India changes mind on AI, demands government approval for launching new models

By releasing a directive requiring "significant" tech companies to obtain government approval prior to releasing new models, India has entered the global AI debate.

The alert was sent to businesses on Friday by India's Ministry of Electronics and IT. The advice, which reporters have seen but which is not in the public domain, also requests that digital companies make sure that their goods and services do not encourage prejudice or discrimination or jeopardize the integrity of the democratic process.

Even though the ministry admits the advice is not legally binding, India's IT Deputy Minister Rajeev Chandrasekhar asserts that the notification shows this is the direction of regulation. He continued by saying that it was an advisory and that you should follow it now.

On Monday, Chandrasekhar said via tweet that the warning does not apply to startups and is intended for unproven AI platforms being deployed on the Indian internet.

In its advice, the ministry references the authority bestowed upon it by the IT Act of 2000 and the IT Rules of 2021. It requests "immediate effect" compliance and requires tech companies to provide the government with an "Action Taken-cum-Status Report" in a timeframe of fifteen days.

The new guidance represents a shift from India's prior hands-off policy to AI regulation. It also requires tech businesses to “appropriately” designate the likely and inherent fallibility or unreliability of the output their AI models generate. The ministry rejected attempts to control the rise of AI less than a year ago, citing the industry's criticality to India's geopolitical objectives.