MeitY AI Advisory on Model Testing, Consent & Deepfake Mitigation
Statutory Overview
Issued under the Information Technology Act 2000 and Intermediary Guidelines, this pivotal advisory directs all digital intermediaries and generative AI platforms to label synthetic media, prevent electoral bias, and prohibit the deployment of unvetted, hallucination-prone AI systems without explicit user consent.
Key Compliance Requirements for AI Companies
Mandatory permanent metadata watermarking and unique identifiers on all synthetic audio, video, and text generated by AI tools.
Clear consent disclaimer informing users that the AI system may generate inaccurate or hallucinatory information.
Prohibition against hosting models that generate deepfakes, non-consensual synthetic intimate imagery, or content disrupting public order.
Immediate 24-hour takedown obligations upon receipt of court or authorized law enforcement notices regarding synthetic fraud.
Legislative Timeline & Milestones
Deepfake Consultations
High-level meetings convened by Union Minister of IT with tech executives following viral celebrity deepfake incidents.
Revised AI Advisory Issued
Clarified scope focusing on commercial intermediaries while exempting early-stage researchers from prior government permits.
Ongoing Monitoring
MeitY Cyber Cell conducts active sweeps for compliance with watermarking standards.
Official Gazette & Regulatory References
Frequently Asked Questions & Legal Interpretations
Q:Do early-stage Indian AI startups need government permission to launch a beta LLM?
No. The revised March 2024 clarification from MeitY removed the prior permission requirement for startups, replacing it with clear labeling and disclaimer mandates.
Q:What are the watermarking requirements for AI video tools?
All generative video services must embed cryptographically verifiable metadata (such as C2PA standards) and visible watermark disclaimers indicating synthetic provenance.