On October 22, 2025, India’s Ministry of Electronics and Information Technology (MeitY) proposed a set of amendments to the existing Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 to respond to the increasing circulation of deepfakes online. So far, deepfakes have been addressed only through publicity rights actions brought typically by celebrities, with no platform-level regulation. I argue that while the publicity rights regime is plagued by overbroad application and doctrinal uncertainty, the proposed rules also reveal several limitations.

The Limits of Publicity Rights

Deepfakes are AI-altered audio, visual, or audio-visual content depicting individuals in a highly realistic manner, generated through deep learning. They include image reconstructions, voice cloning, and pornographic material. As publicity rights protect the commercial appropriation of one’s persona, including name, image, likenesses, and voice, celebrities such as Jackie Shroff, Arijit Singh, Aishwarya Rai, Anil Kapoor, have secured injunctions against platforms and creators for blocking deepfake content. Yet, publicity rights can be an inappropriate avenue to address deepfake harm.

Publicity rights in India lack a coherent doctrinal rationale, with courts frequently conflating the tort of passing off with the constitutional right to privacy. Stemming from the former, publicity rights are commercial in nature and penalise deceptive misrepresentations in trade based on a finding of goodwill; confusion; and resulting damage. Indian courts, however, have repeatedly presumed confusion merely based on the identifiability of a celebrity, without assessing whether consumers were indeed misled. This doctrinally flawed approach, coupled with expansive remedies like dynamic, even ex-parte injunctions, effectively grants celebrities an excessive monopoly over image-related use even when harmless or expressive, thus chilling free speech and trade. For instance, in Mohan Babu, an ex-parte injunction restrained even subscribers of YouTube channels from using the name/image of the actor.

As ex-post facto remedies that don’t prevent deepfake creation, publicity rights actions only help celebrities who can afford expensive litigation, leaving ordinary people vulnerable. Focused on the celebrity’s commercial interests, they also do not address wider misinformation harms that may endanger public safety, such as in the Ankur Warikoo case, where deepfakes offered falsified financial advice, risking serious consumer harm.

Deepfakes involve malicious AI-generated impersonation resulting in privacy, autonomy and dignity harms, which affect vulnerable groups disproportionately. Female celebrities including Rashmika Mandanna, Alia Bhatt, and Katrina Kaif have been subjected to non-consensual sexual deepfakes, a form of technology-facilitated gender-based violence. Publicity rights do not capture these harms that directly threaten the constitutional guarantees under Article 21, as affirmed in Puttaswamy. While there may be an overlap, relying on publicity rights to address deepfakes risks further doctrinal confusion and could suppress legitimate Gen-AI uses.

MeitY’s Platform-Governance Amendments

The proposed amendments mark India’s first step towards deepfake regulation at intermediary level. However, they risk undermining lawful expression.

First, the requirement for platforms enabling the creation of  synthetically generated information (SGI) to label such content with permanent metadata is grounded in an expansive definition. The proposed Rule 2(1)(wa) defines SGI as: ‘artificially or algorithmically created, generated, modified or altered using a computer resource, in a manner that such information reasonably appears to be authentic or true’. While promoting transparency, the definition prioritises technical modification over deceptive intent, sweeping within its ambit even harmless alterations like photo filters, memes, cinematic enhancements, and scientific simulations. Echoing the overreach seen in publicity rights enforcement, this threatens overregulation of creative expressions. By contrast, the EU’s Artificial Intelligence Act focuses on deceptive potential, offering a narrower and more targeted model.

Second, the rule’s effectiveness also depends on whether Gen-AI models qualify as ‘intermediaries’ under the Information Technology Act, 2000 (IT Act). As such an intermediary typically performs a passive role limited to hosting or enabling transmission, rather than content-creation, few AI models would meet the mark, limiting enforceability. This imposes compliance burdens on intermediaries while leaving unregulated model developers, encouraging over-enforcement by platforms without addressing deepfake generation at its source.

Third, the proposed amendments risk undermining the safe harbour principle under Section 79 of the IT Act, which grants intermediaries immunity from liability based on their inability to screen massive volumes of user-generated content. Encouraging proactive content-policing while retaining the immunity fundamentally contradicts the rationale for safe harbour and might incentivise excessive moderation and automated filtering, disproportionately impacting free expression.

The proposed amendments represent a mid-layer regulatory intervention into India’s deepfake crisis,  placing primary obligations on intermediaries rather than upstream developers or trainers who design and deploy technologies capable of deceptive synthetic data generation or the downstream malicious actors who generate and circulate deepfakes. While potentially reducing reliance on expensive litigation, this is likely to encourage precautionary moderation practices resulting in over-censorship and chilling effects on legitimate expression by online creators and speakers, impacting political spoofs, parody, and satire.