Chapter 3 covered the four substantive requirements every invention must satisfy: novelty, non-obviousness, utility, and sufficient disclosure. But before an examiner even evaluates those requirements, a threshold question must be answered: does your invention fall within the categories of subject matter that the law considers eligible for patent protection at all? This chapter examines the boundaries of patentable subject matter -- the exclusions that apply universally, the frameworks that major jurisdictions use to draw the line, and the practical consequences for software, AI, biotechnology, business methods, and standard-essential patents.
Universal Exclusions: What No Country Will Patent
Despite significant differences in how patent offices evaluate applications, a set of core exclusions applies across virtually every jurisdiction. These exclusions reflect a shared policy judgment: certain categories of knowledge must remain freely available to everyone.
| Excluded Category | Why It Is Excluded | Example |
|---|---|---|
| Laws of nature | Fundamental principles govern all of science; monopolizing them would halt progress | Newton's laws of motion, the laws of thermodynamics |
| Natural phenomena | Naturally occurring processes and substances are discoveries, not inventions | Photosynthesis, the electromagnetic spectrum |
| Abstract ideas | Concepts without concrete implementation are not technological contributions | A theory of economic optimization, a mental reasoning process |
| Mathematical formulas (in isolation) | Mathematics is the shared language of science and engineering | The Fourier transform, a statistical regression formula |
| Mere discoveries | Finding what already exists in nature is not creating something new | Identifying a new mineral, observing a new astronomical body |
| Perpetual motion machines | Violate established physical laws and lack credible utility | Any device claimed to generate energy from nothing |
| Aesthetic creations (in isolation) | Protected by copyright, not patents; they lack technical character | A painting, a novel, a musical composition |
These exclusions are not absolute barriers. The critical qualifier -- present in nearly every patent statute -- is the phrase "as such." A mathematical formula in isolation is not patentable, but a concrete technical system that uses that formula to solve a real-world problem can be. A discovery of a natural property is not patentable, but a practical application of that property in a new device or process may well be. The distinction between "as such" and "applied in a technical context" is where the real work of subject matter eligibility analysis takes place.
The practical question is never whether your invention involves an excluded category -- most inventions involve math, natural laws, or abstract reasoning at some level. The question is whether your claims go beyond the excluded category to recite a concrete, technical application.
The U.S. Framework: 35 U.S.C. 101 and the Alice/Mayo Test
The United States defines patentable subject matter broadly. Under 35 U.S.C. 101, a patent may be granted for "any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof." On its face, this language is expansive. The Supreme Court, however, has carved out three judicial exceptions: laws of nature, natural phenomena, and abstract ideas.
The Two-Step Framework
Three landmark Supreme Court decisions -- Mayo Collaborative Services v. Prometheus Laboratories (2012), Alice Corp. v. CLS Bank International (2014), and Bilski v. Kappos (2010) -- collectively established a two-step test that the USPTO applies to every patent claim:
Step 1: Is the claim directed to a judicial exception?
The examiner asks whether the claim is directed to a law of nature, a natural phenomenon, or an abstract idea. If the answer is no, the claim is eligible. If the answer is yes, the analysis proceeds to Step 2.
Step 2: Does the claim recite something "significantly more"?
The USPTO's current guidance breaks this into two sub-steps:
-
Step 2A (Prong 2): Does the claim integrate the judicial exception into a practical application? Indicators of integration include improvements to the functioning of a computer, application of the exception to effect a particular treatment or prophylaxis for a disease, application of the exception with a particular machine that is integral to the claim, or transformation of a particular article to a different state or thing.
-
Step 2B: If the claim does not integrate the exception into a practical application, does it recite an inventive concept -- an element or combination of elements that amounts to significantly more than the judicial exception itself? Routine, conventional, and well-understood activity does not qualify.
The USPTO's Alice/Mayo two-step eligibility analysis
Landmark Cases That Shaped the Framework
Bilski v. Kappos (2010) addressed a business method for hedging energy risk. The Court held the claims were directed to an abstract idea and rejected a rigid "machine-or-transformation" test as the sole eligibility standard, narrowing the path for business method patents.
Mayo v. Prometheus (2012) involved a diagnostic method correlating drug metabolite levels to dosage adjustments. The Court held that adding routine data-gathering steps to a law of nature is not enough to create patent eligibility, establishing the two-step framework.
Alice Corp. v. CLS Bank (2014) challenged a computerized escrow system for financial transactions. The Court held that implementing an abstract idea on a generic computer does not make it eligible -- the most consequential modern eligibility decision.
The Alice decision had seismic consequences. Federal courts invalidated patents at unprecedented rates under Section 101 in the years that followed. The decision made clear that simply reciting "do it on a computer" does not transform an abstract idea into patentable subject matter, while also creating significant uncertainty about where the boundary lies.
Post-Alice Reality: The Alice decision did not make all software unpatentable. It invalidated claims that amounted to abstract ideas implemented on generic hardware. Software inventions that improve computer functionality, solve technical problems in computing, or apply algorithms to produce specific technical results remain eligible. The challenge is drafting claims that clearly articulate the technical improvement rather than the abstract concept. The USPTO issued updated guidance in 2019 (the Revised Patent Eligibility Guidance) and again in 2024 (including AI-specific examples) to help examiners and applicants navigate this landscape.
The EPO Framework: Article 52 EPC and Technical Character
The European Patent Office takes a structurally different approach. Rather than broad statutory language narrowed by judicial exceptions, the European Patent Convention starts with an explicit list of things that are not considered inventions.
Article 52 EPC Exclusions
Article 52(2) of the European Patent Convention lists the following as not being regarded as inventions:
- Discoveries, scientific theories, and mathematical methods
- Aesthetic creations
- Schemes, rules, and methods for performing mental acts, playing games, or doing business
- Programs for computers
- Presentations of information
Article 52(3) adds the crucial qualifier: these exclusions apply only "to the extent to which a European patent application or European patent relates to such subject-matter or activities as such."
The Technical Character Requirement
The EPO's approach revolves around a single concept: technical character. An invention must have technical character to be patentable. Subject matter from the Article 52(2) list is excluded only when claimed "as such" -- that is, when it lacks any technical aspect. The moment excluded subject matter is combined with or applied to a technical purpose, it can form part of a patentable invention.
For computer-implemented inventions, the EPO applies this through its "any hardware" approach combined with the COMVIK framework (named after the T 641/00 decision). If a claim specifies that software runs on a computer or other technical device, it has technical character and passes the threshold for being an "invention." The real question then shifts to inventive step: only the features that contribute to the technical character of the invention are considered when assessing whether an inventive step exists. Non-technical features (such as business logic or aesthetic choices) cannot support inventive step, even if they are novel.
| EPO Concept | What It Means | Practical Effect |
|---|---|---|
| Technical character | The invention must have a technical aspect -- it must relate to a technical field, solve a technical problem, or produce a technical effect | A computer program running on hardware has technical character by default |
| "As such" exclusion | The Article 52(2) list only excludes subject matter that has no technical aspect at all | Pure business methods, pure math, and pure aesthetic creations are excluded; applied versions often are not |
| COMVIK approach | When assessing inventive step, only features with technical character count | Novel business logic implemented on a computer does not contribute to inventive step; only the technical implementation does |
| Technical problem / technical solution | The invention must solve a problem of a technical nature using technical means | A faster algorithm for data compression solves a technical problem; a new way to display advertisements may not |
Key Difference from the U.S.: The EPO does not have a threshold eligibility filter like Alice/Mayo. Instead, it folds eligibility analysis into the inventive step assessment. An invention with any technical character passes the "is it an invention?" threshold. The real test is whether the technical features, considered alone, involve an inventive step over the prior art. This approach is generally considered more predictable than the U.S. framework, though it can be strict: if all of the novelty resides in non-technical features (pure business logic, for instance), the claim will fail for lack of inventive step even if the overall system is technically implemented.
Software Patents: A Global Comparison
Software patentability is the area where jurisdictional differences are most pronounced. The same invention can be straightforwardly patentable in one country and firmly excluded in another, depending on how the claim is drafted and which office examines it.
| Jurisdiction | Approach | Key Standard | Practical Outcome |
|---|---|---|---|
| United States | Permissive but with Alice limits | Must not be directed to an abstract idea; if it is, must show practical application or "significantly more" | Many software patents survive if claims are drafted to emphasize technical improvement; abstract business logic on generic hardware is rejected |
| EPO (Europe) | Technical character required | Software must solve a technical problem using technical means; pure business logic excluded from inventive step | Computer-implemented inventions are widely patented; the technical contribution must be clearly articulated |
| China (CNIPA) | Increasingly open | Must provide a "technical solution to a technical problem" using "technical means" | Software patents are common and growing; CNIPA guidelines have expanded eligibility in recent revisions |
| Japan (JPO) | Relatively permissive | "Creation of technical ideas utilizing the laws of nature" | Software-related inventions are broadly patentable if they use hardware resources; problem-solution based assessment |
| South Korea (KIPO) | Similar to Japan | Must involve technical ideas utilizing natural laws; assessed through a problem-solution lens | Broadly permissive for software with technical effect; follows JPO methodology closely |
Practical Drafting Guidance for Software Claims
Software claims that survive examination across jurisdictions share common traits: they recite a specific technical improvement (not a generic business outcome), identify the technical problem being solved, avoid purely functional language ("a system that optimizes X" without specifying how), and tie the software to a concrete technical context with measurable effects. Claims framed around specific algorithmic steps, data structures, or hardware interactions are more resilient than those describing outcomes in abstract terms.
Business Method Patents
Business method patents have faced skepticism in most major jurisdictions. The core question is whether a new way of conducting business -- pricing strategies, financial instruments, organizational processes -- qualifies as a patentable invention. In the U.S., business methods are restricted post-Bilski and Alice: abstract business methods are not patentable, and implementing them on a generic computer does not change the analysis. At the EPO, methods for doing business are on the explicit Article 52(2) exclusion list, and business logic cannot contribute to inventive step even when implemented on a computer. Japan and China take a more permissive approach, allowing business methods that use hardware resources to produce a genuine technical result -- CNIPA's 2020 guideline revision expanded eligibility for business method inventions containing "technical features."
In practice, pure business method patents are difficult to obtain anywhere. Even in the U.S., where they were once readily granted after the 1998 State Street Bank decision (effectively overruled by Alice), business method claims now face heavy scrutiny. The path to patentability requires demonstrating genuine technical innovation -- not merely automating a known business practice on a computer.
AI and Machine Learning Inventions
Artificial intelligence presents some of the most challenging eligibility questions in modern patent law. The technology sits at the intersection of mathematical methods (which are traditionally excluded), software (which has jurisdiction-specific rules), and practical applications that can produce clear technical results.
What Is Patentable in AI
AI-related patent claims generally fall into three categories, each with different eligibility outcomes:
AI applied to a technical problem. A machine learning model trained to detect cancerous cells in medical images, or a neural network that optimizes semiconductor manufacturing processes, applies AI to solve a concrete technical problem. These claims are generally eligible across all major jurisdictions.
AI training methods and architectures. A novel training algorithm, a new neural network architecture, or an improved method for reducing training data requirements can be patentable if the claims articulate a technical improvement. The USPTO's 2024 AI Subject Matter Eligibility Update provided three new examples (Examples 47-49) illustrating when AI training methods are eligible.
AI models and trained parameters. A trained model itself -- the specific weights and parameters -- is more difficult to patent. It resembles a mathematical model (excluded as such) and raises questions about reproducibility and enablement. Most offices treat trained models as unpatentable in isolation but will consider claims directed to systems or methods that use a trained model to achieve a specific technical result.
The AI Inventor Question: DABUS Cases
One of the most high-profile AI patent debates has been whether an AI system can be named as the inventor on a patent application. Dr. Stephen Thaler filed patent applications worldwide naming DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) as the sole inventor.
| Jurisdiction | DABUS Outcome | Reasoning |
|---|---|---|
| United States | Rejected | Federal Circuit held that "inventor" under the Patent Act must be a natural person (Thaler v. Vidal, 2022) |
| United Kingdom | Rejected | Supreme Court held that an inventor must be a natural person under the Patents Act 1977 (Thaler v. Comptroller, 2023) |
| European Patent Office | Rejected | The Boards of Appeal confirmed that the EPC requires a natural person as inventor |
| Australia | Initially accepted, then overturned | The Federal Court initially allowed AI inventorship, but the Full Court reversed, requiring a human inventor |
| South Africa | Granted | Patent granted with DABUS as inventor; South Africa does not substantively examine patents, so no inventorship analysis was performed |
Current Status of AI Inventorship: As of 2026, no major examining patent office recognizes AI as an inventor. The consensus across the U.S., UK, EPO, Japan, South Korea, and China is that an inventor must be a natural person. However, this does not prevent patenting AI-generated inventions -- it means a human who directed, supervised, or contributed to the inventive process must be named as the inventor. The legal and policy debate continues, with several jurisdictions actively studying whether legislative changes are warranted.
Drafting Strategies for AI Claims
Successful AI patent applications focus on the technical application rather than the underlying math, describe training processes in technical terms (specifying methodology, data structure, and architecture choices), include hardware and system context to avoid abstraction objections, and document human contributions to the inventive process to satisfy inventorship requirements.
Biotechnology and Life Sciences
Biotechnology sits at the boundary between discovery and invention. The central tension: biological materials exist in nature, yet isolating, modifying, and applying them requires genuine inventive effort and substantial investment.
Gene Patents After Myriad
The U.S. Supreme Court's 2013 decision in Association for Molecular Pathology v. Myriad Genetics established that naturally occurring DNA segments are not patentable -- the Court held that merely isolating the BRCA1 and BRCA2 genes associated with breast cancer is a discovery, not an invention. However, synthetic cDNA (complementary DNA created from mRNA, lacking natural non-coding introns) remains patent-eligible. Claims to synthetic constructs, engineered proteins, modified organisms, and methods of using genetic information all survive Myriad.
The EPO Approach: Biotech Directive
The EPO's approach is governed by the Biotech Directive (98/44/EC), implemented through Rule 26-29 EPC. The EPO permits patents on biological material isolated from its natural environment (even if previously existing in nature), plants and animals produced by technical processes, and gene sequences with a disclosed industrial application. The EPO also applies morality exclusions under Article 53(a) EPC, prohibiting patents whose commercial exploitation would be contrary to public order or morality -- applied notably to human embryonic stem cell research (the WARF decision, G 2/06).
| Biotech Area | United States | EPO (Europe) | Key Difference |
|---|---|---|---|
| Naturally occurring gene sequences | Not patentable (Myriad, 2013) | Patentable if isolated and industrial application disclosed | The EPO permits isolation-based claims the U.S. does not |
| Synthetic DNA / cDNA | Patentable | Patentable | Both jurisdictions agree synthetic constructs are eligible |
| Diagnostic methods | Patentable (but face Alice/Mayo scrutiny for natural correlations) | Excluded if practiced on the human body (Art. 53(c) EPC); in vitro diagnostics are patentable | EPO excludes methods "practiced on the human body"; U.S. applies the Mayo two-step |
| Stem cells | Generally patentable (no morality exclusion in U.S. law) | Human embryonic stem cells requiring destruction of embryos are excluded (WARF decision) | EPO applies a morality filter the U.S. does not |
| Plants and animals | Patentable through utility patents (especially GMOs) | Not patentable if produced by "essentially biological processes"; technical processes are patentable | EPO excludes conventional breeding results; U.S. does not draw this distinction |
Diagnostic Method Claims: Diagnostic methods face eligibility challenges in both the U.S. and Europe, though for different reasons. In the U.S., diagnostic claims that correlate a natural biomarker with a disease state are often rejected under Mayo as directed to a law of nature. At the EPO, diagnostic methods practiced on the human body are excluded under Article 53(c) EPC. In both jurisdictions, careful claim drafting -- focusing on in vitro steps, specific technical measurement techniques, or therapeutic applications -- can navigate these exclusions.
Standard-Essential Patents and FRAND
A standard-essential patent (SEP) is a patent that claims technology essential to implementing an industry standard. If a manufacturer cannot comply with a technical standard without using the patented technology, the patent is standard-essential.
How SEPs Arise
Standards-setting organizations (SSOs) -- such as ETSI, IEEE, and ITU -- develop technical standards that enable interoperability. When a company contributes its patented technology to a standard and that technology is adopted, the resulting patents become standard-essential. The 4G LTE and 5G standards incorporate thousands of patented technologies from companies including Qualcomm, Nokia, Ericsson, Huawei, and Samsung. Any manufacturer of a standards-compliant device must use these patented technologies.
The FRAND Commitment
To prevent abuse of this market power, SSOs require patent holders to commit to licensing their SEPs on Fair, Reasonable, and Non-Discriminatory (FRAND) terms:
- Fair and Reasonable: The royalty rate must reflect the actual value of the patented technology, not the leverage from inclusion in the standard.
- Non-Discriminatory: The patent holder must offer comparable license terms to all willing licensees.
FRAND commitments constrain how licensing fees are set, not whether they exist. Determining what constitutes a "fair and reasonable" royalty has been the subject of extensive litigation across the U.S., UK, Germany, and China.
The lifecycle of a standard-essential patent and FRAND licensing
SEPs and Eligibility
SEPs must satisfy the same patentability requirements as any other patent. Because they cover technology essential to technical standards, they typically have clear technical character and rarely face eligibility challenges. However, a patent's claim scope determines both whether it is truly essential to the standard and the FRAND royalty it commands -- creating a direct link between eligibility analysis and commercial value.
Approximate Distribution of Declared SEPs by Technology Area (%)
Chapter 21 provides a comprehensive treatment of SEP licensing, FRAND rate-setting methodologies, global litigation strategies, and the regulatory frameworks emerging around SEPs in the EU, U.S., and China.
Frequently Asked Questions
Eligibility Basics
Software and AI
Biotech and Specialized Areas
What's Next
You now understand which categories of innovation are eligible for patent protection and how major jurisdictions draw the line. Chapter 5 turns to the practical skill that underpins every patent strategy: searching for prior art -- how to find what already exists, assess whether your invention is truly new, and use search results to strengthen your application before filing.