Korean IP policy for AI-assisted output clarifies the standard of human contribution

Managing IP is part of Legal Benchmarking Limited, 1-2 Paris Gardens, London, SE1 8ND

Copyright © Legal Benchmarking Limited and its affiliated companies 2026

Accessibility | Terms of Use | Privacy Policy | Modern Slavery Statement

Cookies Settings

Korean IP policy for AI-assisted output clarifies the standard of human contribution

Sponsored by

HANOL ClassHankyul logo 2.png
Digital AI logo with a finger pointing towards it, and the South Korean flag
Shutterstock

Min Son of HANOL ClassHankyul IP & Law examines South Korea’s new guidance on AI-assisted inventions, designs, and copyright works, and the human contribution required for protection

As generative AI spreads rapidly across the creative sectors, the South Korean government has been working to establish guidelines clarifying the requirements for intellectual property protection of AI-generated outputs. Korea’s policy stance aims to foster a fair intellectual property environment that provides protection for legitimate creative work utilising AI, while filtering out indiscriminate filings for works generated with no or only limited human contribution.

In June this year, the Ministry of Intellectual Property (MOIP) issued guidelines for patent applications involving AI-assisted inventions, followed by the release of guidelines for AI-assisted design applications in July. Prior to this, the Ministry of Culture, Sports and Tourism released guidelines last year addressing copyright issues throughout the entire process – from the training of generative AI models to the generation of their outputs.

This article outlines the key points of the guidelines regarding the eligibility of AI-assisted output for patent, design, and copyright registrations.

AI-assisted inventions: AI may be used as a tool, but ‘substantial human contribution’ still essential

According to statistics from the MOIP, the number of patent applications filed by individuals across a wide range of advanced, highly complex technical fields surged by 289% during the first five months of 2026 compared with the same period the previous year – rising from 4,954 to 19,263, which suggests a sharp increase in AI-generated inventions.

Under current MOIP practice, an AI cannot be named as an inventor. If the AI is listed on an application as an inventor, the application will be rejected (although the Supreme Court’s final ruling on this interpretation remains pending in the so-called DABUS case).

Furthermore, the MOIP’s guidelines clarify that when AI is used as a tool in the inventive process, a human must make a “substantial contribution” to the process for the invention to be eligible for a patent. If there is no substantial human contribution – such as when a person merely provides simple instructions to an AI and files the resulting output as a patent application – the invention cannot be patented, under Article 33 of the Patent Act. Even if a patent is granted, it may later be invalidated.

In practice, if there is reasonable doubt as to whether such contribution was made, an examiner may issue a notice of preliminary rejection and request supporting evidence. This may include lab notes with the inventor’s signature or a signed inventor confirmation statement.

In addition, the guidelines warn that patent applications relying on unverified, AI-generated virtual experimental results may be rejected for lacking enablement (feasibility). This is particularly relevant in the pharmaceutical and biotechnology sectors, where in silico results alone, without supporting in vitro or in vivo data, may be deemed insufficient. Even in cases where a patent application is inadvertently granted, it may later be invalidated on this basis.

AI-assisted designs: human contribution is required for registration

In professional design practice, AI is being used across various stages – from brainstorming and prototyping to concept sketching. The newly released MOIP guidelines clarify the standards for registering designs created with the aid of AI.

The guidelines provide that AI cannot be considered a design creator in view of Article 3 of the Design Protection Act; therefore, AI cannot be entered as a creator on a design application.

However, an AI-assisted design may be registered if a human has made a “substantive creative contribution” – specifically, if a human has conceived the dominant features that define the overall visual impression of the design and has used AI merely as a tool to bring that concept to fruition. For example, if an AI-generated design is submitted for registration exactly as generated – following the input of existing chair design samples along with a generic prompt such as “create an improved chair” – it would not be recognised as a substantive human contribution. However, if specific prompts concerning, for example, the design’s shape, proportions, composition, or colour are provided, and these inputs influence the design’s overall aesthetic impression, then a substantive human contribution may be recognised.

If there is no substantive contribution, a design application may be rejected or the registration subsequently invalidated. During examination, as with patent applications, if there is any doubt as to the extent of human contribution, the applicant may be required to submit evidential documents to substantiate their involvement.

Furthermore, the guidelines recommend maintaining records of the creative process – such as the intent behind the creation and logs of prompts used – to prepare for any future need to verify the creator’s role.

Copyright registration: human creativity must be involved in AI output

The guidelines establish that AI-assisted works – in which “sufficient human creative contribution” is recognised – may be eligible for copyright protection. While AI-generated output itself cannot be registered, it may be registered as a “compilation work” if a human has made a creative contribution to the selection, arrangement, or organisation of that output.

Human creative contributions may be recognised not only when they are made after generative AI output is produced but also when human creative work precedes the use of generative AI. Whether such contribution has been made is determined by factors such as the creator’s ability to control the manner of expression and the final outcome (controllability), and the extent to which the creator can achieve their intended vision (predictability). The mere act of entering a prompt is unlikely to be recognised as a creative contribution, as it is difficult to acknowledge that it entails such controllability and predictability.

The guidelines also include examples of copyright registration decisions involving AI-assisted works. In one case, the Korea Copyright Commission allowed copyright registration for a generative AI-assisted work created by selecting and arranging AI-generated images containing elements that matched those in the author’s sketch, and modifying specific details to achieve overall harmony. However, in another case, registration was denied for a work in which two AI-generated images were combined into one, and only simple post-processing and retouching were performed.

Key takeaways

For AI-assisted works to be eligible for patent, design, or copyright protection, the applicable requirements for human contribution must be met. Accordingly, applicants may need to demonstrate sufficient human involvement in the AI-assisted process both during examination and even after intellectual property rights have been secured.

In light of these requirements, applicants should ensure they establish or refine their internal AI usage policies to align with the principles set out in the guidelines. Furthermore, systematic monitoring and record-keeping throughout the creative process are strongly recommended. This may include:

  • Retaining relevant prompts;

  • Documenting human modifications and selections before and after using the AI; and

  • Keeping records of the sources and results of technical verification and experiments.

Particular care should also be taken when relying on AI-generated in silico predictions in patent applications, as these may raise issues regarding sufficiency and enablement.

more from across site and SHARED ros bottom lb

More from across our site

As clients demand more flexible fees and AI-driven efficiencies, IP firms are rethinking pricing without abandoning the value of their work
HG Law was formed largely by ex Fish & Neave lawyers who left after the latter's merger with Ropes & Gray; now those experiences are shaping its latest combination
David Rodrigues and Evan Lazerowitz – attorneys at Robinson & Cole – offer practical guidance on the steps a business might consider to preserve and maximise IP value before entering financial distress or insolvency
Daniel Law CEO Alicia Daniel-Shores may have stepped into a family business, but she has spent two decades building her own vision around technology and litigation
Law firms don't particularly like saying no to clients. Perhaps they should get better at it
The office, which has a relatively new patent litigation team as well as a wider ‘soft’ IP offering, would remain independent if the rejection is rubberstamped
The IP boutique is betting that growing investment in quantum and semiconductors will fuel demand as technology shifts toward physical systems
Lawyers say the overhaul better reflects commercial reality, with changes spanning novelty searches, digital designs and access to a small claims track, but stopping short of criminalising unregistered design infringement
Carpmaels & Ransford hires its fourth partner of the year and wins access to UPC files, as fresh orders address costs, confidentiality and the use of divisional patents
Peter Ling of Lenz & Staehelin in Zurich discusses trends in patent litigation, the regulation of AI and the benefits of teaching IP
Gift this article