Used an AI filter on your last reel? India now legally requires a label on it.
The IT Rules 2026 created a disclosure regime for AI-generated and AI-altered content. A watermark covering 10% of the frame, a label in the first three seconds, and an ASCI #AIUsed tag are now the floor — and a colour-correction filter that "materially alters" a product can trigger it. Here is what counts, and what gets your post taken down.
By Sumit Kumar
You shot a brand reel, ran it through an AI filter to smooth skin and fix the lighting, and posted it. No watermark, no special tag. A few months ago that was unremarkable. In 2026 it can get the post taken down inside 36 hours and land you on a public non-compliance dashboard.
The reason is a regime most creators haven't read yet: the IT Rules 2026 created a real, enforceable disclosure obligation for AI-generated and AI-altered content. This post explains — in plain language — what the law calls “synthetic” content, what label it now requires, and the surprising case where an ordinary beauty filter trips the rule.
What the law now calls “synthetic”
The framework is the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026. At its centre is a new term: Synthetically Generated Information (SGI) — defined under Rule 2(1)(wa) as any audio, visual, or audio-visual content that is algorithmically created, generated, modified, or altered using a computer in a way that makes it appear indistinguishable from a real person or a real-world event.
Read that definition twice, because the operative word is modified or altered — not just “generated from scratch.” A fully AI-generated avatar is obviously SGI. But so is a real video that's been altered enough to misrepresent reality. That's the part that catches creators off guard.
The label the rules now require
When content is AI-generated or AI-altered, the disclosure is no longer a courtesy hashtag. Per the India AI Governance Guidelines and the ASCI influencer rules, the floor is:
- A visible on-screen watermark that is prominent and high-contrast, and covers at least 10% of the image or video surface area.
- For video, the disclosure must appear in the opening frame and persist for at least the first three seconds.
- An ASCI-aligned label: if the post is also an ad, you now disclose both the commercial nature (#Ad) and the AI use (#AIUsed or #AIGenerated). One does not substitute for the other.
If you work with multinational brands, there's a second layer. Brand contracts increasingly demand C2PA compliance — the Coalition for Content Provenance and Authenticity standard — which means embedding provenance metadata that records which AI tools were used and how much was human-made. This is the spillover from the EU AI Act (Article 50), which came into full effect in August 2026, arriving in your contracts even though you're shooting in Mumbai.
The trap: a colour filter can be “material alteration”
Here is the scenario that catches careful creators. A beauty creator is hired to promote a serum. To get a “flawless” look, the production team uses an AI filter that smooths skin and pushes the lighting and colour saturation beyond what the physical product can actually do. It goes up without the watermark or the metadata.
Within hours a consumer-watchdog monitoring tool flags it for missing synthetic-media disclosure. Because there was no user declaration of SGI at upload — an obligation under Rule 3(3) — the platform takes the content down within 36 hours to protect its own safe-harbour status. The brand then catches a separate notice from the Ministry of Information and Broadcasting for “misleading” advertising, and the creator's name appears on ASCI's non-compliance dashboard.
The thing nobody on the team realised: a colour-correction filter that materially alters the substance of the product's effect qualifies as SGI under Rule 2(1)(wa). It looked like ordinary post-production. The law read it as a synthetic claim about the product.
The line is “material alteration,” not “heavy editing.” Cropping, basic exposure, and stabilisation are not SGI. What crosses the line is alteration that changes what a viewer would believe to be true — making a product look like it does something it doesn't, or making a face/scene appear indistinguishable from an unaltered reality when it isn't. When in doubt on a brand deal, declare it and add the watermark. The cost of an unnecessary #AIUsed is zero; the cost of a missing one is a takedown plus a dashboard listing.
The four-line creator checklist
You can stay clear of all of this with a habit, not a lawyer:
- Ask at upload: did AI create or materially alter this? If yes, treat it as SGI.
- Add the watermark. Visible, high-contrast, ~10% of the frame, present in the opening three seconds.
- Tag it correctly. #AIUsed (or #AIGenerated) and #Ad if it's a paid post — both, in the same language as the content.
- For global brands, keep the provenance. Note which AI tools you used; if the contract names C2PA, make sure the export carries the metadata.
These take seconds at post time and save you a takedown you can't appeal once the 36-hour clock has run.
What SutraOS does about this
Here's the honest division of labour: the watermark and the #AIUsed label have to come from you — no platform can paint them onto your footage. What SutraOS exists to do is make sure you're never caught out by a rule you didn't know existed. When a deliverable could fall under the synthetic-media rules, the platform's job is to flag it, tell you in plain language what the law now expects, and make disclosing the easy, default path — instead of something you find out about after a takedown.
The rules around AI content are brand-new and changing fast. Tracking them so creators and the brands they work with stay informed is a burden SutraOS takes on as a platform. You make the content; we make sure the compliance around it never becomes the reason it gets pulled.
If you're a creator who uses AI tools on brand work and the takedown risk above is new to you, SutraOS is live — create your free account and stay ahead of disclosure rules like these. Working through an agency or brand? Send them this too; the obligation is shared, and we're also taking 3–5 agencies as design partners.
Ready to make this someone else’s problem?
SutraOS is live. You can sign up and set up your account today — self-serve, no waitlist — and run your first compliant campaign. Want it hands-on? The design-partner program adds white-glove onboarding for your first campaigns and direct input on the roadmap.
Share or discuss this post: