AI Deepfakes Are Outpacing the Law: Can India’s New Rules Close the Enforcement Gap?

The rapid spread of artificial intelligence-generated deepfakes is forcing governments to reconsider who should be responsible when fabricated videos, cloned voices and manipulated identities cause real-world harm. India has moved to tighten oversight of synthetic media, placing greater responsibility on digital platforms to identify, label and remove deceptive content. However, the country's emerging regulatory approach exposes a deeper challenge: preventing technological abuse without restricting legitimate expression or overwhelming enforcement institutions.

A study titled "Legal Approaches to Deepfakes: Contextualising India's Intermediary Guidelines Amendments in the Global Regulatory Environment," published in the journal AI for Society, examines this challenge through a comparative assessment of India, China, the European Union and the United States. Authored by José-Miguel Bello y Villarino of the University of Sydney and Digvijay Singh Rana of O. P. Jindal Global University, the research argues that India's 2026 amendments establish a distinctive hybrid model of deepfake regulation.

The researchers consider India's approach a reasonable early intervention, combining content labelling, technical safeguards, accelerated removal and existing legal remedies. However, they caution that its success remains uncertain because enforcement capacity, constitutional protections and the technological limitations of detection systems could determine whether regulatory ambition translates into effective protection.

From Fabricated Videos to Real-World Harm: Why Deepfakes Demand Legal Intervention

Deepfake technology has fundamentally altered the economics of digital manipulation. Realistic synthetic videos and cloned audio that previously required substantial technical expertise can now be produced using relatively accessible software. According to the study, this expansion has transformed deepfakes from a specialised technological capability into a widespread regulatory challenge involving privacy, financial security, political communication and personal identity.

The researchers distinguish deepfakes from ordinary digital editing because convincing synthetic audio and video can create the impression that identifiable people said or did things that never occurred. Their apparent authenticity can influence human decisions, intensify reputational damage and enable deception. Nevertheless, the legal significance of such content depends on its context, intended purpose and potential consequences rather than the technology alone.

India's experience illustrates the diversity of these threats. The study examines political videos involving manipulated speeches and cloned voices, including fabricated content associated with electoral campaigns. Such material can distort political communication, falsely attribute statements to public figures and introduce uncertainty into democratic debate, even when the underlying manipulation is eventually identified.

Other incidents demonstrate the personal and financial consequences of synthetic impersonation. The researchers discuss the non-consensual deepfake involving actress Rashmika Mandanna, legal action protecting actor Anil Kapoor's personality rights, and fraudulent investment promotions misusing the identities of prominent business figures. One documented case involved a Bengaluru woman reportedly losing ₹3.75 crore after encountering a fraudulent investment promotion featuring an AI-generated video of spiritual leader Sadhguru.

These examples reveal why deepfake governance cannot be reduced to misinformation control. Electoral deception, sexual exploitation, identity appropriation and financial fraud involve different legal interests and require different remedies. The research therefore treats India's regulatory response as an attempt to address the circulation of harmful content while preserving existing criminal, consumer-protection, electoral and personality-rights mechanisms.

Three Hours to Act: India Shifts Responsibility Towards Digital Platforms

India's Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, notified on February 10, 2026, and effective from February 20, introduced specific obligations concerning synthetically generated information. Rather than establishing a standalone criminal statute for deepfakes, the amendments regulate intermediaries involved in enabling, hosting, publishing or distributing synthetic content.

A key feature is the reduction of the applicable takedown deadline from 36 hours to three hours. The changes also require appropriate technical measures to prevent specified unlawful synthetic content, alongside prominent labelling of permitted synthetic material. Where technically feasible, intermediaries must preserve provenance information, including identifiers associated with the resources used to generate or modify content.

Significant social media intermediaries, identified in the study as platforms exceeding five million users in India, face additional responsibilities. They must obtain user declarations concerning synthetic content, deploy appropriate technical measures to verify those declarations and display labels when content is confirmed as synthetic.

The regulatory framework also connects compliance to Section 79 of the Information Technology Act, which provides conditional protection against liability for third-party content. Failure to observe prescribed due-diligence obligations can undermine this protection, giving platforms a stronger legal incentive to implement preventive measures and respond to unlawful material.

The researchers describe this architecture as hybrid because it combines technical prevention, transparency and removal requirements with existing legal protections for victims. Labelling addresses uncertainty about authenticity, takedowns limit continued exposure, and criminal or civil mechanisms remain responsible for dealing with perpetrators and individual remedies.

However, the distinction between removing harmful material and delivering justice remains substantial. Removing a fabricated video does not necessarily identify its creator, recover money lost through fraud or compensate someone whose reputation has been damaged. The effectiveness of India's framework consequently depends on coordination between online intermediaries, enforcement authorities and the wider legal system.

China, Europe and America Offer Competing Blueprints for Controlling Synthetic Media

The international comparison places India's approach within three established regulatory traditions. China's framework concentrates on the provision and distribution of synthetic content, imposing requirements on services responsible for creating and circulating manipulated media. Its 2025 labelling measures strengthened obligations concerning visible disclosures and embedded metadata.

The European Union approaches deepfakes through broader AI governance. Its AI Act requires disclosure of relevant artificially generated or manipulated content, while recognising specific considerations for artistic, fictional and satirical material. The study also examines European provisions addressing non-consensual intimate synthetic imagery, demonstrating how general transparency requirements can coexist with prohibitions targeting serious abuse.

The United States has followed a more targeted federal approach. Its 2025 TAKE IT DOWN Act addresses non-consensual intimate imagery, including qualifying digitally forged depictions, and requires covered platforms to remove reported material within 48 hours following valid requests. State-level measures additionally address election-related deception, digital replicas and other forms of manipulated media.

India combines elements found across these systems without reproducing any one model. Its labelling and provenance obligations resemble approaches used in China and Europe, while accelerated removal shares characteristics with the American emphasis on platform response. The distinctive feature is the integration of these responsibilities into India's existing intermediary-liability framework.

The comparison also reveals different regulatory trade-offs. Broad preventive requirements may capture more harmful content but create greater compliance burdens. Narrowly targeted prohibitions can provide clearer legal boundaries yet leave other forms of deception outside their immediate scope.

For countries considering their own deepfake legislation, India's experience illustrates the potential of incremental regulatory intervention. The researchers nevertheless avoid presenting it as a universal template, stressing that institutional context and implementation capacity remain decisive.

The Enforcement Paradox: Stronger AI Rules Could Threaten Legitimate Expression

The study's most consequential warning concerns the distance between regulatory design and operational effectiveness. Mandatory labels, provenance systems and automated detection may strengthen platform accountability, but none guarantees that sophisticated manipulation will be identified accurately or harmful content removed consistently.

Provenance information can be stripped or circumvented, while labelling systems may create misleading assumptions about authenticity. People encountering unlabelled content could incorrectly interpret the absence of a warning as evidence that material is genuine. Disclosure requirements can also affect perceptions of harmless or beneficial AI-generated content.

Accelerated takedowns introduce another risk. Decisions made under compressed deadlines may disproportionately affect satire, journalism, political commentary or other lawful expression, particularly when platforms rely heavily on automated moderation. The researchers therefore emphasise constitutional proportionality, procedural safeguards and effective mechanisms for challenging removal decisions.

Institutional capacity presents an equally significant obstacle. Authorities need forensic expertise, technical infrastructure and specialised personnel to assess whether platforms genuinely comply with detection and provenance requirements. The study proposes considering a dedicated Digital Content Enforcement Unit within CERT-In or the Ministry of Electronics and Information Technology to support audits, common standards and enforcement coordination.

Protection against gender-based digital abuse also requires further attention. The researchers observe that prominent Indian incidents reveal serious harms affecting women, yet the regulatory architecture does not explicitly reflect the gendered dimensions of synthetic-media exploitation. Whether accelerated removal and technical prevention provide sufficient protection remains unresolved.

Cross-border enforcement adds further complexity because manipulated content can be generated outside India and circulated through international platforms. The researchers identify opportunities for cooperation on technical standards, evidence sharing and regulatory coordination, including engagement with European and American systems.

India's approach ultimately reflects a broader dilemma in governing rapidly evolving technologies. Early intervention can address urgent harms while preserving opportunities for regulatory adjustment, but expanding obligations without credible enforcement risks producing rules that are ambitious on paper and inconsistent in practice.

The study's assessment is therefore deliberately conditional. India has established a substantive regulatory structure for synthetic media, but its effectiveness cannot yet be demonstrated. Future evaluation must examine enforcement performance, automated moderation, remedies for victims and safeguards for legitimate content.