Govt's AI advisory needs changes, should focus on risky apps: GTRI

The government's advisory on artificial intelligence (AI) needs certain modifications such as a clear definition of "significant platforms", focus on applications with higher risks of spreading misinformation, and a transparent approval process, for the growth of AI technologies in India, a report said on Thursday.

Days after Google's AI platform's response to queries on Prime Minister Narendra Modi generated controversy, the government earlier this month issued an advisory for social media and other platforms to label undertrial AI models and prevent hosting unlawful content.

In an advisory to intermediaries/platforms on March 1, the Ministry of Electronics and Information Technology warned of criminal action in case of non-compliance. The advisory is applicable to significant players/platforms and untested platforms, not start-ups.

The Global Trade Research Initiative (GTRI) also recommended that the government exempt the AI applications in areas of research medicine, education, disaster management, agriculture, technology (broad spectrum); encouraging self-regulation; engaging with broader range of stakeholders, including AI experts, academia, industry representatives, and civil society, in refining and implementing the advisory; and rolling out the advisory's requirements in phases.

"These modifications in the Ministry of Electronics and Information Technology advisory would support the growth of AI technologies while still addressing the issue of false information," GTRI Founder Ajay Srivastava said.

"There is a need to clearly define what constitutes 'significant platforms' versus startups or smaller companies. AI is a new field and every firm could be a start up. Large platforms could hive off its AI venture as a fresh start up. Even independent start-ups may have investment from large platforms," it said.

It added to narrow the advisory's scope to focus on applications with higher risks of spreading misinformation or causing harm, rather than applying a broad brush approach; and suggested implementing a transparent, and faster approval process with specific timelines for government responses.

"Encourage the development of industry-led guidelines and best practices for AI ethics, reliability, and safety, with government oversight to ensure compliance. Allow companies to conduct self-assessments and audits against these guidelines, reporting to the government only if they identify significant risks," the report said.

It also asked to establish mechanisms for ongoing evaluation of the advisory's impact on AI development and misinformation, adjusting policies as needed based on empirical evidence.

"MeitY advisory on AI is in the right spirit. However, it needs major modifications for rapid growth of the AI industry in India," Srivastava said.