AI Content Labelling Is Now the Law: What India's IT Rules and the EU AI Act Mean for Your Marketing

For the past two years, "should we disclose that this ad was made with AI?" has been an ethics conversation. In 2026 it became a legal one — on two fronts at once.

In February, India notified amendments to the IT Rules requiring synthetically generated images, video and audio to be clearly labelled (Freshfields' analysis of the MeitY amendments). Then on 2 August 2026 the EU AI Act's transparency rules became applicable, requiring AI-generated and AI-manipulated content to carry machine-readable markings and, in specific cases, visible labels (European Commission).

If your brand publishes AI-assisted creative — product renders, campaign films, voiceovers, influencer-style avatars — this affects how you produce, publish and archive that work. Here is what the rules actually say, and what a sensible marketing team should do about them.

What India's IT Rules amendments require

MeitY notified the amendments on 10 February 2026, effective ten days later. The headline concept is synthetically generated information (SGI): content artificially created or altered in a way that looks authentic.

Labels must be clear and prominent. An earlier draft would have required watermarks covering at least 10% of the display area for visual content, with an equivalent for the first 10% of audio. The final rules dropped the fixed percentage in favour of a "clearly and prominently labelled" standard. Visual content needs a visual label, audio needs an audible disclosure, and platforms are expected to embed permanent metadata or identifiers where feasible.

Platforms will ask you to declare. Significant social media intermediaries — platforms with over five million registered users in India, which covers every major channel an Indian brand uses — must require users to declare whether uploads contain synthetic content, and must deploy automated tools to verify those declarations. Expect upload flows to keep tightening.

Text is out of scope. The SGI labelling regime covers images, video and audio; AI-assisted blog posts and ad copy are not covered by it. Every existing rule on misleading or unlawful content still applies, so this is a narrower carve-out than it first sounds.

That is exactly the kind of risk that makes platforms over-comply — which means brands should expect stricter declaration prompts, automated detection sweeps, and faster takedowns of unlabelled synthetic content.

What the EU AI Act adds

The EU's transparency obligations became applicable on 2 August 2026. Providers of generative AI systems must mark synthetic output in a machine-readable way, and deployers — the businesses using those systems — must visibly label two categories: deepfakes, meaning AI content resembling real people, places or events; and AI-generated text published on matters of public interest without human editorial review.

Two details are worth noting. There is a narrow grace period: systems already on the market before 2 August have until 2 December 2026 to meet the machine-readable marking obligation — but the visible labelling duties for deepfakes have been enforceable since August with no extension. And the penalties are not symbolic: up to €15 million or 3% of worldwide annual turnover, whichever is higher, with proportionality considerations for smaller companies.

The Commission has also published a voluntary Code of Practice on the transparency of AI-generated content, which gives companies a concrete way to demonstrate compliance.

Why should an Indian business care about EU rules? Because the AI Act follows the GDPR playbook: it reaches companies whose AI systems or outputs are used in the EU market. If you export, sell to EU customers, or run campaigns reaching EU audiences, you are in scope. And as with GDPR, the practical effect tends to be global — tools and platforms build one compliant pipeline rather than two.

What this means in practice for your marketing

The temptation is to treat this as a legal problem and forward it to a lawyer. The better move is to treat it as an operations problem, because the fixes live inside your content workflow.

Know what you're publishing

Most marketing teams cannot currently answer "which of our live assets are AI-generated?" Start an inventory: campaign films, product imagery, voiceovers, avatar-led videos, AI-upscaled or AI-extended photography. Note which tools produced them, because provenance metadata from the tool is becoming part of your compliance story.

Build disclosure into the template, not the approval

If labelling is a manual step at the end of production, it will be skipped under deadline pressure. Put it in the template instead: a standard visible tag for synthetic visuals, a standard audio disclosure line for generated voiceovers, and a field in your asset-management system recording the declaration you'll make at upload.

Be conservative with realistic humans

Both regimes reserve their sharpest teeth for content depicting real-seeming people. An AI-generated abstract background is low risk. An AI-generated "customer" telling a testimonial story, or a synthetic voice that sounds like a specific person, is exactly what the deepfake provisions target. If a piece of creative could make a viewer believe a real person said or did something, label it prominently — or rethink it.

Watch the declaration flows on ad platforms

Whoever uploads your campaigns needs to know your inventory well enough to answer declaration prompts truthfully. A wrong declaration is now a compliance event, not a typo.

Treat honesty as the brand play it is

There is a marketing upside here. As feeds fill with unlabelled synthetic gloss, a brand that discloses cleanly — and still looks great — earns a form of trust that is hard to buy. We made a similar argument about human-made design earlier this year; regulation is now converging with what audiences already preferred.

The compliance calendar, briefly

India's rules are in force now, and platform-side enforcement is ramping. 2 December 2026 is when the EU's marking grace period ends for systems that predate August. And India's regime will keep evolving — the current rules exclude text, and few expect that to be the final word.

The teams that will find this easy are the ones whose content operations are already documented: clear asset trails, known tools, consistent templates. The teams that will find it painful are the ones where creative lives in personal drives and nobody remembers which renders were generated.

This article is general information about a fast-moving area of regulation, not legal advice. Check your specific obligations with a qualified adviser before relying on it.