China Software and AI Patents: The Technical Solution Requirement

In China, software and AI inventions are patentable when they present a technical solution to a technical problem using technical means and produce a technical effect. A claim to a program alone is not granted. CNIPA's Guidelines for Patent Examination address algorithms and AI. Copyright protects the underlying code separately through voluntary registration.

Can software and AI be patented in China?

The starting point in Chinese practice is that a claim to a computer program alone, divorced from any technical solution, will not be granted. This flows from two sources rather than a single clean statutory carve-out: the China National Intellectual Property Administration (CNIPA) treats a bare program as unpatentable under its Guidelines for Patent Examination, and the Patent Law itself excludes rules and methods for mental activities (Article 25) and requires that an invention be a technical solution. A mathematical method or an abstract algorithm, standing alone, falls outside what CNIPA will grant a patent for. That position sounds broad, but it is not the end of the enquiry. A claim framed as a computer-implemented invention, including many AI-related inventions, can be patented where it does more than recite an abstract algorithm. The question CNIPA asks is whether the claimed subject matter, read as a whole, amounts to a technical solution.

For a software business or an AI developer, this reframes the drafting task. You are not seeking to patent the model, the training method, or the mathematics in the abstract. You are seeking protection for the way a technical means solves a technical problem and produces a technical effect. Getting that framing right at the drafting stage is what separates an allowable claim from a refusal.

The technical solution test

Chinese practice turns on three linked ideas: a technical problem, technical means, and a technical effect. A claim should identify a genuine technical problem, describe technical means (typically the interaction of the algorithm with hardware, data, or a controlled process) that address it, and point to a technical effect that follows from those means. An improvement that is purely commercial, administrative, or presentational, with no bearing on how a system works technically, tends not to clear the bar.

For AI inventions in particular, examiners look at whether the algorithm is tied to a specific technical field and a concrete technical result. An image-recognition method that improves how a device processes visual data, a control method that makes an industrial process more efficient, or a training approach that produces a measurable technical improvement in a system are the kinds of framing that read as technical. A claim that stops at "a better prediction" in a business or financial sense, without a technical anchor, is weaker. The practical lesson is to draft claims and a specification that make the technical contribution explicit rather than leaving it implied.

CNIPA examination guidelines for algorithms and AI

CNIPA has updated its Guidelines for Patent Examination to address algorithmic features and AI more directly, giving examiners a clearer framework for computer-implemented inventions. The guidance recognises that algorithmic features and technical features can work together, and it sets out how such combined subject matter is assessed. It also accepts several claim formats beyond the method and apparatus claim: since the Guidelines were revised in this area, applicants can draft claims as a storage medium carrying a program flow, or as a computer-program product, where those formats carry a technical solution. This matters because it means qualifying software and AI inventions are not confined to method claims alone. Because examination practice here continues to develop, treat any summary as general orientation rather than a fixed rule, and confirm the current version of the Guidelines and their application with CNIPA or local counsel before you rely on a particular claim strategy. The broader policy backdrop, and how different offices are converging or diverging on machine-learning inventions, is covered in our overview of AI and IP across jurisdictions.

On timing and cost, a Chinese invention patent goes through formal examination and substantive examination, and the total time to grant varies with the technology and the backlog. Official fees apply at various stages; confirm the current amounts and timeframes with CNIPA or local counsel rather than budgeting from a figure you have seen quoted elsewhere. For a fuller picture of the invention patent route, utility models, and designs in China, see the China patents pillar.

One point often misunderstood: utility models do not cover software. The Chinese utility model is limited to a technical solution relating to the shape, structure, or their combination, of a product, so it is not a route for a computer program or an AI method. Software and AI inventions that qualify are protected through invention patents, not utility models.

Patents are not the only layer of protection for software in China. Source and object code are protected under the Copyright Law and its dedicated Regulations on the Protection of Computer Software, and the right arises automatically on creation. China also operates a voluntary software copyright registration system, administered by the Copyright Protection Centre of China (CPCC) under the National Copyright Administration, and many businesses register because a registration certificate is useful evidence of ownership and creation date, which helps in enforcement and in commercial dealings. Registration is not a precondition for the right to exist, but it is often worth doing.

The two rights protect different things and work well together. A patent, where available, protects the underlying technical solution regardless of how the code is written; copyright protects the specific expression of the code against copying. Relying on copyright alone leaves the functional idea open to independent reimplementation, which is precisely the gap a well-drafted patent can close.

Where IPEnvoy fits

IPEnvoy is not a law firm and does not provide legal advice. This is general information only. Chinese patent practice for software and AI is evolving, and the outcome for any given invention depends heavily on how the claims are drafted and examined, so confirm the current position with CNIPA and a qualified local IP professional before you file. If it would help to be introduced to a vetted IP firm in China who handles computer-implemented and AI inventions, IPEnvoy can point you in the right direction.

Related

Author: Steffen Hoyemsvoll

Reviewers: pending review