AI Content Output, Deepfake Liability, Generative AI Torts & Copyright Governance

The Legal Crisis of Synthetic Media and Generative AI

Generative Artificial Intelligence (GenAI) has transformed digital content creation, enabling autonomous software agents and neural networks to generate photorealistic imagery, synthetic voices, software code, and literary works in seconds. However, this technological leap has fundamentally destabilized established legal paradigms. Legal systems worldwide—and Indian jurisprudence in particular—were structured around the foundational presumption that creative works and actionable torts emanate from human agency.

Today, artificial intelligence algorithms can scrape billions of copyrighted works without authorization to train commercial foundation models, clone corporate executive voices to execute multi-crore business email compromise (BEC) wire transfers, and create sexually explicit synthetic deepfakes targeting public figures and private citizens. In the absence of an overarching, dedicated Artificial Intelligence statute in India, the legal governance of AI content output is enforced through an intricate synthesis of the Information Technology Act, 2000, the Copyright Act, 1957, the Bharatiya Nyaya Sanhita, 2023, and specialized regulatory circulars issued by the Ministry of Electronics and Information Technology (MeitY).

Our chambers provide strategic legal advisory and high-stakes courtroom representation for enterprises navigating the frontier of artificial intelligence. We assist frontier AI startups in architecting legally defensible training pipelines, defend social media platforms against intermediary liability claims, and protect victims of malicious deepfake defamation through rapid emergency High Court injunctions.

Copyright in the Age of Generative AI: Authorship and Infringement

The intellectual property implications of generative AI split into two distinct, high-stakes legal battlegrounds under Indian copyright law:

1. The Authorship Dilemma (Section 2(d)(vi)): Under Section 2(d)(vi) of the Copyright Act, 1957, in relation to any literary, dramatic, musical, or artistic work which is "computer-generated", the author is defined as "the person who causes the work to be created". This statutory formulation, drafted decades prior to deep learning neural networks, creates severe ambiguity. Does the "person who causes the work to be created" refer to the developer of the AI foundation model, the enterprise fine-tuning the model, or the end-user crafting the natural language prompt? The Indian Copyright Office has historically maintained a strict human authorship requirement, refusing sole registration to autonomous algorithms (as demonstrated in the RAGHAV Artificial Intelligence Painting application). Our chambers structure prompt-engineering workflows and human-in-the-loop creative processes to substantiate legal copyright ownership for enterprise clients.
2. Training Data Ingestion and Copyright Infringement: Foundation models scrape petabytes of copyrighted text, source code, and artistic portfolios from the public internet. AI developers routinely argue that model training constitutes "fair dealing" under Section 52(1)(a) of the Copyright Act for research or private study. However, Indian fair dealing jurisprudence is far more restrictive than the broad US "fair use" doctrine. Commercial model training that ingests proprietary content without a license creates acute exposure to systemic copyright infringement under Section 51, entitling copyright proprietors to seek impounding of trained model weights and statutory account of profits.

Deepfakes, Voice Cloning, and Criminal Tort Liability

Deepfakes—synthetic video, audio, or imagery created using generative adversarial networks (GANs) or diffusion models—pose acute risks to corporate reputation, public order, and individual liberty. When rogue actors deploy deepfakes, multiple statutory provisions are simultaneously engaged:

• Cheating by Personation (Section 66D IT Act & Section 318 BNS): Synthetically cloning the voice of a Managing Director or Chief Financial Officer to instruct accounting staff to initiate urgent wire transfers constitutes cognizable cyber-fraud and cheating by personation, punishable with rigorous imprisonment up to three years.
• Non-Consensual Deepfake Imagery (Section 66E, 67 & 67A IT Act): Generating or transmitting sexually explicit synthetic depictions of individuals violates bodily privacy under Section 66E and constitutes a non-bailable offense under Section 67A, carrying mandatory imprisonment up to five years for a first conviction.
• Personality Rights and Passing Off: In landmark rulings such as Amitabh Bachchan v. Rajat Nagi [CS(COMM) 819/2022] and Anil Kapoor v. Simply Life India [CS(COMM) 652/2023], the Delhi High Court firmly established that a celebrity's name, voice, likeness, image, and digital persona cannot be misappropriated or synthetically simulated through AI tools without express commercial authorization. The High Court granted broad ex-parte omnibus injunctions restraining deepfake generation and AI voice cloning tools globally.

Statutory & Regulatory Framework Matrix

MeitY Generative AI Advisories and Intermediary Due Diligence

In response to escalating deepfake incidents during national elections and viral digital fraud, MeitY issued comprehensive advisories to all digital intermediaries in November 2023 and March 2024 under the IT Rules, 2021:

• Mandatory 24-Hour Nudity Takedown: Under Rule 3(2)(b) of the IT Rules 2021, intermediaries must remove or disable access to any non-consensual synthetic depiction of an individual in a state of nudity, sexual act, or impersonation within 24 hours of receiving a complaint. Failure to do so results in instant forfeiture of Section 79 Safe Harbor.
• Watermarking and Synthetic Content Labelling: MeitY advisories mandate that intermediaries deploying or facilitating generative AI tools must embed permanent, machine-readable metadata, unique identifiers, and visible watermarks indicating that the output is synthetically created or modified by AI.
• Permissibility and Model Testing: Digital platforms hosting untested or unreliable AI foundation models are cautioned against deploying systems that generate unlawful content, bias, or threaten the integrity of electoral processes. Platforms must ensure their Terms of Service explicitly inform users that uploading or generating deepfakes is an offense punishable under criminal law.

Emergency Court Remedies for Deepfake Victims and Brand Owners

When an enterprise or individual is subjected to a viral deepfake campaign, standard civil litigation is wholly inadequate—content must be scrubbed within hours. Our chambers execute rapid-response High Court litigation strategies:

1. Filing Commercial Suits for Personality Infringement and Defamation: We initiate urgent commercial suits before the High Court, presenting expert forensic affidavits under Section 63 of the Bharatiya Sakshya Adhiniyam, 2023 demonstrating synthetic artifact anomalies and spectral voice deviations.
2. Securing Dynamic John Doe Injunctions: We secure ex-parte ad-interim injunctions ordering social media platforms (X, Meta, YouTube, Telegram) to immediately scrub the infringing synthetic clips, suspend offending channel URLs, and disable algorithmic dissemination.
3. DoT and ISP Blocking Orders: High Court orders are served directly upon the Department of Telecommunications (DoT) and Ministry of Electronics and IT to issue emergency blocking directives to Indian internet service providers, terminating access to hosting domains within 24 hours.

Frequently Asked Questions