Generative AI now significantly replicates the superficial appearance of patent infringement assessments. When hypothetical factual backgrounds are inputted into generative AI models such as ChatGPT, Gemini, and Claude, they derive conclusions by combining existing infringement theories and judicial precedents. Although it outwardly appears as though an "assessment" has been conducted, whether this can truly be deemed a "legal judgment" requires separate and careful scrutiny.
The utilization of AI is rapidly expanding within recent patent practice as well. AI substantially enhances productivity in tasks such as searching and analyzing vast volumes of prior art documents, dissecting claim limitations, and establishing correspondence between claim elements. As an analytical tool, AI is already becoming entrenched in practice, and the utilization itself can be evaluated as a natural progression.
The issue lies in the boundary between the tool and the subject of judgment. A distinction needs to be made between instances where an individual utilizes AI as a supplementary aid and instances where AI-generated infringement assessments are provided in the form of a service without independent verification. Particularly in the latter case, a review of the mechanism of judgment formation and the attribution of liability is essential.
A patent infringement assessment can hardly be reduced to simple logical operations. It is a complex process encompassing claim construction, the application of the Doctrine of Equivalents, the evaluation of differences from prior art, and the selection and organization of facts. The standards of "identity of the principle of problem-solving" and "substantial identity" presented by the Supreme Court in its determination of equivalence also demand interpretation and evaluation that go beyond literal comparison. This structure of judgment inherently presupposes a responsible subject.
Such issues may become more pronounced when AI-based analysis is provided in the form of a service. This boundary is distinctly revealed especially when such analyses are conducted in areas like patent trend surveys or patent mapping. These tasks are originally understood as information analysis work supporting research and development (R&D) decision-making, and in practice, they are not infrequently performed by specialized investigation and analysis firms rather than patent attorneys.
However, if a report includes an evaluation of the possibility of infringement alongside the applicability of a specific technology, its character may transcend the scope of ordinary information provision. This is because the moment a technology is contrasted against prior art on the premise of product application, it substantially performs a function analogous to an infringement assessment.
What becomes problematic at this juncture is the relationship with relevant legislation. The Certified Patent Attorney Act restricts non-patent attorneys from performing services such as agency or expert opinion (appraisal) regarding patents. In light of this, depending on the manner in which AI-based analysis results are provided, legal issues similar to the performance of patent attorney services by unlicensed entities may potentially arise. This should be understood not to definitively conclude the illegality of individual services, but as a structural issue requiring examination under the current system.
The attribution of liability is also a critical point of contention. If an error occurs after the relevant assessment has been utilized for business decision-making, the subject to whom liability is attributed may be ambiguous. AI cannot be a subject of liability, and the scope of liability may remain unclear between the service provider and the user. A structure where the generation of judgment and the attribution of liability are separated becomes a potential cause of future disputes.
If a company incurs billions of won in damages or faces delays in product launches due to an error in an infringement assessment made by AI, will the company providing the AI-based analysis service bear the responsibility? Or will the corporate representative who made decisions based on that assessment bear the responsibility?
It is a natural phenomenon for the pace of technological development to outpace institutions. However, in domains accompanied by legal judgment, responsibility and trust must take precedence over speed. This aligns with the context of why unauthorized legal acts have been continuously regulated.
There is no need to restrict the utilization of AI itself. However, its role and limitations need to be more clearly established. While AI is useful as a tool to assist analysis, additional discussions regarding institutional and liability structures must precede its functioning as a subject that substitutes for legal judgment.
Ultimately, the core of the issue is not the technology, but the role. How to interpret and utilize the results provided by AI, and how to establish the corresponding liability, still belongs to the human domain. Unless this basic premise is maintained, technological advancement may operate not as a convenience, but as a risk.
Source: AI Patent Infringement Assessment: How Far Can It Be Permitted? < Legal Forum < Contribution < Opinion < Article Body - The Law Times (Beobryul Shinmun)