On the Pen Losing Its Might: Contemplating the Digital News Industry’s Fate in Light of ANI v OpenAI
SpicyIP
On the Pen Losing Its Might: Contemplating the Digital News Industry’s Fate in Light of ANI v OpenAI
By Subhalaxmi Mukherjee
“The first rule of any technology is that creative destruction is a beautiful thing.”
– Marc Andreessen
A single-judge bench of the Delhi High Court is presently hearing submissions in the ongoing matter between Asian News International (“ANI”) and OpenAI (previously discussed here, here and here), including those advanced by Senior Advocate Rajshekhar Rao on behalf of the Digital News Publishers Association (“DNPA”). The DNPA, comprising the digital arms of traditional legacy media organisations, intervened in January 2025 in this highly contentious dispute, aligning itself with ANI’s allegations that OpenAI unlawfully trained its models on copyrighted content produced by these media houses. Interestingly, its counsel underscored that permitting generative artificial intelligence (“Gen-AI”) companies such as OpenAI to continue their practices without restraint would prompt the decline of digital news, driven by shrinking web traffic, diminished visibility, and the consequent loss of revenue. Pursuant to this, DNPA is advocating for a fair compensation regime for publishing houses.
This blog explores this argument through two questions: first, whether the establishment of such a regime is possible at all; and second, whether such a mechanism would necessarily resolve the challenges faced by digital news today, or whether they must look towards adaptation strategies.
Understanding the Stakes: The Context Behind DNPA’s Argument
Before deliberating upon DNPA’s argument and its viability as a cure-all, we must first understand why it is being advanced at all. The answer lies in how digital news websites make money, and the manner in which Gen-AI models are walling off that revenue stream.
Digital news platforms typically operate on a subscription-based model, an ad-based model or a hybrid of both. The subscription-based model generates revenue by charging customers a recurring fee for continuous access to content. Conversely, the ad-based model earns income by selling advertising space to other businesses. Notably, a high volume of website traffic and strong web visibility are essential prerequisites for the ad-based revenue model. AI-generated news summaries and snapshots directly impact media outlets that rely on this model by pulling traffic and visibility away from their websites. This effect is most clearly established in the case of ad-based sites (see here and here); however, it is reasonable to assume that subscription-based models may also be affected. A decrease in casual visitors, who are often potential subscribers, could logically lead to a decrease in new subscriptions. However, empirical data on this specific point remains sparse.
Illustrating my earlier point, the Pew Research Centre recently published a small-sample study, which found that users who encountered an AI summary (in this case, the Google AI Overview) clicked on a website in the search results only about half as often as users who browsed without such AI intervention. Further, it was also found that the most commonly linked sources in these summaries are websites like Wikipedia, YouTube and Reddit. Hence, the ‘injury’ to digital news websites is two-fold.
In addition to such empirical evidence of consumer behaviour, several leading American publishing houses, including the New York Times, Business Insider, HuffPost, and the Washington Post, have reported a sharp decline in website traffic over the past three years. While it is difficult to establish an absolute causal link between this decline and the activities of Gen-AI companies, the timelines align closely enough to make such a connection a plausible explanation. However, whether establishing such a connection truly achieves anything remains an open question. Given that newer technologies, which increase consumer convenience, are always preferred, the mere fact that consumers gravitate toward one technology over another cannot, by itself, constitute a cognizable injury. It is in this context that arguments based on intellectual property conveniently come to the fore, with both foreign and domestic media houses (the present case being an example) filing copyright infringement suits against Gen-AI companies like OpenAI, seeking fair compensation.
Copyright? No Copyright? Chasing Remedies in the Fog
The DNPA’s rationale for alleging copyright infringement rests on the pressing need for digital news companies to secure an alternative revenue stream in light of their inability to control shifting user behaviour. A finding of copyright infringement would create two possible remedial avenues: securing compensation in the specific case at hand, or using the decision to push the legislature towards establishing a statutory licensing mechanism. The two approaches differ significantly in their practicality. In the former, even if copyright infringement were established here, companies would still need to approach the courts afresh for compensation in every subsequent instance, a route that would both be economically burdensome and lead to multiple litigations. By contrast, a statutory licensing framework (previously discussed here) would allow Gen-AI companies to use copyrighted material by paying a predetermined fee, thereby ensuring consistent and predictable compensation without repeated litigation. Yet, both paths hinge on one foundational question, i.e. whether copyright infringement exists in the first place. Without that answer, neither remedy can be meaningfully pursued.
Gen-AI companies most often attempt to sidestep copyright liability by framing their use of protected material, such as news content, as either a non-expressive use or transformative fair use. While the ultimate determination of the validity of these claims rests with the court, both arguments carry considerable weight. For both these counter-arguments, it is pertinent for us to understand how AI learns from the input data. Essentially, it mimics human learning by breaking down the information it receives, identifying patterns within it, and using those patterns to predict responses to user queries. As a result, for AI, the specific expression of an idea matters less than the underlying facts and information. This strengthens the first counter-argument of non-expressive use, since the model relies on the underlying idea to generate its outputs rather than reproducing the creative expression itself (discussed at length here). This also strengthens the second counter-argument, as the copyrighted work is used for an entirely different purpose, namely, training the model, which makes the use transformative (discussed at length here). The production of news summaries or snapshots is incidental since the training was not intended specifically for such outputs, and the information drawn from the news materials can also be used to answer other, unrelated queries.
Given the prevailing uncertainty over which way the Delhi High Court would sway, one must reckon with the possibility that the digital news industry may be denied the remedies it seeks – could that, as DNPA’s counsel describes it, mark the industry’s curtain call?
The End isn’t the End: Where the Industry Could Turn Next
The phenomenon of a new technology rendering older ones obsolete is a truth that keeps repeating, whether in the replacement of fax machines or telephone booths. However, I believe the digital news industry may not answer the curtain call just yet, as commercial, technical and marketing cures exist for its malady.
An instance of a commercial solution is the growing trend of striking private licensing deals with Gen-AI companies, exchanging content access for a fixed stream of alternative revenue without the requirement of statutory intervention (see here, here, here, and here). In addition, many publishers plan to focus on native content for social platforms to maintain traffic to their sites, while also investing in newsletters, mobile apps, and personalised content experiences to foster a direct connection with their audiences. It must be remembered that news generated by Gen-AI companies offers little beyond convenience. This leaves room for the digital news industry to re-strategise and cultivate a personal connection with its audience, one that encourages readers to go the extra mile and make that click.