Draft regulations on AI-generated content: protection of the digital space

These are important first steps. Celebrity deepfakes are being used to create sexually suggestive videos of celebrities for fraud (the federal finance minister was recently targeted) and to circumvent strict regulations by avoiding explicit content. Tools like Sora and Dall-E evolve almost every quarter to create increasingly convincing, completely fictional images and video clips. Celebrities, aware that their faces, voices and mannerisms can easily be copied without their consent, are asking courts for injunctions against unauthorized use of their “likenesses.”

But the first step requires a stronger follow-through. The draft definition “all synthetically generated information” appears to be aimed at AI-generated text such as ChatGPT output. However, no guidance is provided on how to label or fingerprint such content. The draft also does not say how the center plans to deal with media produced by underground tools, where checks on nudity and gore can be easily broken. These may require separate legislation. This approach is increasingly being adopted in many regions.

The consultation period provides an opportunity to address these gaps. We need to consider further steps to limit the damage caused by text-based misinformation, while avoiding heavy-handed approaches that stifle AI adoption and innovation. India is right to be proactive, but in an area as complex as AI, precision is required. The more difficult task ahead lies in building institutions of national digital literacy and media trust, the only real antidote to the slide into alternative realities where facts are increasingly contested. The draft is the foundation and the rest of the legislative structure must now be planned.