Independent guidance for spotting, reporting and understanding synthetic media in India

What is Synthetically Generated Information? (SGI Legal Guide)

Deep dive into the legal and technical definition of Synthetically Generated Information (SGI) under new laws and its impact on content creation.

When India’s IT Amendment Rules 2026 came into force, they introduced a new legal term that developers, content creators, and legal researchers need to understand: Synthetically Generated Information (SGI). Here’s the complete breakdown.

The Official Definition of SGI

Under the IT Rules Amendment 2026, Synthetically Generated Information is defined as any digital content – including audio, video, image, or text – that has been created or materially altered using AI or machine learning algorithms in a manner that misrepresents the identity, actions, or statements of a real person, or creates a false impression of real events that did not occur.

What Is NOT SGI: The “Good Faith Editing” Exemption

The rules explicitly carve out good faith editing from the SGI definition. The following do NOT qualify as SGI:

  • Standard photo filters (brightness, contrast, saturation adjustments)
  • AI-assisted background removal that doesn’t alter the subject’s identity
  • Auto-captioning and transcription tools
  • Grammar and spell-check AI tools operating on text
  • AI upscaling of resolution without identity modification

The key legal test is whether the algorithmic manipulation materially changes how a real person is perceived or represented. A beautification filter passes. Replacing someone’s face does not.

Technical Markers That Define SGI

From a forensic standpoint, content is likely to be classified as SGI if it exhibits:

  • GAN artifacts: Texture inconsistencies in hair, teeth, or background edges typical of Generative Adversarial Networks
  • Facial geometry anomalies: Asymmetric blinking, ear misalignment, or unnatural head-turn artifacts
  • Spectral inconsistencies in audio: Vocoders and neural TTS systems leave frequency fingerprints detectable via spectrogram analysis
  • Missing C2PA metadata: Genuine AI-generated content from compliant tools carries Content Credentials. Absence of these in AI-looking content is itself a red flag

Implications for Developers and Creators

If you build or use generative AI tools in India, you are now legally required to embed C2PA-compliant metadata into any output that could qualify as SGI. Distributing SGI without disclosure exposes you to liability under both the IT Act and the new amendment. Platforms that host SGI without proper labelling lose Safe Harbour protection under Section 79.

Join the discussion

Your email address will not be published. Required fields are marked *

Teal signal passing through layered editorial pages

Stay informed

Make better decisions in an age of synthetic media.

Follow practical explainers, detection guides and India-focused policy coverage from Deepfake India.

Read the latest