Delhi High Court grants Vijay Deverakonda an interim injunction against AI and deepfake misuse of his persona

Key takeaways

• The order is interim and ex parte. It applies "till the next date of hearing" and makes no final determination of liability.
• Paragraph 59(i) names AI, generative AI, machine learning, deepfakes and face morphing as covered technologies.
• Marketplaces (defendants 5 to 8), registrars (9 to 16), MeitY and the Department of Telecommunications have 72 hours from being communicated a copy of the order by the plaintiff. Affidavits are due within two weeks from defendants 5 to 8 and, as the order reads, "defendant nos. 9 and 16"; a compliance affidavit is also directed for the two departments. Meta, Google LLC and X have no hour limit stated.
• The Court declined, for now, passing-off protection for the nickname "Rowdy" but covered AI-morphed images that use it.

What did the Delhi High Court order, and is it final?

It is an interim order. On 8 October 2026, Justice Vikas Mahajan granted the plaintiff, the actor Vijay Sai Deverakonda, an ex parte ad-interim injunction "till the next date of hearing". The Court found a prima facie strong case, a balance of convenience in his favour and a risk of irreparable loss "may not be compensated in monetary terms" (paragraph 54). It also noted at paragraph 55 that, for the time being, counsel was pressing only prayers (a) and (g) of the application.

The order makes no final determination of liability, although paragraph 58 records an interim view that the Annexure-A content shows unauthorised use of the plaintiff's attributes. The Court issued notice, allowed four weeks for replies, listed the suit before the Joint Registrar on 3 December 2026 and before the Court on 2 February 2027. The order records the plaintiff's account of fake handles, merchandise, wallpaper sites, unauthorised apps and AI-generated or morphed videos, and describes those as contentions.

What conduct does the injunction cover when AI is involved?

Paragraph 59(i) restrains the defendants and persons claiming through them from using the plaintiff's name and initials "VD", voice recordings and duplications, signatures, fashion style, photographs, movie stills, performance and other attributes exclusively identifiable with him, without his consent, "through the use of any technology including but not limited to Artificial Intelligence, Generative Artificial Intelligence, Machine Learning, Deep-Fakes, Face Morphing". The clause also reaches GIFs and any medium, listing websites, mobile apps, the Metaverse and social media.

The Court drew one line on the nickname. At paragraph 56 it declined, at this stage, an ex parte passing-off injunction on "Rowdy", because acquired distinctiveness has to be shown on material. It added that using an AI-generated or morphed image of the plaintiff with that expression "will come within the ambit of violation of the personality rights of the plaintiff". At paragraph 58 it also observed that some Annexure-A URLs contain abusive, obscene and vulgar content about him.

Who has to act under the order, and by when?

Paragraph 59 sets different steps for different defendants. The order lists the content and domains by URL in Annexure-A.

The 72-hour trigger is the Court's phrase "within 72 hrs of being communicated a copy of this order by the plaintiff". The registrar clause directs "Defendant nos. 9 to 16" to lock and suspend domains, but then says "Defendant nos. 9 and 16 will file an affidavit of compliance within two (2) weeks". We quote it as written and do not resolve the difference. For marketplaces, the affidavit sentence follows the 72-hour sentence without stating its own trigger, and for MeitY and the Department of Telecommunications the order says only "within two (2) weeks".

What can a blocked site or an intermediary do if the order reaches too far?

The order includes two release valves. Under paragraph 60, defendants 2 to 16 may approach the plaintiff to ask in what manner a listed URL is covered. Under paragraph 61, a website or account that is not primarily an infringing one and is blocked or locked may approach the Court with an undertaking that it does not intend to infringe the plaintiff's personality rights, after which the Court would consider modifying the injunction.

For new content, paragraph 62 leaves the plaintiff free to notify the relevant intermediary or take other legal steps. Whether that notice alone triggers any duty on a platform is not addressed in the order.

What should platforms, marketplaces and counsel do now?

The specific takedown steps are addressed to named defendants, so the first question for any intermediary is whether it is one. Paragraph 59(i) is wider: it is addressed to defendants (including John Does), associates, persons claiming through them "and all other persons". Whether it applies to a given non-party is for counsel.

• Named defendants: record the date and time the plaintiff's copy of the order is received, since the 72-hour periods (defendants 5 to 18) run from that communication, and diarise any two-week affidavit period that applies to you. Confirm with counsel how the affidavit clause for defendants 9 to 16 applies to you.
• Check each Annexure-A entry against what you host or register before acting, and use the paragraph 60 route for any item you think is outside the order.
• Other platforms and registrars: review how your intake process handles complaints about AI-generated likeness and voice, using this order as an example of the relief a Delhi court has granted, and take advice on paragraph 59(i)'s "all other persons" wording.
• IP and media counsel: note that the injunction text lists AI techniques by name, which can be reused in drafting prayers. The Court granted it at the interim stage on the plaintiff's material only.
• Talent and brand teams: keep dated evidence of each AI or morphed item, since the Annexure-A format ties each URL to a platform and a pleaded category.

Which dates matter in this order?

What we did not verify

What we opened: the full 30-page order as a PDF (published by LiveLaw), including paragraphs 53 to 69 and Annexure-A, on 25 October 2026 (UTC).

What we did not open: the plaint and its documents, the Delhi High Court order portal copy, any defendant's reply, any compliance affidavit, and any later order. We did not check whether any defendant has complied or challenged the order. The order cites no statute for the takedown directions, so we state none. We did not count or test the Annexure-A URLs.

What we refuse to claim: that the order is a final finding, that it does or does not bind anyone who is not a party (paragraph 59(i) names "all other persons"), that it sets a rule for all AI content in India, or that the allegations in the plaint have been proved. Paragraph 59(iv) directs defendants 9 to 16 to lock domains but names "defendant nos. 9 and 16" for the affidavit; we have not confirmed the Court's intent. The order does not state what triggers the two-week affidavit period.

Related briefings

• AI Regulation News hub
• Asia-Pacific AI regulation tracker
• Attorneys and AI regulation
• Delhi High Court personality-rights takedown order in the Ravi Kishan matter
• India IT Rules on synthetic media labelling
• MeitY AI governance guidelines on actor liability
• Beijing Internet Court on AI face swapping and personality rights