information Technology Minister Rajeev Chandrasekhar said the measures are advisory for the time being. However, in the future they may become mandatory and will be enshrined in law.

In this way, Indian authorities are seeking to ensure greater "rigour" and accountability for AI solution providers in the wake of a number of recent incidents, including Google's Gemini system making erroneous remarks about the country's prime minister.

Source: The Register