Twenty months after ANI sued OpenAI for copyright infringement, Delhi High Court delivered a lengthy interim judgment, pending trial, in favour of the tech company. ANI had made two claims. One, that the San Francisco-based AI company was reproducing copyrighted content belonging to ANI, on its own servers, for the purposes of training its LLMs. Two, that OpenAI was reproducing and displaying its copyrighted content to its users, in response to their queries. OpenAI denied the allegations on both facts and law, including that it was protected under copyright law's fair-dealing doctrine.
The court ruled in favour of the tech biggie on all counts, despite conceding that there were several disputes of facts that could be decided only through a trial. Ideally, when facts are in dispute, the court should direct litigants to a trial. Instead, the court decided to form a prima facie opinion about the facts - basically, an informed guess on the outcome after a trial.
This is a difficult exercise when facts are disputed, and the facts have a huge implication for the law. For example, if OpenAI was using 'text and data mining' techniques on ANI's content, it may not amount to copyright infringement. Reproduction of entire articles, on the other hand, is more likely to count as copyright infringement. When these facts are unclear, forming a prima facie opinion is essentially a coin toss.
On the law, the judgment makes far-reaching conclusions, some of which are evidently erroneous. For example, at one point, the court makes a fundamental error of law by concluding: 'In the present case, it is not the case of ANI that OpenAI obtains copies of ANI's works from unauthorised sources, or by breaking through the paywall of ANI. It is not disputed that OpenAI obtains copies from information/data that is freely available on ANI's website. Therefore, in any event, it cannot be said that OpenAI has obtained an infringing copy of ANI's works.'
This is a fundamental misunderstanding of copyright law. The test for copyright infringement is whether reproduction is authorised by the copyright owner, not whether it is freely available without a paywall.
The more controversial aspect of the judgment is its interpretation of Section 52(1)(a), a fair-dealing clause in Indian copyright law invoked by OpenAI in its defence. This clause declares that the fair dealing of a work for the purposes of 'private or personal use, including research' shall not constitute copyright infringement.
ANI argued that 'private or personal use' applied only in contexts involving humans as litigants. This was a reasonable argument since both these concepts generally pertain to humans and not corporations. Delhi High Court disagreed, and ruled that 'private or personal use' would apply to not just human activity but also machine learning algorithms. It did so by invoking the 'purposive construction' rule. This is a doctrine to be invoked when a literal interpretation of the law results in patently absurd results.
In this case, if OpenAI was precluded from invoking the 'personal and private, including research' exception since it was a corporation, it may have had to negotiate a licence with ANI. This is hardly an absurd result in a case involving copyright law. Except, Delhi High Court decreed that any view limiting 'the exception' to only use 'by a human would be of a regressive view' and limit 'societal progress', and, hence, a purposive interpretation was warranted.
Make no mistake, this is judicial activism, even if it is in favour of a litigant worth $850 bn and not 'the bewildered and the oppressed'. The final limb of the court's analysis was whether OpenAI's use of copyrighted works for 'private or personal use' qualified as 'fair dealing'. The court approaches this issue by first evaluating the economic consequences for ANI, if OpenAI continued with its activities, and second whether public interest is served by OpenAI using the content.
The first is clearly a factual issue requiring evidence to be led through a trial. Nevertheless, without a shred of evidence, the court reached a prima facie opinion that ANI will not be harmed. On the issue of 'public interest', it concluded that the progress of AI contributes to scientific knowledge in a manner that 'resonates with the broader constitutional and statutory commitment' to advancing knowledge. The unwritten subtext being that the copyright owner may be sacrificed.
There are judgments that are the result of the law, and others that are the result of judicial activism. This judgment lies squarely in the latter camp, with the beneficiary being OpenAI and other developers of AI.
