Delhi High Court rules OpenAI ANI training is fair dealing

This ruling emerges against a global pattern of copyright litigation testing whether machine learning ingestion constitutes fair use or fair dealing. Historically, courts in common law jurisdictions have distinguished between expressive copying and non-expressive, functional use of copyrighted works. In text-and-data mining (TDM) contexts, precedents in the EU and Japan have statutorily exempted computational analysis from infringement, while US jurisprudence has relied on transformative use doctrine. The Delhi High Court’s invocation of 'research' aligns with Commonwealth traditions where fair dealing is purpose-specific rather than open-ended. Prior to this case, no Indian court had explicitly applied these exceptions to generative AI training. The pattern suggests a judicial tendency to require proof of market substitution or verbatim reproduction before granting preliminary relief in technology cases. This approach mirrors earlier disputes involving search engine indexing and cached content, where courts frequently denied injunctions absent evidence of direct consumer confusion or identical replication. The current controversy continues this trajectory, treating AI training as a distinct category of use requiring specific evidentiary thresholds for interim restraint.

On July 24, 2026, the Delhi High Court issued a significant interim ruling in the copyright dispute between Asian News International (ANI) and OpenAI, declining to grant ANI’s request for injunctive relief. According to reports from the Economic Times and News India Times, the court determined that OpenAI’s use of ANI’s proprietary news content for training its ChatGPT model did not constitute copyright infringement at this stage of the proceedings. The judgment explicitly classified the ingestion of copyrighted material for large language model development as falling under the 'fair dealing' exception for research purposes within India’s Copyright Act.

The court’s reasoning, as detailed by News India Times, hinged on two primary findings. First, the bench found that ANI failed to demonstrate that ChatGPT had reproduced its original articles verbatim. Second, the court accepted the argument that using such material for AI model training qualifies as research, thereby invoking statutory fair dealing protections. This interim order denies ANI immediate relief but does not dispose of the broader lawsuit, which remains ongoing. The decision marks India's first major judicial articulation regarding the intersection of generative AI training and domestic copyright law.

OpenAI has maintained throughout the litigation that its training practices comply with Indian law, arguing that processing publicly available content for transformative technological development is permissible. Conversely, ANI alleged unauthorized commercial exploitation of its proprietary news archive without licensing or consent. Despite these allegations, the court’s refusal to issue an interim injunction suggests a judicial skepticism toward claims of irreparable harm in the context of AI training, at least where direct reproduction is not evident.

Legal analysts cited by Asteris AI have noted that this ruling provides AI developers with a stronger legal foothold in the Indian market, a jurisdiction where language, media, and data rights are deeply localized. The decision is viewed as establishing a precedent that may insulate AI companies from preliminary injunctions in similar future cases, provided they can demonstrate a research nexus and lack of verbatim output. However, because this is an interim ruling rather than a final judgment on the merits, the long-term legal landscape remains subject to the outcome of the full trial. The court’s emphasis on the absence of verbatim reproduction indicates that future liability may depend heavily on technical evidence regarding model outputs versus training inputs.

Asian News International faces delayed enforcement and must prove verbatim reproduction or market harm at trial. OpenAI secures continued operational freedom in India during litigation, reducing immediate business disruption risk. The ruling affects all AI developers using Indian copyrighted content, potentially lowering licensing costs if fair dealing holds. Magnitude is currently procedural rather than financial: no damages awarded, no injunction imposed, but the precedent could influence outcomes in parallel US/EU cases involving similar fair use/dealing arguments. If upheld finally, it may establish India as a jurisdiction with favorable AI training safe harbors; if reversed, it could trigger retroactive licensing demands across the sector.