Japan Software and AI Patents: How Computer-Implemented Inventions Are Protected

Can you patent software in Japan?

Software businesses often assume Japan treats a program as unpatentable subject matter, in the way some jurisdictions resist. That assumption is wrong. Japanese patent practice recognises software and AI inventions, but it frames them in a particular way, and understanding that framing is the difference between a claim the Japan Patent Office (JPO) will examine on its merits and one that falls at the subject-matter hurdle. The starting point is that an invention must be a creation of technical ideas utilising a law of nature. A pure business method, an abstract algorithm, or a mathematical formula standing alone does not meet that bar. What makes the difference is how the software interacts with a machine.

The hardware-resources requirement

The central concept is that a computer program is patentable as a computer-implemented invention where the information processing performed by the software is concretely realised using hardware resources. In plain terms, the claimed invention has to show that the software and the hardware work together to produce a specific technical result, rather than describing an idea in the abstract. The JPO looks for the information processing to be embodied in a concrete way through the cooperation of the program and the computer's resources, such as processing, memory, and input and output.

This shapes how you draft. Claims that recite a purely mental or administrative scheme, dressed up with the words "using a computer", tend to struggle. Claims that describe how data is processed by defined hardware to achieve a technical effect stand on much firmer ground. For an AI product, that often means articulating the technical contribution of the model in operation: how inputs are transformed, what the system computes, and the concrete result delivered, rather than claiming the underlying mathematics in isolation. This is the same discipline you would bring to a patent filing generally, and the broader route considerations are covered on the Japan patents pillar.

How the JPO examines AI-related inventions

The JPO has published dedicated guidance on AI. Its Examination Guidelines, together with the AI-related case examples in its Examination Handbook, specifically address AI-related inventions, illustrating how the office applies eligibility, novelty, inventive step, and the description requirements to machine-learning and data-driven inventions. These examples are worth studying closely before drafting, because they show the kinds of technical detail examiners expect to see.

Two themes recur. First, the description (enablement and support) requirement matters a great deal for AI. If a claim asserts that a trained model achieves a particular performance, the specification generally needs to make that credible, for instance by explaining the relationship between the input data, the training, and the claimed effect, so that a skilled reader could reproduce it. A claim that promises a result the specification does not substantiate is vulnerable. Second, inventive step for AI inventions is assessed against what a skilled person would find obvious, and simply applying a known machine-learning technique to a new dataset is not automatically inventive. The Examination Guidelines and the worked case examples in the Handbook are the best current signal of where the line sits, and they are updated over time, so confirm you are working from the latest version published by the JPO. The wider cross-border picture, including how different offices approach these questions, is discussed in our overview of AI and IP internationally.

Copyright protects the code, patents protect the method

Patents are not the only layer. In Japan, software is separately protected by copyright as a program work under the Copyright Act. This protection arises automatically on creation and covers the expression of the code, not the underlying idea or method. The two rights complement each other rather than overlap: copyright guards against someone copying your source or object code, while a patent can protect the functional method or system your software implements, which copyright does not reach. A robust position for a software or AI company usually combines both, alongside trade-secret protection for training data, model weights, and know-how that you do not wish to disclose in a published patent.

Practical steps and where to get help

If you are considering protection in Japan, the sequence is familiar: assess whether the technical contribution can be framed as a computer-implemented invention, draft claims that tie the information processing to hardware resources, and file either directly with the JPO or through an international route. Substantive examination in Japan is not automatic; it must be actively requested within a statutory period that runs from the filing date. That deadline is fixed by statute rather than something that drifts, but several distinct official fees apply across prosecution, including a request-for-examination fee that varies with the number of claims, and these are revised from time to time. Confirm the current deadline and the current fee amounts with the JPO or local counsel rather than relying on figures quoted elsewhere.

IPEnvoy is not a law firm and does not provide legal advice. This is general information, and the position on software and AI patentability evolves as the JPO updates its guidance and as case law develops, so confirm the current position with the JPO and a qualified local IP professional before acting. If it would help, we can introduce you to a vetted IP firm in Japan to assess your specific invention and handle the filing.