I Wonder …
What if our “belief” in something turns into our “faith” in it? For the last few months, I have been wondering if our belief in “fair dealing” (or broadly, “limitations and exceptions”) has silently slipped into our “faith” in it – a faith that demands complete surrender to it while blinding us to the harm it covertly causes to the public domain. The calls for revisiting copyright in the wake of generative AI – with proposals to create new exceptions or design remuneration models for authors – further fortify this “faith,” I believe.
In this post, I argue why this transition is problematic. The overreliance on fair dealing, I argue, inadvertently expands the scope of copyright law and harms the public domain. Note: Fair dealing and fair use are construed to be conceptually different, the former being limited to an exhaustive list and the latter being open. However, recent research shows the fading boundaries of these differences. While I present my case from a fair dealing perspective, the kernel of the argument (with some contextual nuance, of course) can be applied to fair use.
What Fuels Faith in the First Place?
The reasons for our faith in fair dealing may vary between litigation and policy spheres, considering the intricate distinctions between adjudication and legislation. There are, however, some “seemingly valid” (or “practical”) reasons that support our reliance on fair dealing on both fronts. For one, it helps one escape from the clutches of copyright and avoid any liability. Speaking of litigation, especially, if the defendant is a small entity standing against a tech giant, invoking fair dealing makes more (economic) sense than thinking of the larger policy issues of the public domain, diversity, equality, etc.
Perhaps this is what makes it part of a litigation strategy: when other infringement arguments falter, fair dealing steps into the spotlight. This way, litigation strategies seem to follow a bi-modal approach in copyright infringement cases: denial mode, and defense mode. In the former, efforts are directed to proving the non-infringement. For example, the defendant could contest the originality of the plaintiff’s work or argue that their work is dissimilar to the plaintiff’s work. Then comes the defense mode where the focus shifts to defending the defendants from any harm caused by copyright. Essentially, this defense mode acknowledges that the defendant’s use is prima facie infringement. Here, fair dealing’s “more concrete” presence in the copyright statute can help lawyers frame their arguments with more definite values (like research and freedom of speech) and tether them to the statute’s language.
To some extent, the said “faith” also flows from the commonly celebrated desire to “balance” the copyright system – an urge that entices us to “tinker” with the status quo, grounded in the oft-claimed access-incentive paradigm. So, the seemingly distorted scales of balance can be brought back to their desirable state, depending upon which side of the balance one associates with. And fair dealing nicely fits in one side of the scale, opposing copyright claims. Doesn’t it?
The upshot is that fair dealing is ultimately a defense. It rescues us when the (f)act of infringement is prima facie proven. Put simply, if one invokes an exception, one accepts that one has infringed the copyrighted work. Reliance can be placed on Madras High Court’s E.M. Forster And Anr. v A.N. Parasuram which noted that “Fair Dealing” arises only if substantial reproduction infringement is proven. Otherwise, such a claim does not arise.
Faith-based Fair Dealing Expands Copyright Law
This bi-modal litigation strategy, as an economic activity, may make some sense, especially in David-versus-Goliath-type scenarios, like a small entity battling a tech giant. However, flowing with such faith and making policy proposals for new fair dealing exceptions might be a step too far. It’s simple, if there is no infringement, why invoke fair dealing or create any exceptions?
Let me exemplify this with a semi-hypothetical (as it is partially real) example:
Bubu made audiobook summaries and got sued for the infringement of Plaintiff’s abridgment right in the literary work. Bubu claimed fair dealing (e.g. that the use was for review purposes). Supposedly, Bubu succeeds. This is a fair dealing “victory” and “loss” for copyright holders. Right? But no, it only appears so. Why? It confirmed that “audio summaries” are prima facie “abridgment.” However, was that ever the case? The current definition of “abridgment,” appositely, suggests that it happens from literary to literary work, e.g. a book is abridged into another literary form like a pamphlet, etc. (Read Section 2(iii) of the Indian Copyright Act). But since the fair dealing claim succeeded, the abridgment definition inadvertently gets expanded (which means the scope of rights on literary works broadens) to “audio” summaries.
Now, if someone decides to use this “fair dealing success” case (and other such cases) to argue the expanded scope of rights, theoretically, courts should not reject the claims. At least in Indian law, precedents possess legal lode and can shape future judgments. As Arthur L. Goodhart noted: “The logic of the argument, the analysis of prior cases, the statement of historical background may all be demonstrably incorrect in a judgment but the case remains a precedent nonetheless.”
Beware of TDM (or any new) Exceptions!
So, if one asks for a TDM exception or any other exception for training of generative AI, one accepts that TDM/training AI is per se infringing. It translates to – without an exception, I would be liable. However, the issue is that creating a TDM exception blurs the boundaries between expressive and non-expressive uses of works. Theoretically and historically, copyright has only been concerned with the expressive use of works. As Prof. Severine Dusollier argues, the exploitation/communication of work by the public is what copyright should be concerned with.
Faith in fair dealing also keeps us clinging to the external limit of copyright, ignoring that copyright has both external and internal limits placed upon it. Fair dealing/use is an external limit, specifically recognized in the statutes of certain laws (similarly for fair use and limitations and exceptions for other laws). However, the system has internal limitations like the idea-expression dichotomy (or the Spillover principle), scenes-a-faire, substantial similarity, de-minimis copying, etc. Focusing on fair dealing too much, the internal limits may become invisibilized. Or, as Prof. Ruth L. Okediji rightly remarked “[c]onceiving L&Es as a tool to achieve copyright goals reduces the pressure to design copyright law to serve large-scale socially beneficial outcomes. It allows copyright protection to grow unhindered because it assumes that whenever there is an imbalance, some L&E will fix it”.
Once a new exception is created, all those arguments for TDM/Training AI being non-infringing will be nullified. (see, e.g., here, here, here, here, and here.) Moreover, if this “new-technology/new-exception” urge gains international traction and becomes concretized, there is no retreat! Rather, it will pressure countries with weak bargaining power to make such an exception if they wish to legalize the use of TDM in their countries.
With that, I conclude this piece with a strong recommendation to read Prof. Oren Bracha’s piece called The Work of Copyright in the Age of Machine Production which rightly reminds us that “Instead of attempting to blindly apply rules to new phenomena, we need to ask what the basic principles behind the rules are, what purpose those principles are designed to serve, and whether these principles and the rules derived from them still serve that purpose in the new context or require reconsideration.”
Thanks to Shivam Kaushik for his input on this piece.