India revises decision requiring new AI models to get government approval before market launch
India has made a U-turn on its approach to AI regulation following widespread backlash from both local and global stakeholders. The Ministry of Electronics and IT had earlier released an advisory that mandated government approval for the deployment of AI models in the South Asian market.
The directive was met with sharp criticism from multiple high-profile individuals including Martin Casado, a partner at venture firm Andreessen Horowitz, who called the move “a travesty.”
Consequently, the government has shared an updated advisory with industry players and instead of reporting to the government, AI founders will be required to label under-tested or unreliable AI models to avoid misleading users about the capabilities of these technologies.
The earlier advisory was also a departure from India's previous laissez-faire stance on AI regulation.
The government had just a year ago stated that it would refrain from regulating the AI market as that would discourage innovation making India lose out on the AI wave.
Similar to the previous advisory, the new advisory has not yet been published online, but reputable sources have been able to get and review a copy of it.
The ministry notes that the advisory isn't legally binding, but asserts its significance as a guiding principle for future regulation, emphasising the need for industry compliance.
Some key directives within the advisory include the prohibition of AI models from disseminating unlawful content or perpetuating bias, discrimination, or threats to electoral integrity. AI companies are urged to implement “user consent” pop-ups and similar mechanisms to transparently convey the unreliability of AI-generated outputs.
The issue of deepfakes was also addressed in the advisory with AI developers instructed to label or embed content with unique metadata or identifiers. Another notable change with the new advisory is that AI makers are no longer required to devise methods for identifying the “originator” of any particular message.
India's recalibration of its AI advisory reflects a nuanced approach aimed at balancing innovation with regulatory oversight, signalling a willingness to engage with stakeholder feedback and adapt to evolving technological landscapes.
