# Artificial Intelligence and Patents: Protecting AI Technology As AI technology develops rapidly, patent activity in the field is also growing. Source: https://www.iplexlaw.co.kr/en/blog/843118 HOME / NEWS & INSIGHTS NEWS & INSIGHTS Artificial Intelligence and Patents: Protecting AI Technology As AI technology develops rapidly, patent activity in the field is also growing. AI & Software 2023.05.12 published IPLEX 4 min read Hello, this is patent attorney Yongduck Kim of IPLEX IP Law Firm. As AI technology develops rapidly, patent activity in the field is also growing. In this post, we'll explain which of the most used deep learning technologies in AI can be protected by patent . We will omit the specific description of the technical terms, assuming that you have an understanding of AI technology. Potential areas for patent protection The following six parts of a deep-learning implementation may contain patentable technical contributions. Patentability depends on the specific invention and applicable requirements. 1 Acquisition of training data set 2 Data preprocessing method 3 Deep learning model structure design 4. Training methods 5 Hardware design 6 Deep Learning Applications Other technical aspects may also merit assessment. Acquiring training datasets Simply downloading a public dataset does not establish patentability. A technically distinctive method of obtaining data may warrant assessment. For example, a specific data-acquisition rule that solves a technical problem may be protectable if it meets novelty, inventive-step and other requirements. There are currently not many applications in relation to the field. Data Preprocessing Method After obtaining the training data set, the data set is usually processed by preprocessing. Preprocessing can include noise removal, normalization and augmentation. A specific improvement or adaptation that solves a technical problem may support patent protection. Deep Learning Model Structural design After preprocessing, an existing deep-learning model may be selected or a new architecture designed. Using a familiar model such as a CNN, RNN or LSTM is not itself a patentable distinction. The assessment concerns specific architectural changes or combinations addressing a technical problem. Such changes can be demanding to develop, and this article notes that architecture-focused inventions account for relatively few of the AI patents discussed. Selecting a training method Training follows preprocessing. Applying a conventional training method alone does not establish patentability. A training method that differs from the prior art may warrant patent assessment. For example, a training method could constrain hidden-layer outputs to fit a defined distribution. The potentially protectable contribution lies in the technical difference from prior art. The number of patents associated with the learning method is next to those related to data preprocessing. Hardware Design Implementation of a trained model and the processing used to generate predictions may also warrant protection. These issues can overlap with hardware and semiconductor design, where companies have developed substantial patent portfolios. Deep Learning Application Many AI applications concern using deep learning in a particular field. Merely applying a known model does not by itself establish patentability. The application should explain the specific training data, processing and technical effects that distinguish the invention. Read the Korean source This article reflects the information available when it was published. Contact us to discuss your circumstances. Discuss this topic ↗ All articles TALK TO IPLEX Discuss your IP questions We consider your technology and business needs together. ↗ Contact us Newer Publication Announcement: Design Protection Law — Protect Your Designs ↗ Older AI Times Column: Preparing for Special Listing under Revised Evaluation Criteria ↗ Related insights AI & Software 2026.10.01 FiX: fine-grained forgetting in softmax attention Yongduck Kim examines FiX’s feature-wise gates, numerical implementation and paged cache, distinguishing reported gains from unresolved limitations. ↗ Read article AI & Software 2026.09.30 MHAR: Reading earlier layers through different feature subspaces Yongduck Kim examines Multi-Head Attention Residuals: depth routing, reported training results, implementation costs and the relationship between technical features and effects. ↗ Read article AI & Software 2026.09.28 Column: Claude Computer Use and the Data That Trains AI Agents Writing for AI Times, IPLEX Managing Partner Yongduck Kim examines the training data behind computer-operating AI agents through U.S. Patent No. 12,585,862. ↗ Read article - https://www.iplexlaw.co.kr/en - https://www.iplexlaw.co.kr/en/blog/category/ai - https://www.iplexlaw.co.kr/forum/view/843118 - https://www.iplexlaw.co.kr/en/contact - https://www.iplexlaw.co.kr/en/blog - https://www.iplexlaw.co.kr/en/contact - https://www.iplexlaw.co.kr/en/blog/843161 - https://www.iplexlaw.co.kr/en/blog/831543 - https://www.iplexlaw.co.kr/en/blog/fix-fine-grained-forgetting-attention - https://www.iplexlaw.co.kr/en/blog/fix-fine-grained-forgetting-attention - https://www.iplexlaw.co.kr/en/blog/mhar-multi-head-attention-residuals - https://www.iplexlaw.co.kr/en/blog/mhar-multi-head-attention-residuals - https://www.iplexlaw.co.kr/en/blog/claude-computer-use-agent-patent - https://www.iplexlaw.co.kr/en/blog/claude-computer-use-agent-patent