← 글 목록← All articles

Intelligent robot inventions in Korea: which examination guidelines, accelerated examination and examining units apply?

Brief summary

  • Examination standard: The Examination Practice Guidelines by Technology Field (기술분야별 심사실무가이드) of the Ministry of Intellectual Property (MOIP, 지식재산처) contain Chapter VI, “Intelligent Robot Field” (제6부 지능형로봇 분야). Chapter VI applies to inventions in which there is an organic combination of AI technology and a robot. It contains the description requirements, the method of assessing inventive step, the method of assessing medical practice and 19 cases on issues.
  • Accelerated examination: Applications related to advanced robotics technology are designated as eligible for accelerated examination (우선심사; expedited examination) by MOIP Announcement No. 2025-18. Eligible applications are those for which accelerated examination is requested from 1 November 2025 to 31 October 2026. Both of two requirements must be met. An advanced robotics-related patent classification (CPC) must be assigned as the main classification. The application must be an application by a company that manufactures, or is preparing to manufacture, advanced robotics-related products, devices and the like in Korea, or an application for the results of a national research and development program related to advanced robotics technology.
  • Dedicated organization: The examination of patent applications in the intelligent robot field is under the Digital Convergence Examination Bureau (디지털융합심사국) of MOIP, and that Bureau has the Intelligent Robot Examination Division (지능형로봇심사과).
  • Provisions that exist only in this field: The main text of the Patent Act (특허법) does not contain the word “로봇” (robot). No provision of the Patent Act that applies only to robot inventions was found.

The statute is the Patent Act in force from 11 November 2025. The examination practice follows Chapter VI of the Examination Practice Guidelines by Technology Field (March 2026). Last verified on 10 October 2026.

Last verified

The intelligent robot field at a glance

As of October 2026.

Item Content Basis
Examination standard Chapter VI, “Intelligent Robot Field”, of the Examination Practice Guidelines by Technology Field (“the Practice Guidelines”). The opening note states establishment in December 2021 and revision in May 2023. The division in charge of revision and management is the Intelligent Robot Examination Division Opening note of Chapter VI of the Practice Guidelines [2]
Scope Intelligent robot inventions in which there is an organic combination of AI technology and a robot. If the feature lies only in AI technology, Chapter I (Artificial Intelligence Field) applies. If the feature lies in control technology and component technology, the general Examination Guidelines apply Practice Guidelines 1.2 [2]
Description of the invention The assessment standard for the enablement requirement and five items of the assessment method Practice Guidelines 2.1 [2]; Patent Act, Art. 42(3), subparagraph 1 [1]; Examination Guidelines, Part 2, Chapter 3 [3]
Claims What must be clearly stated in a method invention and in a product invention Practice Guidelines 2.2.1, 2.2.2 [2]; Patent Act, Art. 42(4) [1]; Examination Guidelines, Part 2, Chapter 4 [3]
Medical practice The claims of an invention relating to a robot used for the surgery, therapy or diagnosis of humans are checked in three steps Practice Guidelines 2.2.3 [2]; main clause of the Patent Act, Art. 29(1) [1]; Examination Guidelines, Part 3, Chapter 1, 5.1 [3]
Inventive step The order of assessment and four items of the specific method Practice Guidelines 3.2 [2]; Patent Act, Art. 29(2) [1]; Examination Guidelines, Part 3, Chapter 3 [3]
Cases on issues 19 cases. Seven on the enablement requirement, 11 on inventive step and one on medical practice Practice Guidelines 4.1 to 4.19 [2]
Designation for accelerated examination A patent application or utility model application that is directly related to advanced robotics technology and that meets both the main classification requirement and the requirement of manufacturing in Korea or a national research and development program. Applications for which accelerated examination is requested from 1 November 2025 to 31 October 2026 Patent Act, Art. 61, subparagraph 2 [1]; Enforcement Decree, Art. 9(1), subparagraph 2-3 [4]; Public Notice, Art. 4(1), subparagraph 2, item (r) [5]; Announcement No. 2025-18 [6]
Fourth Industrial Revolution classification A patent application that utilizes intelligent robot technology and to which MOIP has assigned the new patent classification Z03R Enforcement Decree, Art. 9(1), subparagraph 2-2 [4]; Public Notice, Art. 4(1), subparagraph 2, item (p), attached Table 3 [5]
Dedicated organization Intelligent Robot Examination Division of the Digital Convergence Examination Bureau Organization Decree, Art. 17; Enforcement Rule of the Organization Decree, Art. 14 [7]

The only statutory provision that Chapter VI of the Practice Guidelines states is the main clause of Art. 29(1) of the Patent Act. The other Patent Act provisions in the table are the provisions that the chapters of the Patent Examination Guidelines (특허·실용신안 심사기준, “Examination Guidelines”), which the Practice Guidelines state are to be followed, give as “관련 규정” (relevant provisions). Part 2, Chapter 3 of the Examination Guidelines (description of the invention) gives Art. 42(3), Part 2, Chapter 4 (claims) gives Art. 42(4), and Part 3, Chapter 3 (inventive step) gives Art. 29. [1][2][3]

Examination standard: Chapter VI, “Intelligent Robot Field”, of the Practice Guidelines

The main text of the Patent Act does not contain the words “로봇” (robot) and “인공지능” (artificial intelligence). The amended Patent Act that will enter into force on 16 March 2027 (Act No. 21955) does not contain the two words either. Intelligent robot inventions are also examined under the patentability requirements (Art. 29) and the description requirements for the specification (Art. 42(3) and (4)). A violation of either provision is a ground for rejection (Art. 62, subparagraphs 1 and 4). [1]

The examination practice in the intelligent robot field is in Chapter VI of the Examination Practice Guidelines by Technology Field. Chapter VI states that it explains the examination practice guidelines that apply to applications for inventions in the intelligent robot field. The Practice Guidelines call these inventions “intelligent robot inventions” (지능형로봇 발명). [2]

The date markings of Chapter VI differ from place to place. The opening note states “제정 2021.12.” (established December 2021) and “개정 2023. 5.” (revised May 2023). The history at the end of the book gives January 2022 for establishment and May 2023 for revision. At the foot of the pages, pages marked “2023년 5월 추록” (supplement of May 2023) are mixed with pages that have no marking. The MOIP website posts these Practice Guidelines as the March 2026 edition. [2]

Chapter VI has no Addenda or transitional measure that sets from which applications it applies. The foreword of the Practice Guidelines was written in December 2020 in the name of the Commissioner of the Korean Intellectual Property Office (특허청장). The foreword states two roles that it hopes the Practice Guidelines will play. Internally, the role is accurate and consistent examination work. Externally, it is a guide that helps start-up and venture business applicants who file in new technology fields. [2]

Chapter VI does not contain a single case number of the Supreme Court of Korea or the Intellectual Property High Court of Korea (Patent Court of Korea). Examples of the assessment are presented as the cases on issues (쟁점 사례) in section 4. [2]

Scope: inventions in which AI technology and a robot are organically combined

Section 1.1 of the Practice Guidelines explains an intelligent robot by three functions. An intelligent robot is a robot that perceives the external environment (Perception), judges the situation by itself (Cognition) and operates autonomously (Mobility & Manipulation). The Practice Guidelines state that it differs from existing robots, which repeatedly perform a fixed task as programmed. The Practice Guidelines do not state a statute or document that is the source of this explanation. [2]

Section 1.2 of the Practice Guidelines divides the element technologies of an intelligent robot into three. They are AI technology, control technology and component technology. The Practice Guidelines state that an intelligent robot invention means an invention whose feature lies in AI technology. [2]

Which standard applies depends on where the feature of the invention lies (Practice Guidelines 1.2). [2]

Feature of the invention Standard that applies
A general robot invention whose feature lies in control technology and component technology General Examination Guidelines
The case where the feature lies only in AI technology Practice Guidelines for AI-related inventions (Chapter I)
An intelligent robot invention in which there is an organic combination of AI technology and a robot Practice Guidelines for intelligent robot inventions (Chapter VI)
Which standard applies to an invention relating to a robot?

For that reason, not every invention relating to a robot is covered by Chapter VI. The Practice Guidelines call an invention whose feature lies in control technology and component technology a “general robot invention” (일반적인 로봇 발명), and state that the general Examination Guidelines apply to it. The Practice Guidelines do not give examples of the three cases. [2]

Chapter VI is organized as follows. [2]

Place What it covers
1. Overview Explanation of the technology (1.1), technical features (1.2)
2.1 Requirements for the description of the invention Assessment standard for the enablement requirement (2.1.1), assessment method (2.1.2)
2.2 Requirements for the claims Assessment standard for the description requirements (2.2.1), assessment method (2.2.2), method of assessing medical practice (2.2.3)
3. Assessment of patentability requirements Assessment of patentability requirements in general (3.1), method of assessing inventive step of intelligent robot inventions (3.2)
4. Cases on issues 19 cases (4.1 to 4.19)

Relationship with the general Examination Guidelines and the Practice Guidelines for the AI field

Chapter VI states, for each requirement, the standard to be followed. The basic standard is the Patent Examination Guidelines. An invention whose feature is only the AI of a robot follows the Practice Guidelines for the AI field. [2]

Requirement Standard followed as the basis An invention whose feature is only the artificial intelligence (AI) of the robot, without an organic combination with the hardware of the intelligent robot
Description of the invention (2.1.1) Examination Guidelines, Part 2, Chapter 3 “2.1 발명의 설명” (2.1 Description of the invention) of the Practice Guidelines for AI-related inventions
Claims (2.2.1) Examination Guidelines, Part 2, Chapter 4 “2.2 청구범위” (2.2 Claims) of the same Practice Guidelines
Patentability requirements (3.1) Examination Guidelines, Part 3, Chapter 1 (industrial applicability), Chapter 2 (novelty) and Chapter 3 (inventive step) “3.특허요건” (3. Patentability requirements) of the same Practice Guidelines

Section 3.1 of the Practice Guidelines states that the assessment of industrial applicability and novelty of an intelligent robot invention is the same as the existing assessment method. Section 3 therefore deals only with the method of assessing inventive step. [2]

Chapter VI has no section that deals with subject matter eligibility (발명의 성립성). “3. 특허요건” (3. Patentability requirements) of the Practice Guidelines for the AI field (Chapter I) contains “3.1 발명의 성립요건” (3.1 Subject matter eligibility requirements) and “3.2 신규성, 진보성” (3.2 Novelty, inventive step). Section 3.1.1 of Chapter I states that the standard for assessing subject matter eligibility of AI-related inventions is, in principle, the same as the standard for assessing subject matter eligibility of computer- and software-related inventions. [2]

The current version of the Patent Examination Guidelines is MOIP Established Rule No. 7. It has been in force from 13 March 2026. [3]

The content of the Practice Guidelines for the AI field is covered in “How are AI inventions examined in Korea?“. Subject matter eligibility of software inventions is covered in “Is software patentable in Korea?“, and program claims are covered in “How are computer program claims written in Korea?”.

Description of the invention: the specific content of the AI technology

Section 2.1.1 of the Practice Guidelines states the condition for an intelligent robot invention to be easily practiced. The specific content of the AI technology that the invention implements must be stated in the description of the invention. The Practice Guidelines give four things as that content. They are the kind of training data, the correlation between the input data and the output data for training, the kind of learning model, and the method of using the output data derived from the learning model. [2]

Section 2.1.1 has a “다만” (however) clause attached. It concerns the case where the specific matters are not stated. If they can be clearly understood by a person skilled in the art in view of the common general technical knowledge at the time of filing, it is not found for that reason that the invention cannot be easily practiced. [2]

Section 2.1.2 of the Practice Guidelines states the assessment method in five items. [2]

Item Requirement “다만” (however) clause in the Practice Guidelines
(1) The means for training, such as the definition and kind of the training data and the kind and structure of the learning model, must be specifically stated. The standard is the extent to which the invention can be accurately understood and reproduced without excessive trial and error or the addition of special knowledge None
(2) The definition of the training data, the form of the input data and the form of the output data must be stated clearly and in detail If it is difficult to clearly define the training data, an embodiment that states an example of the training data must be stated
(3) The structure of the input layer, the intermediate layer and the output layer must be stated specifically and in detail according to the kind of learning model This does not apply if the structure of the learning model is common general technical knowledge that is self-evident to a person skilled in the art, or can be understood and reproduced from an embodiment
(4) If the correlation between the input data and the output data of the trained model is not specifically stated, the enablement requirement is regarded as not met If a person skilled in the art can infer the correlation through an embodiment, or it is found that the correlation is formed by the learning model, the requirement can be regarded as met
(5) If the measured values of a sensor and the input data of the learning model do not correspond to each other, or the preprocessing process that converts the measured values into input data is not specifically disclosed, the enablement requirement cannot be regarded as met This does not apply if that relationship and the preprocessing process can be inferred as self-evident from the description of the invention

Item (2) has three examples of input data and output data. For AI that reads welding defects from the shape of a weld, they are an image of the weld shape and the kind of welding defect. For path search of an autonomous driving robot, they are distance data to obstacles, a map and GPS data, and the driving path of the robot. For a robot that responds to a user’s utterance, they are the user’s voice data and the robot’s utterance sentence. [2]

Item (4) explains the case where the correlation is specifically stated as the case where four things are present. The training data is specified. Between the characteristics of the training data, there is a correlation for solving the technical problem. The learning model or the training method is specifically stated. A trained model is generated by that training data and that training method. [2]

Claims: the entity that performs the step and the organic combination

Section 2.2.1 of the Practice Guidelines states that three things must be stated clearly and concisely in the claims of an intelligent robot invention. They are the functions of the elements, the organic combination between the elements, and the category of the invention. [2]

Section 2.2.2 of the Practice Guidelines states them separately by category. [2]

Category What must be clearly stated
Method invention Elements such as the training data and the structure of the learning model. The entity that performs the training step (the hardware)
Product invention The training data and the structure of the learning model. The organic combination of the learning model and the control configuration of the robot

The sentence of the Practice Guidelines on method inventions concerns the case where an intelligent robot invention includes an AI training step. The Practice Guidelines state that the training step can be expressed as a series of processes or operations connected in time series, that is, as steps. If those steps are specified, the invention can be stated in a claim as a method invention. For product inventions, the Practice Guidelines give as the reason that an intelligent robot invention consists of an organic combination of an AI learning model and a hardware configuration such as the robot body. [2]

The general standard for claim formats is covered in “Which claim formats does Korea accept?“.

Robots used for surgery, therapy or diagnosis: the method of assessing medical practice

Section 2.2.3 of the Practice Guidelines deals separately with inventions relating to robots used for the surgery, therapy or diagnosis of humans. If the claims include medical practice (의료행위), the invention falls under an invention that is not industrially applicable, and the ground for rejection under the main clause of Art. 29(1) of the Patent Act can apply. The basic assessment standard follows Part 3, Chapter 1 of the Examination Guidelines. [2]

The specific assessment method that the Practice Guidelines state has three steps. [2]

  1. Check whether the claim ends with “방법” (method). For a claim that ends with “물건” (product), assessment of medical practice is unnecessary.
  2. Check whether medical practice is performed in each step of the claim. In the examples of the Practice Guidelines, “수술 도구를 조작하는 단계” (a step of manipulating a surgical tool) includes an actual act of surgery and is medical practice. “수술 도구를 이동시키는 제어부” (a control unit that moves a surgical tool) states only a configuration without an act of surgery and is not medical practice.
  3. Check whether each step of the claim restricts the medical practice of a medical practitioner, and then finally assess whether it is medical practice.

The Practice Guidelines state in one sentence what is important in the assessment. It is whether a restriction on the medical practice of a medical practitioner arises if the right holder asserts patent infringement according to the statement of the claim. [2]

There is also a sentence on the performing entity. It concerns the case where it is unclear whether the entity that performs each step of the claim is a medical practitioner or a medical device. The Practice Guidelines state that it is necessary to notify, together, a ground for rejection to the effect that the performing entity is unclear, so that the claims are made clear. [2]

Case 19 is an example of this assessment method. Its heading is “A case found not to be medical practice because, as a method of operating a medical device, it does not substantially include an act of surgery”. The claim was a method for operating a surgical navigation system. According to the description of the invention, the navigation module is used attached to a surgical tool and a bone. The Practice Guidelines found that the attaching step, which could fall under medical practice, is excluded from the claim and that only the information processing of the system is stated. The Practice Guidelines state that each step of the claim cannot be regarded as a case that has a direct and non-temporary effect on the human body, and does not include an act of surgery that interacts with the human body. They therefore found it to be a method of operating a medical device that does not include substantial medical practice. [2]

The words used in the assessment of Case 19 are in the general Examination Guidelines. Part 3, Chapter 1, 5.1 (1) ③ of the Examination Guidelines does not accept as industrially applicable an invention of a process that includes medical practice as at least one step or as an inseparable element. Section 5.1 (2) ② of the same chapter treats an invention of a method of operating a medical device, or of a measuring method using a medical device, as industrially applicable. Two cases are excluded. They are the case where the interaction between the human body and the medical device has a direct and non-temporary effect on the human body, and the case where substantial medical practice is included. [3]

The general standard for medical practice is covered in “Are method-of-treatment claims allowed in Korea?”.

Inventive step: the principle of solving the problem, the learning model and the training data are looked at in that order

Section 3.1 of the Practice Guidelines states the elements of an intelligent robot invention as three. They are the kind of training data of the AI technology applied, the kind of learning model, and the method of using the output data derived from the learning model. When inventive step is assessed, the difficulty of the technical configuration, the distinctiveness of the purpose and the remarkableness of the effect that result from these elements must be taken into account. [2]

The order of assessment in section 3.2.1 of the Practice Guidelines has three steps. The invention of the application is identified from the claims. The closest prior art is identified as the primary (主) prior art invention. The invention of the application is compared with the primary prior art invention, and it is assessed whether a person skilled in the art could easily invent the difference. In doing so, other prior art can be combined in the assessment. [2]

On the primary prior art invention, the Practice Guidelines state two things. It is desirable to select as the primary prior art invention an invention that is closest in technical features to the invention of the application, or that has the same or a similar technical problem, effect or use. The other concerns the case where the technical feature of the invention of the application lies in AI technology. It is desirable to select a prior art invention that is in the same technical field or is for solving the same problem, and that used training data of the same form as the invention of the application. [2]

Section 3.2.2 of the Practice Guidelines states the specific method in four items. [2]

  1. Check whether the invention of the application has a feature in AI technology. If it does not have that feature, it is assessed under the general standard for assessing inventive step.
  2. Check whether the principle of solving the problem is the same in the invention of the application and the prior art invention. If the principle of solving the problem of the invention of the application could be easily conceived from the prior art invention, the difference in the learning model is reviewed and the assessment is made.
  3. Check whether the learning model is the same. If there is a difference, whether a publicly known learning model can be applied is reviewed and the assessment is made.
  4. Check whether the training data is the same. If there is a difference, it is reviewed whether it is something that a person skilled in the art could appropriately select, or whether there is a difference in effect, and the assessment is made.

The 19 cases on issues

Section 4 of the Practice Guidelines contains 19 cases. Seven concern the enablement requirement (Cases 1 to 7), 11 concern inventive step (Cases 8 to 18) and one concerns medical practice (Case 19). [2]

There is a note at the beginning of section 4. The examination cases and cited inventions included were processed and edited into simplified expressions in order to explain examples of the assessment. The result of the assessment in each case may differ from the actual examination result of an examiner. [2]

Enablement requirement: met in two cases, not met in five

The headings in the table below are those that the Practice Guidelines attach to each case on the body pages. The headings on the table of contents pages differ slightly in wording in six cases (Cases 1, 3, 7, 8, 9 and 14). [2]

Case Heading in the Practice Guidelines Case in the Practice Guidelines
1 A case found to meet the enablement requirement even though the description of the invention states only the name of the means for implementing machine learning and has no specific statement related to data processing A robot guide system that analyzes ultrasound images with a machine learning algorithm and detects the positions of a target and a robot finger
2 A case where it was found that the invention can be practiced even though the correlation between the input data and the output data of the trained model is not specifically stated A robot in human form that learns dance movements by deep learning
3 A case where, for an AI invention based on reinforcement learning, it was found that the invention cannot be practiced because the reinforcement learning method is not specifically stated A reinforcement learning system that moves a cleaning robot to a charging device
4 A case where it was found that the invention cannot be practiced because the specific method of learning from data measured by the sensors and the like of the robot is unclear, and the correlation with the control commands derived from the learned data is unclear A robot that cooperates with a person and learns the behavior of the robot with three state variables
5 A case where it was found that the invention cannot be easily practiced because, although a plurality of neural networks are used, each learning model is not specifically disclosed A robot system that selects one of a plurality of trained artificial neural networks and recognizes the motion of the user
6 A case where it was found that the invention cannot be easily practiced because the method of using the output data of the artificial neural network to solve the problem is not specifically stated A service robot system that displays a facial expression according to the inferred situation of the user
7 A case where it was found that the invention cannot be easily practiced because the training data, and the correlation between the training data and the output data, are not clearly stated in the description of the invention A method of diagnosing faults of a reduction gear device that determines from vibration data whether there is a fault and the type of the fault

In Case 1, the Practice Guidelines gave two things as the basis. Distinguishing the form of an object from a captured image is publicly known art that is widely used. That invention is a simple application of machine learning to work that a program or a person performed. At the end of the assessment, the Practice Guidelines added one line in parentheses. It is that whether patentability is acknowledged is a separate requirement. [2]

In Case 2, the Practice Guidelines state the degree of statement in one sentence. It is sufficient to state the input data and the output data only to the extent that it can be inferred that they have a physical relationship with each other. It is not necessary to state them so that they have a mathematical relationship. [2]

Case 3 concerns reinforcement learning. The Practice Guidelines state two ways. One is to specifically state the essential elements such as the agent, the environment, the state, the action and the reward. The other is to state the reinforcement learning method to the extent that it can be inferred through embodiments. The description of the invention in that case stated only a reinforcement learner, a driving sensor and a driving device. [2]

For Cases 4 to 7, the assessment states what was missing from the description of the invention. In Case 4, the method of calculating the third state variable from two state variables was not clearly disclosed. The correlation between the training data set and the control commands was also unclear. In Case 5, it was not specifically stated which of the plurality of neural networks is selected under which condition. In Case 6, the means for solving the problem, for outputting personality and voice from the inferred situation information of the user, was not specifically disclosed. In Case 7, the training data was unclear because it was not specifically defined what the pre-trained learning information that is the object of comparison is, and the method of comparison and its correlation with the presence or absence of a fault were also not clear. [2]

Inventive step: denied in seven cases, acknowledged in four

Case Heading in the Practice Guidelines Conclusion Case in the Practice Guidelines
8 A case where inventive step was denied through the mere addition of publicly known AI technology, where the training data and the learning model are not clearly limited and are stated only as ‘use of artificial intelligence’ Inventive step denied A temperature observation system that inputs sensor data into artificial intelligence and determines whether a welding device is operating normally
9 A case where inventive step was denied by using a cited document that states a learning model of a more generic concept than the learning model stated in the claim Inventive step denied A mobile robot alarm system that analyzes camera images with a convolutional neural network. The cited invention stated a deep learning technique
10 A case where inventive step was denied where the same learning model is used and the training data differs Inventive step denied A door recognition method of a mobile robot that determines the presence of a door with a deep learning model. The cited invention determined the kind of home appliance
11 A case where inventive step was denied because the invention merely systemizes work performed by a person by using publicly known AI technology Inventive step denied A logistics robot system that sets an optimal transport path on the basis of deep learning reinforcement learning
12 A case where inventive step was denied because it was found that there is no effect arising from the difference in the training data of an AI-related invention Inventive step denied A method of predicting the cutting characteristics of a laser 3D cutting robot that inputs the supplied current value into an artificial neural network. The cited invention input the vibration frequency
13 A case where inventive step was denied because, although there is a difference in the input data applied, the learning model is stated only to the extent of simply using it Inventive step denied A calibration device that carries out training of a neural network with position information as the input and the difference between estimated force information and force information as the input signal. The input data of the cited invention, which used a neural network, was a disturbance detection value
14 A case where inventive step was denied because, even though the AI technology applied to the claimed invention and to the cited invention differs, it is no more than a simple change of publicly known AI technology Inventive step denied A reception robot for disease control that models the degree of contamination with AI based on a deep neural network. The cited invention used a GAN algorithm
15 A case where inventive step was acknowledged because, although a learning model of the same scheme is used, the purpose of the invention and the principle of solving the problem differ (that is, the field of use differs) Inventive step acknowledged A device that controls dangerous situations in robot work by competitive training of a generator network and a discriminator network. The cited invention, which used the same scheme, was a counterfeit banknote detection device
16 A case where inventive step was acknowledged because there is a difference in the preprocessing of training data that derives the main characteristics from the input data Inventive step acknowledged A control method of a weld line tracking system that divides an image into patches and calculates the variance of the pixel gray levels
17 A case where inventive step was acknowledged because the feature lies in the use of the output values of an artificial neural network model Inventive step acknowledged A control method that allocates visiting points to a plurality of mowing robots by the optimal travel speed calculated by an artificial neural network
18 A case where inventive step was acknowledged because the way the training result is used differs from the ordinary use of a neural network training result Inventive step acknowledged A collision detection system of a robot that uses the prediction data of a neural network trained on the collision-free state as the reference value for determining a collision

In four of the seven cases where inventive step was denied, the Practice Guidelines give the degree of statement as the basis. In two of them, it is the statement of the claim. The claim in Case 8 did not specifically limit the training data and the learning model. The claim in Case 10 was stated only to the extent of using a deep learning model. In the other two, it is the statement of the description of the invention. The description of the invention in Case 11 did not specifically state how optimization is done by using deep learning reinforcement learning. The assessment in Case 14 presupposes the case where the method of implementing the deep neural network is not specifically stated in the description of the invention. [2]

Cases 10 and 12 contain sentences on the case where only the training data differs. The Practice Guidelines find that, even for training applied to a different technical field, there is ease of combination if it has input data of the same kind of form (Case 10). Even if the training data changes, the training can be found to be easy if the data can be provided in the same format (Case 12). The examples of format that the Practice Guidelines give are images and numerical values. [2]

Case 9 has a “다만” (however) clause. When training by AI, the problem to be solved cannot always be solved in an arbitrary technical field. Therefore, where the learning model is not specified, it is desirable to use a prior art document to the extent that the identity of the technical field and of the problem to be solved can be maintained. [2]

The differences in the four cases where inventive step was acknowledged are stated in the headings. They are the purpose of the invention and the principle of solving the problem, the preprocessing of training data, the use of the output values, and the way the training result is used. [2]

In Cases 16 and 17, the Practice Guidelines added generalized sentences. If there is a detailed difference in the data preprocessing method, whether there is inventive step can be assessed by taking into account the remarkableness of the effect (Case 16). The degree of simply applying an artificial neural network is easy to derive. If the feature lies in the use of the output result and that use goes beyond the scope of ordinary creative ability, inventive step can be acknowledged (Case 17). [2]

The fast track: the advanced robotics designation covers applications for which accelerated examination is requested by 31 October 2026

Applications related to advanced robotics technology are designated as eligible for accelerated examination. The basis has four levels. [1][4][5][6]

  1. Patent Act, Art. 61, subparagraph 2: where it is found necessary to urgently process a patent application prescribed by Presidential Decree
  2. Enforcement Decree of the Patent Act (특허법 시행령), Art. 9(1), subparagraph 2-3: a patent application related to advanced technology that is important to the national economy and the enhancement of national competitiveness, such as semiconductors. It is limited to a patent application for which the Minister of Intellectual Property (지식재산처장) determines and announces the specific cases eligible for accelerated examination and the request period
  3. Public Notice on Request for Accelerated Examination on Patent or Utility Model (특허·실용신안 우선심사의 신청에 관한 고시; Notification, “the Public Notice”), Art. 4(1), subparagraph 2, item (r): an application related to advanced technology such as semiconductors, for which the Minister of Intellectual Property determines and announces the eligible cases and the request period
  4. MOIP Announcement No. 2025-18 (31 October 2025): the eligible cases and the request period for applications related to advanced robotics technology

Art. 9(1), subparagraph 2-3 of the Enforcement Decree was newly inserted by Presidential Decree No. 32973. This Decree was promulgated on 1 November 2022 and entered into force on that day. [4] The current version of the Public Notice is MOIP Notification No. 2026-3 and has been in force from 23 February 2026. The amended provisions of this Public Notice apply from requests for accelerated examination that are filed on or after its enforcement date (Addenda, Art. 2). The former provisions apply to applications for which accelerated examination was requested before its enforcement (Addenda, Art. 3). [5]

Eligible cases and requirements

Announcement No. 2025-18 covers patent applications or utility model applications that are directly related to advanced robotics technology. Both of the two requirements below must be met. [6]

  1. An advanced robotics-related patent classification (CPC) is assigned as the main classification
  2. The application is an application by a company that manufactures, or is preparing to manufacture, advanced robotics-related products, devices and the like in Korea, or an application for the results of a national research and development program related to advanced robotics technology

The Announcement states four advanced robotics-related patent classifications. [6]

CPC Explanation in the Announcement
B25J 1 to 21 Robot arms
G05B 1 to 11, 15 to 99 Robot control or regulating systems
G05D 3 to 99 Control of non-electric variables of robots
G06F 3 Robot interfaces

A request for accelerated examination can be filed for an application for which examination has been requested (Public Notice, Art. 4(1)). [5]

The common documents to be filed that the Announcement states are the Request for Accelerated Examination and the written explanation of the request for accelerated examination. The written explanation states that the request is made under Art. 4(1), subparagraph 2, item (r) of the Public Notice. It also states specifically the reason why the application is one directly related to advanced robotics technology. [6]

Is the application covered by the advanced robotics designation (MOIP Announcement No. 2025-18)?

Request period and extension

The request period of Announcement No. 2025-18 is from 1 November 2025 to 31 October 2026. Eligible applications are those for which accelerated examination is requested in that period. The Announcement states “1년 한시적 시행 후 연장여부 검토 예정” (operated for a limited time of 1 year, after which whether to extend is to be reviewed). [6]

The advanced robotics designation started in 2025. Announcement No. 2025-18 describes itself as an announcement of change (extension) of Korean Intellectual Property Office (KIPO, 특허청) Announcement No. 2025-66. The table in Part 7, Chapter 4, 4.20 of the Patent Examination Guidelines states announcements only up to No. 2025-66, and No. 2025-18 is not in that table. [3][6]

Announcement Date announced Request period
KIPO Announcement No. 2025-66 19 February 2025 19 February 2025 to 31 October 2025
MOIP Announcement No. 2025-18 31 October 2025 1 November 2025 to 31 October 2026

The wording of the eligible cases is the same in the two announcements. The main classification requirement, the four patent classifications, and the requirement of manufacturing in Korea or a national research and development program are unchanged. On extension, the earlier announcement stated “한시적 시행 후 ’25년 11월 연장여부 검토 예정” (operated for a limited time, after which whether to extend is to be reviewed in November 2025). [6]

As of 10 October 2026, there is no announcement on the MOIP notices and announcements board that gives notice of a designation from 1 November 2026. The newest post on the board is that of 8 October 2026. The only posts with “로봇” (robot) in the title are the two announcements above. The newest posts with “우선심사” (accelerated examination) in the title are the two announcements issued again on 30 January 2026, and they do not concern advanced robotics. [6]

Can foreign applicants use it?

Announcement No. 2025-18 has no wording that asks about the nationality or address of the applicant. The main clause of Art. 3 of the Public Notice states that any person can file a request for accelerated examination. The proviso concerns an application that concerns the duties of the State or a local government (Art. 4(1), subparagraph 2, item (d)). [5][6]

However, the second requirement of the Announcement is based on manufacturing in Korea and a national research and development program. An application that falls under neither of the two is not eligible for accelerated examination as set by this Announcement. [6]

The wording of the requirement of manufacturing in Korea is “국내에서 생산하거나 생산준비 중인 기업” (a company that manufactures, or is preparing to manufacture, in Korea). The Announcement does not limit the nationality of that company or the location of its head office. The examples of supporting documents that the Announcement gives are a factory registration certificate, a goods supply contract, a delivery confirmation, a quality agreement and a publicly disclosed business report. The Announcement describes these as documents that prove that manufacturing and the like take place in Korea. [6]

The supporting document for the national research and development program requirement is a document in which the applicant is specified as the lead research and development institution or a joint research and development institution of a national research and development project related to the invention of the application. The Announcement gives an agreement and a research and development plan as examples. [6]

The way to decide whether an application is eligible is stated in one sentence in the Examination Guidelines. Part 7, Chapter 4, 4.20 of the Examination Guidelines states that whether an application is eligible for accelerated examination is decided by referring to the text of the announcement or to internal guidelines set separately. [3]

Other grounds that do not state manufacturing in Korea as a requirement

Art. 4(1), subparagraph 2, item (p) of the Public Notice is a patent application that utilizes technology related to the Fourth Industrial Revolution. The technologies stated in item (p) include intelligent robots. It must be a patent application to which MOIP has assigned a new patent classification of attached Table 3. Attached Table 3 states the classification of “지능형 로봇” (intelligent robots) as Z03R. [5]

Part 7, Chapter 4, 4.15 of the Examination Guidelines states the following on this ground. Utility model applications are not eligible. The new patent classification system applies from applications filed on or after 1 January 2018, and the new patent classification is used only as a secondary classification or additional information. No separate supporting document needs to be submitted when accelerated examination is requested. It is sufficient for the requester to state in the written explanation of the request for accelerated examination that the application is technology that falls under one or more of the technology fields of item (p). There is also a sentence on a patent application to which the patent classification agency has not assigned a new patent classification. The examiner must look at the content of the invention and assess whether the application is directly related to the Fourth Industrial Revolution. If the examiner finds that the existing patent classification is wrong, the examiner can add a new patent classification by ex officio correction, and can accordingly acknowledge the application as eligible for accelerated examination. [3]

Art. 4(1) of the Public Notice has other grounds as well. Three of them are below. [5]

  • Subparagraph 1: an application for which it is found that a third party is working the filed invention for business purposes after the application is laid open
  • Subparagraph 3: a patent application that the Minister of Intellectual Property has agreed with the head of a foreign patent authority to examine on an accelerated basis, that meets the request requirements of attached Table 2, and that falls under item (a) or (b) of the same subparagraph. It is limited to the case where the supporting documents of attached Table 1 are attached
  • Subparagraph 2, item (q): an international patent application for which MOIP carried out the international search under the Patent Cooperation Treaty and for which the document under Art. 203 of the Patent Act has been filed

The requirements and request procedure for these grounds are covered in “How do I get accelerated examination in Korea?“.

Classifications that overlap with the artificial intelligence designation

Applications related to artificial intelligence technology are also designated under the same item (r). It is MOIP Announcement No. 2026-22 (30 January 2026). The artificial intelligence-related patent classifications that this Announcement states include classifications on robots and control. They are B25J 9 (robot manipulators), G05B 13 (AI adaptive control systems), G05B 19 (intelligent control systems) and G05D (systems for controlling physical quantities). This Announcement is an announcement of change (extension) of MOIP Announcement No. 2025-19. [6]

The classifications of the two announcements partly overlap and partly diverge. G05B 13 is not in the range of the advanced robotics announcement (G05B 1 to 11, 15 to 99) and is in the artificial intelligence announcement. B25J 9, G05B 19 and G05D 3 to 99 are in both announcements. The rest of B25J and G06F 3 are only in the advanced robotics announcement. [6]

Announcement No. 2026-22 also requires that both the main classification requirement and the second requirement be met. Eligible applications are those for which accelerated examination is requested from 1 February 2026 to 31 October 2026. The wording of the second requirement differs from that of the advanced robotics announcement. The artificial intelligence announcement also states an application by a company that is developing artificial neural networks, training methods, systems and the like in Korea. [6]

Super-accelerated examination

Art. 4(2) of the Public Notice calls accelerated examination for being examined earlier than other applications for which accelerated examination has been requested super-accelerated examination (초고속심사). There are two cases in which an application under item (r) is eligible for a request for super-accelerated examination. They are the case where it also falls under an application that serves as the basis of a priority claim under a treaty (paragraph (1), subparagraph 2, item (h)) (paragraph (2), subparagraph 2), and the case where at least one of the applicants at the time of first filing is a start-up company announced by the Minister of Intellectual Property (paragraph (2), subparagraph 3). MOIP Announcement No. 2026-60 (23 February 2026) included advanced robotics in the technology fields eligible for a request under subparagraph 3. For advanced robotics, AI application technology must be included, and the person qualified to file the request is any one of a start-up company under the Support for Small and Medium Enterprise Establishment Act (중소기업창업 지원법), a venture business under the Special Act on the Promotion of Venture Businesses (벤처기업육성에 관한 특별법) and a technology-innovative small and medium enterprise under the Act on the Promotion of Technology Innovation of Small and Medium Enterprises (중소기업 기술혁신 촉진법). Eligible applications are those for which accelerated examination is requested by 31 December 2026. [5][6]

Dedicated organization

The examination of patent applications and utility model applications in the intelligent robot field is a matter assigned to the head of the Digital Convergence Examination Bureau (Decree on the Organization of the Ministry of Intellectual Property and Institutions under Its Jurisdiction (지식재산처와 그 소속기관 직제, “Organization Decree”), Art. 17(3), subparagraph 1). The Organization Decree states convergence technology as artificial intelligence and big data, the Internet of Things, bio and healthcare, intelligent robots, and autonomous driving and smart manufacturing. [7]

The Digital Convergence Examination Bureau has the Intelligent Robot Examination Division (Enforcement Rule of the same Organization Decree, Art. 14(2)). The fields assigned to the head of the Intelligent Robot Examination Division are robot arms, robot manipulators, robot control, robot interfaces, robot monitoring and robot simulation (Art. 14(6)). [7]

Systems that exist only in this field: there are none in the main text of the Patent Act

A search of the main text of the Patent Act for “로봇” (robot) and “인공지능” (artificial intelligence) returns no result. No provision on filing, examination or rights that applies only to robot inventions was found in the main text of the Patent Act. The scope searched is the main text and the Addenda of the Patent Act in force from 11 November 2025 and of the version scheduled to enter into force on 16 March 2027. [1]

What exists only in this field therefore comes down to three things. They are Chapter VI of the Practice Guidelines, the advanced robotics designation for accelerated examination together with the Fourth Industrial Revolution classification, and the dedicated examination division. [1][2][5][6][7]

Where the Practice Guidelines are out of date

The opening note of Chapter VI and the history at the end of the book give May 2023 as the last revision. The places that changed after that, and the places where the notation does not match, are as follows. [2][3]

  • Revision of the Practice Guidelines for the AI field. The Practice Guidelines for the AI field (Chapter I), which Chapter VI states are to be followed, were revised in January 2026. The opening note of Chapter I states “개정 2026. 1.” (revised January 2026), and the foot of the pages of Chapter I is marked “2026년 1월 추록” (supplement of January 2026). The names of the sections that Chapter VI points to (“2.1 발명의 설명”, “2.2 청구범위”, “3. 특허요건”) are still in Chapter I of the March 2026 edition. Chapter VI states that an invention whose feature is only the AI of a robot follows those sections of Chapter I, and does not state which edition of Chapter I is followed.
  • Version of the Examination Guidelines. The current version of the Patent Examination Guidelines, which Chapter VI states are to be followed, is MOIP Established Rule No. 7. It has been in force from 13 March 2026. The marking at the foot of the pages of Part 3, Chapter 1, 5.1, which deals with medical practice, is “2021년 12월 추록” (supplement of December 2021) in the current version as well.
  • Names and numbers of chapters. Section 3.1 of Chapter VI states the place of novelty as “제3부제2항” (Part 3, paragraph 2). In the table of contents of the Examination Guidelines, novelty is Part 3, Chapter 2. Chapter VI states Part 2, Chapters 3 and 4 as “발명의 설명 기재요건” (requirements for the description of the invention) and “청구범위 기재요건” (requirements for the claims). The chapter names in the table of contents of the Examination Guidelines are “발명의 설명” (description of the invention) and “청구범위” (claims).
  • Marking of the time of establishment. The opening note of Chapter VI gives December 2021 for establishment, and the history at the end of the book gives January 2022.
  • Headings of the cases. The headings of Cases 1, 3, 7, 8, 9 and 14 differ in wording between the table of contents pages and the body pages. Case 14 has “단순 적용한 정도” (no more than a simple application) in the table of contents and “단순 변경한 정도” (no more than a simple change) in the body. The tables in this article follow the headings on the body pages.

On the accelerated examination side, the advanced robotics designation was created on 19 February 2025. Chapter VI does not deal with accelerated examination. The table in Part 7, Chapter 4, 4.20 of the Examination Guidelines does not yet contain Announcement No. 2025-18 of 31 October 2025. [2][3][6]

What applicants should do

In general, DEFINE IP Law Firm suggests the following.

  • Before filing, first decide where the feature of the invention lies. The standard to read differs depending on which it is: control technology and component technology, AI technology only, or an organic combination of AI technology and a robot (Practice Guidelines 1.2).
  • Check whether the description of the invention states the kind of training data, the correlation between the input data and the output data, the kind of learning model and the method of using the output data (Practice Guidelines 2.1.1). Also check the preprocessing process by which the measured values of a sensor become the input of the learning model (Practice Guidelines 2.1.2 (5)).
  • For an invention that uses reinforcement learning, check whether the agent, the environment, the state, the action and the reward are stated in the description of the invention (Case 3).
  • Check that the claim does not stop at the level of “using artificial intelligence”. In the cases where the Practice Guidelines acknowledged inventive step, the difference lay in the preprocessing, the use of the output values, the way the training result is used, and the principle of solving the problem (Cases 15 to 18).
  • For a method claim of a robot used for surgery, therapy or diagnosis, state clearly whether the entity that performs each step is a medical device or a medical practitioner (Practice Guidelines 2.2.3).
  • To use the advanced robotics designation for accelerated examination, accelerated examination must be requested by 31 October 2026. First check whether both the main classification requirement and the requirement of manufacturing in Korea or a national research and development program are met. If the application does not fit those requirements, look at other grounds such as the Fourth Industrial Revolution classification (Z03R).

Please confirm with your Korean agent which standard and which ground for accelerated examination fit an individual application.

FAQ

Does Chapter VI of the Practice Guidelines apply to every invention relating to a robot?

No. Chapter VI applies to intelligent robot inventions in which there is an organic combination of AI technology and a robot. Section 1.2 of the Practice Guidelines states that the general Examination Guidelines apply to a general robot invention whose feature lies in control technology and component technology. If the feature lies only in AI technology, the Practice Guidelines for the AI field apply.

Must the structure of the learning model be stated in the description of the invention?

Section 2.1.2 (3) of the Practice Guidelines states that the structure of the input layer, the intermediate layer and the output layer must be stated specifically and in detail. There is a proviso. This does not apply if the structure of the learning model is common general technical knowledge that is self-evident to a person skilled in the art, or can be understood and reproduced through an embodiment.

Can a method of controlling a surgical robot be patented in Korea?

It depends on the statement of the claim. Section 2.2.3 of the Practice Guidelines states that it is checked whether medical practice is performed in each step of a method claim and whether each step restricts the medical practice of a medical practitioner. For a claim that ends with “물건” (product), assessment of medical practice is unnecessary. In Case 19, the Practice Guidelines found a method of operation in which only the information processing of the system is stated not to be medical practice.

Can a foreign company that does not manufacture in Korea also use the advanced robotics designation for accelerated examination?

Announcement No. 2025-18 has no wording that asks about the nationality or address of the applicant. However, together with the main classification requirement, the application must be an application by a company that manufactures, or is preparing to manufacture, in Korea, or an application for the results of a national research and development program. If it falls under neither of the two, it is not covered by this Announcement. Art. 4(1) of the Public Notice has other grounds that do not state manufacturing in Korea as a requirement.

Will the advanced robotics designation for accelerated examination continue after November 2026?

As of 10 October 2026, this cannot be known. Announcement No. 2025-18 states “1년 한시적 시행 후 연장여부 검토 예정” (operated for a limited time of 1 year, after which whether to extend is to be reviewed). As of that day, there is no announcement on the MOIP notices and announcements board that gives notice of a designation from 1 November 2026.

How DEFINE IP Law Firm handles this

Before drafting the specification of an intelligent robot invention, DEFINE IP Law Firm first checks where the feature of the invention lies. The question is which of AI technology, control technology and component technology it is. For an invention whose feature lies in AI technology, we look at the description of the invention and the claims before filing. We check whether the training data, the learning model and the method of using the output data are stated. For accelerated examination, we file the request after checking, for each application, which ground’s requirements it meets against the wording of the announcement and the Public Notice. For questions about Korean applications for intelligent robot inventions, please write to contact@defineip.kr.

Sources

  1. Ministry of Intellectual Property (지식재산처), Patent Act (특허법), Act No. 21134, in force from 11 November 2025, Art. 29, Art. 42, Art. 61, Art. 62. Version scheduled to enter into force (Act No. 21955, promulgated 15 September 2026, in force from 16 March 2027). https://www.law.go.kr/법령/특허법 (verified 10 October 2026)
  2. Ministry of Intellectual Property (지식재산처), Examination Practice Guidelines by Technology Field (기술분야별 심사실무가이드), March 2026, Chapter VI, Intelligent Robot Field (제6부 지능형로봇 분야; the opening note states establishment in December 2021 and revision in May 2023), the foreword, the history at the end of the book, and the opening note (revised January 2026), table of contents and 3.1.1 of Chapter I, Artificial Intelligence Field. https://www.kipo.go.kr/ko/kpoContentView.do?menuCd=SCD0200147 (verified 10 October 2026)
  3. Ministry of Intellectual Property (지식재산처), Patent Examination Guidelines (특허·실용신안 심사기준), MOIP Established Rule No. 7, in force from 13 March 2026, table of contents; the “관련 규정” (relevant provisions) of Part 2, Chapters 3 and 4 and of Part 3, Chapter 3; Part 3, Chapter 1, 5.1; Part 7, Chapter 4, 4.15 and 4.20. https://www.law.go.kr/행정규칙/특허·실용신안심사기준 (verified 10 October 2026)
  4. Ministry of Intellectual Property (지식재산처), Enforcement Decree of the Patent Act (특허법 시행령), Presidential Decree No. 35809, in force from 1 October 2025, Art. 9. Presidential Decree No. 32973 (promulgated and in force on 1 November 2022). https://www.law.go.kr/법령/특허법시행령 (verified 10 October 2026)
  5. Ministry of Intellectual Property (지식재산처), Public Notice on Request for Accelerated Examination on Patent or Utility Model (특허·실용신안 우선심사의 신청에 관한 고시), MOIP Notification No. 2026-3, in force from 23 February 2026, Art. 3, Art. 4, Addenda Arts. 2 and 3, attached Table 3. https://www.law.go.kr/행정규칙/특허·실용신안우선심사의신청에관한고시 (verified 10 October 2026)
  6. Ministry of Intellectual Property (지식재산처), Announcement No. 2025-18 “첨단기술 관련 출원(첨단로봇 기술)의 우선심사 대상 지정 공고” (Announcement designating advanced technology applications (advanced robotics technology) eligible for accelerated examination, 31 October 2025), KIPO Announcement No. 2025-66 (19 February 2025), MOIP Announcement No. 2026-22 (artificial intelligence technology, 30 January 2026), No. 2026-60 (super-accelerated examination for start-up support in 2026, 23 February 2026), and the notices and announcements board. https://www.kipo.go.kr/ko/kpoContentView.do?menuCd=SCD0200225 , https://www.kipo.go.kr/ko/kpoBultnMgmt.do?menuCd=SCD0200610&parntMenuCd2=SCD0200049 (verified 10 October 2026)
  7. Ministry of the Interior and Safety (행정안전부) and Ministry of Intellectual Property (지식재산처), Decree on the Organization of the Ministry of Intellectual Property and Institutions under Its Jurisdiction (지식재산처와 그 소속기관 직제), Presidential Decree No. 36514, in force from 21 July 2026, Art. 17; Enforcement Rule of the same Organization Decree, Prime Ministerial Decree No. 2138, in force from 21 July 2026, Art. 14. https://www.law.go.kr/법령/지식재산처와그소속기관직제 (verified 10 October 2026)