A new global default for synthetic ads
On August 2, 2026, Article 50 of the EU AI Act became applicable, and with it the first continent-wide rule requiring that AI-generated audio, image, video and text be machine-marked and, where it resembles a real person, disclosed to the viewer as synthetic. Under the European Commission's own summary of the rule, providers of generative AI systems must apply machine-readable, detectable marks to their outputs, while deployers — including advertisers and marketers — must clearly disclose deepfakes "at the moment of first exposure," with narrow carve-outs for art and satire. Non-compliance can draw fines up to €15 million or 3% of global turnover, among the stiffer transparency penalties anywhere.
Crucially, Article 50 explicitly reaches advertising. A synthetic celebrity endorsement, an AI-cloned voice reading a sponsored script, or a "realistic synthetic influencer" demonstrating a product all trigger the deepfake disclosure duty, regardless of whether anyone intended to deceive. That is a meaningful expansion beyond the EU's borders in practice: any Indian agency, brand or creator producing synthetic promotional content for a European audience is now inside the rule's reach — well before India finishes writing its own equivalent.
India was already building toward this, in pieces
India's regulatory architecture for synthetic advertising is assembling in parallel, but through two separate, only loosely coordinated tracks. The Ministry of Electronics and IT notified the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 on February 10, which took effect February 20. The amendment writes "synthetically generated information" into intermediaries' due-diligence duties for the first time, requiring visual labels that are "prominent, easily noticeable, and adequately perceivable" and audio content to carry a prefixed disclosure, alongside a sharply compressed takedown clock — court-ordered removals cut from 36 hours to 3, and flagged deepfake-category content cut to 2 hours, according to law firm Khaitan & Co.'s analysis of the notified rules.
Separately, the Advertising Standards Council of India — India's self-regulatory advertising body — released Draft Guidelines for Responsible Labelling of AI-Generated Content in Advertising on May 8, 2026, for stakeholder consultation. ASCI's own announcement confirms the guidelines exist and are framed around consumer impact rather than the technology itself; industry analysis of the draft describes a three-tier structure: fabricated endorsements, unauthorised deepfakes and fake authority figures are prohibited outright as "high risk"; virtual influencers and AI likeness/voice replication are "medium risk" and require disclosure (suggested language: "Audio/Video created using AI"); routine touch-ups sit in a "low risk," no-label tier. MeitY's broader appetite for platform rulemaking hasn't slowed either — an August 20 consultation with the National Commission for Protection of Child Rights on a children's social-media code, reported by MediaNama, shows the ministry is still actively expanding its rulemaking footprint on platforms that carry influencer advertising.
Steelmanning the case for mandatory labelling
The strongest argument for rules like these is not hypothetical. Synthetic celebrity endorsements have already been used to push fraudulent crypto and stock-trading schemes in India, trading on a real person's face and voice without consent. A consumer scrolling Instagram cannot be expected to distinguish a cloned voice from a real one, or know whether a "customer testimonial" was ever spoken by a human being. Disclosure at the point of exposure — not buried in a terms page — is a genuinely low-cost intervention against a real and rising harm, and ASCI's own risk-tiering shows an awareness that not all synthetic content deserves the same treatment.
Where the design goes wrong
The problem is not the principle; it's the plumbing. India now has two regimes converging on the same conduct — MeitY's binding IT Rules obligations on intermediaries, and ASCI's still-draft, non-binding code for advertisers — using different definitions, different risk thresholds, and different enforcement machinery, with no confirmed mechanism to reconcile them. A small creator or performance-marketing agency producing an AI-voiced product demo faces platform-level labelling duties under one track and advertiser-level disclosure expectations under another, with ASCI's "materially influence consumer decisions" test doing a lot of interpretive work that only case-by-case rulings will clarify. The EU's Article 50, whatever its flaws, is at least a single statute with one compliance target. India's approach risks the opposite: two well-intentioned bodies writing overlapping rules on parallel tracks, raising fixed compliance costs that large platforms and agencies can absorb far more easily than independent creators — the same asymmetry that has dogged platform regulation everywhere.
The fix is convergence, not more rules
Before ASCI finalises its guidelines, MeitY and ASCI should publish a single shared definition of what counts as disclosure-triggering synthetic content, and a joint statement on which body enforces what. A proportionate regime — clear labelling for content that could deceive, no labelling burden for routine AI-assisted editing, and one compliance standard rather than two — protects consumers from real fraud without taxing India's creator economy for building on the same generative tools every other market is adopting.