Retrofitting the Law: Delhi HC’s AI "Fair Dealing" Call and the 3-Hour Takedown Rule Upend Indian Tech Practice

For technology and intellectual property lawyers in India, the writing on the wall for 2026 is unmistakably clear: stop waiting for a bespoke Artificial Intelligence Act. The Ministry of Electronics and Information Technology (MeitY) has firmly signaled its intent to govern the frontier of AI not through new, tailor-made legislation, but by aggressively force-fitting generative AI into existing statutory frameworks. This regulatory "retrofitting" has turned the Copyright Act of 1957, the Information Technology Act of 2000, and the nascent Digital Personal Data Protection (DPDP) Act of 2023 into the most heavily contested battlegrounds in Indian commercial law.

Three massive developments this year—the Delhi High Court’s stance on AI training, a draconian amendment to the IT Rules, and the Supreme Court’s scrutiny of the DPDP-RTI clash—demand immediate strategic pivots from practicing advocates and in-house counsels.

The Section 52(1)(a) Safe Harbor: A Bitter Pill for Media Houses

The most consequential IP development of 2026 is undeniably the Delhi High Court’s prima facie view in the ongoing dispute between news agency ANI and OpenAI. The Court indicated that OpenAI’s ingestion of ANI’s literary works to train its Large Language Models (LLMs) falls squarely within the "fair dealing" exception under Section 52(1)(a) of the Copyright Act.

The Court reasoned that such ingestion is limited to backend training, does not substitute the commercial market for the original news reports, and serves a broader public interest in technological innovation. This is a monumental shift. For years, Indian IP practitioners have debated whether Text and Data Mining (TDM) constitutes infringement. By pegging AI training to the "fair dealing" doctrine rather than requiring an explicit statutory exception, the judiciary is handing AI developers a massive operational shield.

"If the Delhi High Court’s prima facie view holds, traditional media and publishing houses must abandon basic infringement claims against AI developers and instead bear the heavy evidentiary burden of proving actual market substitution."

Practice Point: If you are representing content creators or media houses, sending standard cease-and-desist notices for copyright infringement based merely on web scraping is no longer viable. Your litigation strategy must now pivot to demonstrating how the AI output directly cannibalizes your client's specific commercial market. Conversely, lawyers advising AI startups should aggressively audit their clients' training datasets to ensure outputs cannot be classified as derivative works that compete with the original sources.

The 3-Hour Takedown: Intermediary Liability on a Knife’s Edge

While the judiciary is offering safe harbors to AI developers, the executive is tightening the noose on intermediaries. The February 10, 2026, amendment to the IT Rules has dramatically lowered the takedown threshold for synthetically generated information (deepfakes and AI hallucinations). More alarmingly, it has compressed the response time for certain notice-based takedowns to a staggering three hours.

This effectively neuters the traditional safe harbor protections under Section 79 of the IT Act. A three-hour window to adjudicate the veracity of synthetic content is a logistical nightmare, especially given that distinguishing between malicious deepfakes and legitimate AI-generated satire is highly subjective.

Practice Point: In-house counsels at social media platforms and digital intermediaries must immediately overhaul their grievance redressal architectures. Human review within three hours is practically impossible at scale, meaning platforms will be forced to rely on automated, algorithmic takedowns. Lawyers must prepare for a surge in writ petitions challenging arbitrary content removal, as political entities and corporations will inevitably weaponize these three-hour notices to silence legitimate criticism.

DPDP vs. RTI: The Supreme Court Steps In

On the privacy front, the Supreme Court’s February 16, 2026, decision to refer the constitutional challenges against the DPDP Act’s amendment of the Right to Information (RTI) Act to a larger bench is a critical holding pattern. However, the Court pointedly refused to stay the operation of the DPDP Act.

The controversy stems from Section 44(3) of the DPDP Act, which amends Section 8(1)(j) of the RTI Act. Previously, personal information could be disclosed under the RTI Act if it served a larger public interest. The DPDP amendment obliterates this public interest caveat, creating a blanket exemption for the disclosure of personal information. Petitioners rightly argue this severely cripples the transparency regime built over the last two decades.

Simultaneously, the Centre has notified the administrative rules for the DPDP Act. With the substantive commencement of the Act expected by May 13, 2027, and the consent-management infrastructure mandated by November 13, 2026, corporate India is on a strict countdown.

Practice Point: The era of broad, bundled consent is dead. Under the DPDP Act, lawyers must draft consent notices that are free, specific, informed, unconditional, and unambiguous. Crucially, the withdrawal of consent must be as frictionless as providing it. Corporate lawyers must immediately audit their clients' data architectures. If your client requires users to navigate three different menus to withdraw consent that took one click to give, they are already in breach of the upcoming compliance mandates.

Ultimately, 2026 is the year Indian tech law lost its novelty and became a brutal exercise in statutory interpretation. Whether you are defending an LLM under the Copyright Act, racing against a three-hour IT Rules takedown, or building a DPDP-compliant consent manager, the legal engineering required today is more complex and high-stakes than ever before.