By reserving the term “intellectual contribution” exclusively for humans, current journal policies create a conceptual paradox regarding the actual utility of generative AI in drafting manuscripts. In July 2026, the European Journal of International Law established a formal framework designed to govern the use of these computational tools within legal scholarship. This move by editors Sarah Nouwen and Michael Waibel mirrored a broader movement across the academic publishing landscape, aligning the journal with the standards of other prominent publications that emphasize transparency and human authorship. However, the specific implementation of the human-AI “outweighing” approach introduced a unique regulatory hurdle that demands a human author’s contribution clearly surpass any machine involvement. While the intent was to protect the sanctity of original legal thought, the policy has sparked intense debate regarding its underlying logic and practical feasibility. Critics argue that the framework rests on shaky conceptual foundations that do not reflect the technological reality of 2026. This policy intervention arrives at a time when the distinction between a researcher’s independent thought and their interaction with digital assistants is becoming increasingly difficult to quantify or define with legal precision.
The Logical Inconsistency: Comparing Humans to Machines
A significant point of contention within the current framework is the reliance on a direct analogy between generative artificial intelligence and human research assistants. The policy posits that if an AI tool performs a task that would traditionally warrant a co-authorship credit if completed by a human, the outweighing standard is effectively breached. This comparison is fundamentally fragile because the boundaries between an assistant’s acknowledgment and a co-author’s credit have remained notoriously subjective, even in purely human settings, for decades. Decisions regarding authorship frequently involve normative and subjective judgments about the substance of a contribution, which are nearly impossible for a third-party editor to reconstruct retroactively. By using a blurry standard from the human world to regulate a precise technological output, the journal introduces an element of unpredictability into the submission process. This uncertainty leaves researchers without a clear roadmap for how much machine assistance is considered too much, potentially discouraging the use of sophisticated tools that could otherwise improve the clarity and depth of international legal analysis.
The conceptual paradox is further deepened by the journal’s specific definition of what constitutes intellectual work. By consistently reserving the term “intellectual contribution” for human actors while simultaneously denying that machine outputs can be intellectual, the policy creates a logical trap that is difficult to escape. If a machine output is inherently non-intellectual, it remains theoretically unclear how such an output could ever truly replace human intelligence in a scholarly context. The policy treats machine work as both significant enough to outweigh human effort and yet fundamentally devoid of the very intellectual quality that defines scholarship. This dual treatment suggests that the framework is more concerned with the mechanics of text production than with the actual cognitive processes driving the research. If the AI contribution is viewed as a mere mechanical byproduct, then the requirement for human effort to outweigh it becomes a quantitative measure that ignores the qualitative framing and intent that only a human author can provide.
Furthermore, a distinct disconnect exists between the core principle of non-replacement and the specific quantitative standard applied by the editors. A human researcher might provide only 30 percent of the literal text in a manuscript while the generative tool produces 70 percent, yet this statistical distribution does not automatically imply that human intelligence was replaced. The 70 percent of text could be a direct result of thousands of specific, iterative human prompts that directed every nuance of the argument. Conversely, a human could contribute 90 percent of the text but follow a structure entirely suggested by a machine, which would constitute a much deeper form of intellectual replacement without triggering the outweighing test based on volume. By conflating the threshold for authorship with the relative weight of visible contributions, the policy lacks a clear explanation for why certain levels of technical assistance are deemed impermissible while others are celebrated.
Intertwined Intelligence: The Reality of Modern Co-creation
To fully understand the limitations of current journal policies, one must examine how legal scholars actually engage with generative tools in 2026. The relationship is not a linear path of copy-pasting but rather a dynamic interplay or a dialectic process that occurs in real time. In this iterative cycle, the machine provides exploratory possibilities and structural suggestions, while the human author provides the necessary framing, intent, and critical selection. These processes are so tightly interwoven through constant brainstorming and refinement that it is practically impossible to separate and weigh the human elements against the machine elements. When a scholar uses a tool to test the robustness of a legal argument or to find counter-arguments in foreign case law, the resulting text is a hybrid product of a shared cognitive environment. Attempting to disentangle these threads for the purpose of a disclosure form often requires an arbitrary division of labor that does not reflect the fluid reality of the creative process.
The intensity of interaction with these tools also differs qualitatively from traditional human-to-human collaboration. Researchers can engage in repetitive, blunt, or exceptionally high-frequency dialogue with a digital assistant that would be socially unacceptable and physically impossible with a human research assistant. This allows for a level of research intensity and divergent thinking that traditional methods simply cannot match. The labor involved in prompting, refining, and fighting with an AI to develop a coherent legal framework is itself a significant form of intellectual work that requires deep expertise. As noted by experts such as Jessica Tillipman, some legal frameworks require months of persistent interaction with digital models to ensure accuracy and limit bias. Consequently, the outweighing test fails to recognize that managing and directing a sophisticated tool often requires more human cognitive effort than the traditional drafting process, making the weight of the contribution a poor proxy for the originality of the thought.
Rethinking Authorial Credit: From Generation to Accountability
As the academic community critiques restrictive weighing tests, a model of absolute authorial accountability is emerging as a more robust alternative. This principle suggests that human authors should be held fully and exclusively responsible for every word, citation, and claim within their work, regardless of the technological tools employed during the drafting phase. This approach shifts the editorial focus away from the hidden mechanics of text generation and toward the verifiable quality and accuracy of the final scholarly product. It recognizes that the true value of scholarship lies in the purposeful and meaningful communication curated by a human author who stands behind the argument and accepts the professional consequences of its publication. By emphasizing accountability over the process of creation, journals can maintain high standards of integrity without needing to police the private research habits of their contributors.
Navigating the limitations, hallucinations, and inherent biases of generative models is rapidly becoming a core competency of modern human intelligence in 2026. Skillful use of these tools is not a shortcut to thinking but rather a new form of labor-intensive framing and selection that requires a high degree of domain expertise. When an author successfully extracts a meaningful insight from a complex dataset or a linguistic model, the reduction of friction in the writing process should not be mistaken for outsourced thinking. An accountability-based model effectively solves the idea vacuum created by policies that might suppress valuable arguments simply because they were initially sparked by a machine suggestion. If a scholar adopts, verifies, and expands upon a machine-generated idea, that idea becomes part of their intellectual contribution through the act of professional endorsement and rigorous verification.
The Disclosure Paradox: Stigma and Practical Impossibility
The mandatory disclosure requirement included in the EJIL policy introduces what many researchers describe as the transparency dilemma. Recent studies in 2026 indicate that individuals who openly disclose their use of generative tools are often trusted less and perceived as less creative by their peers, regardless of the objective quality of their research. This persistent stigma can bias peer reviewers and editors against a manuscript from the moment of submission, creating an uneven playing field. By forcing authors to disclose machine assistance, the policy effectively encourages them to voluntarily lower the perceived value and authenticity of their own work. This creates a psychological barrier to entry for honest researchers while doing very little to deter those who are willing to hide their use of technology to gain a competitive advantage in the publishing market.
Enforcing these disclosure rules also presents a series of practical impossibilities that could overwhelm editorial boards. Detecting sophisticated, human-tuned text generated by machines is notoriously difficult, as researchers can easily adjust the tone and structure to bypass automated detection systems. More importantly, machine-generated ideas leave no digital footprint if they are eventually expressed in text written entirely by a human hand. This creates a perverse incentive structure where honesty pays less than dishonesty; ethical authors who provide detailed logs of their prompts face potential stigma and scrutiny, while those who hide their interactions with technology go unpunished. This detection arms race undermines the very integrity and trust that journals seek to protect, as the rules are essentially unenforceable against those determined to circumvent them.
The long-term risks for early-career researchers and doctoral students are particularly concerning under these mandatory disclosure frameworks. Because academic careers are built on a foundation of reputation and perceived originality, a public disclosure of machine assistance could follow a scholar for decades. If institutional norms or tenure requirements become even more restrictive between now and 2030, a past disclosure could be used to devalue a researcher’s entire body of work during crucial career milestones. Requiring authors to provide technical details like prompt logs or verification measures is a burdensome solution that offers the appearance of oversight without any meaningful ability to verify the integrity of the scholarly process. For many young scholars, the risk of being labeled as an AI-dependent researcher outweighs the benefits of transparency, leading to a culture of silence and hidden usage.
Adaptive Frameworks: Navigating the New Scholarly Landscape
The integration of generative tools into the workflow of international law scholars represented an unavoidable technological evolution that mirrors the earlier adoption of digital databases and personal computers. Rather than attempting to regulate the minute details of how a manuscript was produced, academic institutions and journals benefited from reconfiguring their review architectures to meet the challenges of a data-rich era. This transition encouraged the incorporation of digital tools into the peer-review process itself, allowing editors to handle an increased volume of submissions while focusing more heavily on the validity of the data and the logical coherence of the legal arguments. By treating technology as a standard part of the research toolkit, the academic community moved toward a more inclusive and technologically literate definition of what it meant to produce original scholarship.
The development of these policies highlighted the fundamental tension between traditional values and the practical necessity of embracing revolutionary technology. The academic community found that the most sustainable path forward involved treating these tools as an unstated default in the research process, much like the use of spellcheck or legal search engines. By shifting the emphasis toward absolute authorial accountability, journals successfully maintained their role as gatekeepers of quality while remaining open to diverse styles of investigation. This approach ensured that the focus of legal scholarship remained on the quality of the ideas and the strength of the legal reasoning. Ultimately, the successful navigation of this transition required acknowledging that human and machine intelligences became permanently intertwined in the pursuit of knowledge, making the transparency of the tool less important than the integrity of the scholar.