Govt Takes Back Controversial Advisory on Prior Permission to Deploy AI Models After Industry Uproar – News18

In a significant reversal of its recent stance on artificial intelligence regulations, the government has walked back on a controversial advisory that required prior approval for deploying certain AI models. The U-turn comes after criticism from many local and global AI industry insiders.

In early March, the government had issued an advisory outlining strict guidelines for the deployment of AI models on Indian internet. It mandated explicit permission from the Government of India for the use of under-tested or unreliable AI systems, generating confusion and sparking backlash across various sectors.

Following the criticism, the Ministry of Electronics and IT clarified that the requirement for seeking government approval applied only to large tech companies, exempting startups from the stringent regulation.

However, in a recent development, the ministry released an updated advisory, signalling a significant shift in approach. The new guidelines, shared with industry stakeholders, no longer mandate government approval before launching or deploying AI models.

Under the revised guidelines, firms are now advised to label under-tested and unreliable AI models, ensuring users are informed of their potential fallibility or unreliability. This move aims to balance fostering innovation and ensuring accountability in AI deployment.

The advisory, dated March 15, was issued by the Ministry’s Cyber Law and Data Governance Group. It emphasises the importance of intermediaries and platforms adhering to due diligence obligations outlined in the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021.

Here’s a look at the highlights of the new advisory:

• Ensuring AI models do not permit the hosting, display, or transmission of unlawful content, in compliance with IT Rules and other relevant laws.
• Preventing bias or discrimination in AI systems and safeguarding the integrity of the electoral process.
• Requiring the labelling of under-tested or unreliable AI models or “consent popup” or the use of equivalent mechanisms to inform users about potential fallibility or unreliability.
• Informing users about consequences for dealing with unlawful information through terms of service and user agreements.
• Any intermediary permitting or facilitating the creation, generation, or modification of content with the potential for misuse as misinformation or deepfakes should label or embed such content with permanent unique metadata. In the case any changes are made by a user, the metadata should be so configured to enable identification of such user or computer resource that has effected such change.

The advisory also reiterates the consequences of non-compliance with IT Act, 2000 and IT Rules, including prosecution under applicable laws for intermediaries, platforms, and their users.

All intermediaries are instructed to ensure immediate compliance with the new guidelines, signalling a more flexible approach to AI regulation in India. This shift is expected to ease concerns among entrepreneurs and investors while fostering innovation in the AI sector.