Author: JS Shreeya, PES University
LinkedIn Profile: www.linkedin.com/in/shreeya-js-016239378
To the Point:
Artificial Intelligence (AI) has revolutionised the digital ecosystem by enabling the creation of highly sophisticated synthetic media commonly known as deepfakes. Through advanced machine learning models, particularly Generative Adversarial Networks (GANs), deepfakes can generate realistic videos, audio recordings, and images that are often indistinguishable from authentic content. While such technology offers legitimate applications in education, entertainment, accessibility, and digital innovation, its misuse has created significant legal and societal concerns.
Deepfakes are increasingly being employed for political misinformation, financial fraud, identity theft, cybercrime, non-consensual intimate imagery, and reputational attacks. The rapid dissemination of such content through social media platforms has exposed weaknesses in existing intermediary liability frameworks. This raises a critical question: can online platforms continue to enjoy safe harbour protection when their systems facilitate, amplify, and monetise the spread of harmful synthetic content?
Safe harbour provisions were designed to protect intermediaries from liability arising from user-generated content, provided they remained neutral facilitators and complied with statutory obligations. However, the rise of algorithm-driven platforms challenges this traditional understanding. In the age of deepfakes, intermediaries are no longer mere hosts; they increasingly influence visibility, engagement, and dissemination through recommendation systems.
This article examines whether the doctrine of safe harbour remains viable in an era where digital platforms possess unprecedented power over information flows and whether legal frameworks must evolve to ensure accountability without undermining innovation and freedom of expression.
Use of Legal Jargon:
The discourse surrounding deepfakes and intermediary liability engages several significant legal doctrines and principles, including:
• Safe Harbour Protection
• Intermediary Liability
• Due Diligence Requirements
• Actual Knowledge Standard
• Constructive Knowledge Doctrine
• Algorithmic Accountability
• Defamation
• Identity Misappropriation
• Invasion of Privacy
• Vicarious Liability
• Digital Tort Law
• Negligence Standards
• Freedom of Speech and Expression
• Content Moderation
• Platform Governance
• Notice-and-Takedown Mechanisms
• Constitutional Proportionality
• Informational Privacy
Under Section 79 of the Information Technology Act, 2000, intermediaries receive immunity from liability for third-party content if they function merely as facilitators and observe due diligence obligations. However, immunity may be withdrawn where platforms actively participate in unlawful conduct or fail to respond appropriately after receiving notice of illegal content.
The emergence of deepfakes has complicated the distinction between passive hosting and active participation, particularly where recommendation algorithms amplify harmful content to wider audiences.
The Proof:
Deepfakes are generated through sophisticated AI techniques capable of replicating human faces, voices, and expressions with remarkable accuracy. Unlike traditional forms of digital manipulation, modern deepfakes can create entirely fabricated scenarios that appear authentic to ordinary viewers.
The dangers associated with deepfakes are no longer hypothetical.
Political Manipulation:
Political deepfakes have emerged as a significant threat to democratic governance. Manipulated videos portraying political leaders making inflammatory statements or engaging in misconduct can influence voter behaviour, spread misinformation, and undermine public trust in democratic institutions.
During election cycles across several jurisdictions, fabricated videos have circulated widely before fact-checking mechanisms could respond. The viral nature of online platforms enables such content to reach millions of users within hours, causing irreparable damage to electoral integrity.
Financial Fraud:
Deepfake technology has also become a powerful instrument of financial crime. Criminals increasingly employ AI-generated voice cloning to impersonate executives, government officials, and trusted individuals.
Several high-profile incidents have demonstrated how employees were deceived into authorising substantial financial transfers based solely on synthetic audio communications. Such attacks highlight the growing convergence between artificial intelligence and cybercrime.
Non-Consensual Deepfake Content:
One of the most troubling applications of deepfake technology involves the creation of non-consensual intimate imagery. Victims, predominantly women, frequently discover fabricated content depicting them in explicit situations they never participated in.
The psychological trauma, reputational damage, and invasion of privacy resulting from such content often persist long after removal. Existing legal remedies frequently prove inadequate due to jurisdictional challenges and the rapid replication of content across multiple platforms.
Threat to Judicial and Evidentiary Integrity:
Historically, photographs, audio recordings, and videos possessed significant evidentiary value because they were regarded as objective representations of reality. Deepfakes undermine this assumption.
Courts increasingly face challenges concerning authenticity and admissibility. Sophisticated synthetic media can blur the distinction between truth and fabrication, complicating criminal investigations and civil litigation. The existence of deepfakes also creates what scholars describe as the “liar’s dividend,” whereby individuals dismiss authentic evidence as manipulated content.
Platform Amplification:
The legal challenge extends beyond the creation of deepfakes to their amplification. Modern social media platforms utilise recommendation systems designed to maximise engagement. Content that provokes strong emotional reactions often receives greater visibility.
Deepfakes are particularly suited to exploit such systems because sensational content attracts clicks, shares, and comments. Consequently, platforms may indirectly contribute to the spread of misinformation by prioritising engagement metrics over informational accuracy.
Evolution of Safe Harbour: From Passive Hosting to Active Governance
The doctrine of safe harbour emerged during the early development of the internet when service providers merely transmitted or stored information generated by users. Legislators recognised that imposing publisher-level liability on intermediaries would inhibit technological innovation and economic growth.
Accordingly, jurisdictions around the world adopted intermediary immunity frameworks. In India, Section 79 of the Information Technology Act, 2000 protects intermediaries subject to compliance with due diligence obligations.
The rationale underlying such protections was straightforward. Intermediaries lacked the practical ability to monitor every communication occurring on their platforms. They were therefore treated as neutral conduits rather than publishers.
However, the digital landscape has changed dramatically. Contemporary platforms curate, recommend, personalise, rank, and monetise content through algorithmic systems. These functions influence public discourse and shape user behaviour.
The central question therefore, becomes whether a platform that actively determines content visibility can continue to claim neutrality. Deepfakes have intensified this debate because algorithmic amplification frequently contributes to the widespread dissemination of harmful synthetic content.
Deepfakes and the Emerging Crisis of Digital Trust:
Deepfakes present a challenge not merely to individual rights but to the broader concept of trust within digital societies.
Information ecosystems depend upon a shared assumption that audiovisual evidence generally reflects reality. Deepfakes undermine this assumption by making fabrication accessible, inexpensive, and scalable.
The resulting erosion of trust affects multiple institutions:
• Democratic processes
• Journalism and media
• Corporate governance
• Criminal investigations
• Judicial proceedings
• Public administration
When citizens can no longer distinguish between authentic and manipulated content, confidence in information itself begins to deteriorate. This phenomenon threatens the foundational principles upon which democratic societies operate.
Moreover, deepfakes may be weaponised by state and non-state actors to influence public opinion, destabilise institutions, and create social unrest. Their capacity to exploit existing political and social divisions makes them particularly dangerous in polarised environments.
The Indian Regulatory Framework and Deepfake Governance:
India currently lacks dedicated legislation regulating deepfakes. Nevertheless, several statutory provisions may be invoked depending upon the nature of the harm involved.
Section 66C of the Information Technology Act criminalises identity theft, while Section 66D addresses cheating by personation through computer resources. Deepfakes involving impersonation or fraudulent representations may therefore attract liability under these provisions.
Similarly, defamation provisions under the Bharatiya Nyaya Sanhita, 2023 may be utilised where synthetic content damages an individual’s reputation.
The constitutional right to privacy recognised in Justice K.S. Puttaswamy v. Union of India provides an additional basis for challenging unauthorised use of a person’s likeness, image, or voice.
Furthermore, the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 require intermediaries to exercise due diligence and establish grievance redressal mechanisms. However, these regulations were formulated before the explosive growth of generative AI technologies and may not adequately address the unique challenges posed by deepfakes.
The absence of a comprehensive statutory framework creates uncertainty regarding platform obligations, victim remedies, and enforcement mechanisms.
Comparative International Approaches:
Different jurisdictions have adopted varying approaches to addressing deepfake-related harms.
United States:
The United States continues to rely primarily upon existing legal frameworks such as privacy law, intellectual property law, defamation law, and Section 230 of the Communications Decency Act.
Several states have enacted legislation targeting election-related deepfakes and non-consensual synthetic imagery. Nevertheless, debates continue regarding whether broad intermediary immunity remains appropriate in the context of algorithmic amplification.
European Union:
The European Union has adopted a more interventionist approach through the Digital Services Act and the Artificial Intelligence Act.
These frameworks impose transparency obligations, risk assessment requirements, and accountability mechanisms upon large online platforms. The emphasis is on proactive governance rather than reactive enforcement.
China:
China has implemented regulations requiring synthetic media to be clearly labelled and identifiable. Service providers must undertake measures to prevent misuse and ensure transparency concerning AI-generated content.
Although it differs significantly from liberal democratic models, the Chinese approach reflects a growing global willingness to regulate synthetic media directly.
Challenges in Imposing Liability on Intermediaries:
Despite growing demands for accountability, imposing liability on intermediaries presents several practical and constitutional challenges.
First, billions of pieces of content are uploaded daily, making comprehensive monitoring difficult.
Second, automated detection technologies remain imperfect. Deepfake detection systems frequently generate both false positives and false negatives.
Third, excessive liability may incentivise over-censorship. Platforms may remove lawful content merely to avoid legal risk, thereby undermining freedom of expression.
Fourth, establishing causation is often complicated because content may pass through multiple platforms before causing harm.
Therefore, any reform of safe harbour protections must carefully balance accountability against innovation, privacy, and constitutional freedoms.
The Need for Algorithmic Accountability:
A growing body of scholarship argues that the future of intermediary liability should focus not only on content but also on the systems responsible for its dissemination.
Algorithms significantly influence public discourse by determining which content receives visibility. Where platforms knowingly deploy systems that amplify harmful deepfakes, claims of neutrality become increasingly difficult to justify.
Several reforms have therefore been proposed:
• Mandatory transparency reports.
• Independent audits of recommendation systems.
• Deepfake detection and labelling requirements.
• User notification mechanisms.
• Rapid grievance redressal systems.
• Enhanced due diligence obligations for large platforms.
Such measures seek to preserve innovation while ensuring that platforms assume responsibility proportionate to their influence.
Abstract:
Safe harbour provisions were established to protect intermediaries from liability arising from user-generated content and to facilitate the growth of the digital economy. However, the emergence of deepfake technology has fundamentally altered the digital landscape. Deepfakes challenge conventional assumptions regarding intermediary neutrality because modern platforms actively curate, recommend, and amplify content through sophisticated algorithmic systems.
This article argues that traditional safe harbour frameworks are increasingly inadequate in addressing harms associated with synthetic media. Through an examination of Indian and international jurisprudence, intermediary liability principles, constitutional considerations, and emerging regulatory approaches, the article contends that greater platform accountability is necessary. While complete abolition of safe harbour protections would be undesirable, enhanced due diligence obligations and algorithmic accountability mechanisms are essential to ensure that digital innovation does not come at the expense of truth, privacy, and justice.
Case Laws:
1. Shreya Singhal v. Union of India (2015) 5 SCC 1
The Supreme Court upheld Section 79 of the Information Technology Act while clarifying that intermediaries are required to remove unlawful content only upon receiving a court order or notification from an appropriate governmental authority.
The judgment remains foundational to intermediary liability jurisprudence in India. However, the speed with which deepfakes spread raises concerns regarding the effectiveness of post-facto judicial intervention.
2. MySpace Inc. v. Super Cassettes Industries Ltd. (2016)
The Delhi High Court held that intermediaries are not expected to proactively monitor all user-generated content. Nevertheless, immunity remains contingent upon compliance with due diligence obligations.
The reasoning becomes increasingly relevant where platforms possess technological capabilities capable of detecting harmful synthetic media.
3. Christian Louboutin SAS v. Nakul Bajaj (2018)
The Court distinguished passive intermediaries from entities that actively participate in content promotion or optimisation. Active involvement may result in the loss of safe harbour protection.
This principle has significant implications for platforms employing recommendation algorithms.
4. Justice K.S. Puttaswamy v. Union of India (2017) 10 SCC 1
The Supreme Court recognised privacy as a fundamental right under Article 21 of the Constitution.
Deepfakes involving unauthorised use of a person’s likeness, voice, or image directly implicate informational privacy and individual autonomy.
5. Google India Pvt. Ltd. v. Visaka Industries (2020)
The Supreme Court reiterated that intermediary immunity is conditional rather than absolute and depends upon statutory compliance.
The judgment supports the proposition that platforms may lose protection where they fail to exercise reasonable diligence.
6. Delfi AS v. Estonia (2015)
The European Court of Human Rights upheld liability against an online platform for harmful third-party content where the platform exercised significant control over dissemination.
The case reflects growing judicial willingness to scrutinise intermediary conduct beyond passive hosting.
Conclusion:
The doctrine of safe harbour was conceived during a period when online intermediaries functioned primarily as passive conduits of information. The rise of deepfake technology has fundamentally altered this reality. Today, digital platforms possess unprecedented influence over the creation, visibility, and dissemination of information.
While safe harbour protections remain essential for preserving innovation and freedom of expression, blanket immunity is increasingly difficult to justify where platforms actively amplify harmful synthetic content through algorithmic systems. The law must therefore evolve to reflect contemporary technological realities.
Future regulatory frameworks should emphasise transparency, accountability, due diligence, and algorithmic responsibility while safeguarding constitutional freedoms. Dedicated deepfake legislation, improved detection technologies, expedited grievance mechanisms, and platform-specific obligations may provide a balanced path forward.
Ultimately, the challenge posed by deepfakes extends beyond technology. It is a test of whether legal systems can preserve truth, privacy, and democratic integrity in an era where seeing is no longer believing. If the law fails to adapt, society risks entering a digital environment where authenticity becomes suspect and individuals are effectively treated as “guilty until proven checked.”
