BharatNotes What is Deepfake?
A deepfake is synthetic media — audio, image or video — in which a person's face, body or voice is fabricated or swapped using artificial intelligence so convincingly that it appears authentic. The term combines "deep learning" with "fake". Deepfakes are typically produced using Generative Adversarial Networks (GANs), where two neural networks compete — one generating fakes and the other detecting them, driving steadily more realistic output — and increasingly using diffusion models that synthesise images and video from text prompts.
How the Technology Works
A GAN pairs a generator that produces synthetic content with a discriminator that judges whether it is real or fake. Through repeated rounds, the generator learns to defeat the discriminator, yielding highly realistic forgeries. Face-swapping autoencoders and, more recently, large diffusion and voice-cloning models have lowered the skill and cost barrier, making convincing fakes accessible to non-experts.
Why It Matters
| Risk area | Concern |
|---|---|
| Personal harm | Non-consensual intimate imagery, harassment, reputational damage |
| Fraud | Voice/face cloning for financial scams and impersonation |
| Disinformation | Fake political clips threatening electoral integrity |
| Erosion of trust | "Liar's dividend" — genuine media dismissed as fake |
In India, deepfakes targeting actors (Rashmika Mandanna, 2023) and public figures including PM Modi (flagged by him on 17 November 2023) catalysed regulatory attention.
Legal and Regulatory Status in India
India lacks a single dedicated anti-deepfake statute, but multiple provisions apply:
- IT Act, 2000 — Section 66C (identity theft), 66D (cheating by impersonation using a computer resource), 66E (privacy violation).
- Bharatiya Nyaya Sanhita, 2023 — provisions on defamation and on false/misleading information causing public mischief.
- IT (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 — basis for MeitY advisories requiring platforms to act against unlawful synthetic content.
The most significant recent step is the IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2025, notified by MeitY and effective from 15 November 2025. They define "synthetically generated information" and mandate labelling/disclosure of AI-generated content, requiring that the disclaimer cover at least 10% of the visual display area (or the initial 10% of an audio clip's duration), along with permanent metadata identifiers that may not be removed or suppressed.
The Delhi High Court has been notably proactive, granting personality-rights injunctions and takedown orders in cases involving figures such as Shashi Tharoor, Rajat Sharma, Bhuvan Bam and Ankur Warikoo (2024–25), restraining unauthorised AI/deepfake use of name, voice and likeness.
UPSC Angle
For GS3, link deepfakes to cyber security, internal security and disinformation; for GS2, to platform regulation and the IT Rules; for GS4, to AI ethics and the erosion of trust. Note the regulation-versus-free-speech tension — the 2025 rules drew criticism for lacking explicit exceptions for satire, news and research (as of the stakeholder consultation, 2025).
