Patent Eligibility · § 101
Software Patents After Alice
Alice Corp. v. CLS Bank (2014) changed what software claims must say — it didn't kill software patents. Claims that articulate a specific technical improvement to computer functionality still survive. Here's what passes and what fails the two-step test, with real case examples.
The post-Alice rule
Software claims must do more than take an abstract idea and say “do it on a computer.” They must claim a specific technical improvement to computer functionality — something that improves how the computer works, not just what the computer is used for. The more concrete and specific the technical steps, the stronger the claim.
Alice analysis in practice
What survives and what fails
Specific technical improvements to computer functionality
Claims that describe a specific algorithm that improves how a computer processes data — not just 'process it faster' but the specific technical steps that achieve the improvement.
Enfish LLC v. Microsoft Corp. (Fed. Cir. 2016) — self-referential database table claims survived; the improvement was specifically to how the computer functions, not to business outcomes.
Specific improvements to computer security
A specific algorithm for detecting and blocking a novel class of computer attacks, where the claims describe the technical mechanism of the detection.
BASCOM Global Internet Services v. AT&T (Fed. Cir. 2016) — filtering content at a remote server using individualized filters; the inventive concept was in how the filtering technology was arranged, not just that filtering occurred.
Specific solutions to technology-specific problems
Claims directed to a specific technical problem unique to digital environments — such as reducing latency in real-time streaming, reducing compression artifacts in a specific image format, or a specific method of error correction.
DDR Holdings v. Hotels.com (Fed. Cir. 2014) — retaining website visitors by using a specific technique to generate hybrid web pages; the solution was rooted in computer technology.
Organizing information through mathematical relationships
A method of categorizing financial transactions using an algorithm — the core of the claim is organizing information, which is an abstract idea. A computer merely executes it.
Alice Corp. v. CLS Bank (2014) — settlement risk intermediation using a computer; the idea of intermediation is abstract, and adding 'apply it on a computer' adds nothing inventive.
Collecting, storing, and displaying information
A method of collecting user data, storing it in a database, and displaying it in a report. The steps are generic data manipulation without a specific technical improvement.
Electric Power Group v. Alstom (Fed. Cir. 2016) — collecting power grid data and displaying it in real time; the claims amounted to 'nothing more than selecting information and presenting it using generic computer tools.'
Fundamental economic practices implemented digitally
A method of managing credit risk or insurance using a computer — the business concept existed long before computers; adding 'on a computer' doesn't make it patentable.
Bilski v. Kappos (2010) — hedging risk in commodities markets; the concept is a longstanding commercial practice, not a technological advance.
USPTO 2019 Revised Guidance
How the USPTO applies the Alice test
The 2019 Revised Guidance clarified Step 2A by breaking it into two prongs and gave examiners more specific instructions for when to find an abstract idea.
Step 1
Directed to a judicial exception?
Examine what the claim is 'directed to.' Three categories of abstract ideas under 2019 guidance: (1) Mathematical concepts — mathematical relationships, formulas, calculations, equations; (2) Certain methods of organizing human activity — fundamental economic principles, managing relationships/behavior, concepts performed in the mind; (3) Mental processes — concepts performed purely in the human mind. If the claim is not directed to any of these, it's eligible — no further analysis needed.
2A Prong 1
Practical application?
Even if directed to an abstract idea, does the claim integrate the abstract idea into a practical application? A practical application exists if: the claim reflects an improvement to the functioning of a computer or other technology; the claim applies the abstract idea with a particular machine or transformation; the additional elements impose meaningful limits on the abstract idea beyond applying it. If yes → eligible, stop. If no → continue.
2A Prong 2
Significantly more?
If no practical application found: do the additional elements (beyond the abstract idea itself) amount to 'significantly more' than the abstract idea? Generic computer hardware (processor, memory, display) performing their routine functions does NOT qualify. Specific technical improvements, non-generic hardware arrangements, or specific technical steps beyond what was routine in the field may qualify.
FAQ
Software patent questions
Can you still patent software after Alice Corp. v. CLS Bank?
Yes — software patents are still available after Alice Corp. v. CLS Bank International (Supreme Court 2014), though they require more careful drafting. Alice changed where the line falls, not whether software can be patented. Software claims survive when they: (1) claim a specific technical improvement to how a computer functions (not just applying an abstract idea on a computer); (2) solve a technical problem with a specific technical solution rooted in computer technology; (3) provide a specific, narrowly described algorithm that achieves a concrete result through defined technical steps. Software claims fail when they: (1) recite an abstract idea (mental process, mathematical calculation, fundamental economic practice) and merely implement it on a generic computer; (2) collect, store, or display data without a specific technical improvement; (3) use functional language that reads entirely on the abstract idea. In 2023, the USPTO allowed software-related patents — the agency has issued guidance helping applicants frame claims around technical improvements rather than abstract functionality.
What is the Alice two-step test for software patent eligibility?
The Alice/Mayo two-step test for patent eligibility under 35 U.S.C. § 101 works as follows: Step 1 — Are the claims directed to one of the judicial exceptions (abstract idea, law of nature, natural phenomenon)? Under the USPTO's 2019 Revised Guidance, abstract ideas are: mathematical concepts (formulas, algorithms, calculations), certain methods of organizing human activity (fundamental economic principles, managing relationships, concepts performed in the mind), and mental processes (concepts performed purely in the mind). If not directed to an exception, the claim is eligible — stop here. If directed to an exception: Step 2A Prong 2 (practical application) — Does the claim integrate the abstract idea into a practical application? If yes, the claim is eligible: examples include claims that improve the technical functioning of a computer or another technology, apply the abstract idea with a specific machine or transformation, or meaningfully limit the claim to a practical application. If no practical application: Step 2B (inventive concept) — Do the additional elements (beyond the abstract idea) amount to significantly more than the abstract idea itself? Generic computer components (processor, memory, display) performing generic functions do not qualify as significantly more.
How should software patent claims be drafted after Alice?
Drafting software claims that survive Alice requires focusing on the specific technical improvement, not the abstract outcome. Best practices: (1) Lead with the technical problem, not the business problem — if you're solving a latency issue in streaming, say so in the claim; if you're solving a financial reporting problem, the focus should be on how the computer achieves it technically, not the reporting outcome. (2) Use specific technical language — describe the specific algorithm, data structure, or process steps, not just 'process data to achieve X.' Claims that are too functional (i.e., read only on the abstract concept) fail Alice. (3) Include concrete limitations — a specific number of steps, a specific data structure, a specific architectural arrangement. The more specific and concrete the technical claim, the stronger the Alice defense. (4) Avoid claiming the result — 'a method that improves customer retention' is an abstract outcome; 'a method comprising generating a hybrid webpage by dynamically combining elements of a host website and a third-party website based on a stored set of parameters' is more concrete. (5) Claim hardware-software interactions specifically — if the software improvement works because of a specific interaction with hardware (memory management, parallel processing, real-time signal processing), make that interaction explicit in the claim. (6) Write a strong specification — document the technical problem and how your specific technical solution (not just the idea) solves it. Alice analysis often considers the specification to understand what the claim is 'directed to.'
Are Beauregard claims (software on a computer-readable medium) still valid?
Beauregard claims — claims directed to computer-readable media (non-transitory computer-readable medium) storing instructions that, when executed by a processor, perform specific steps — are still available and commonly used. The term 'non-transitory' is required to distinguish from transitory signals (electromagnetic waves, carrier waves), which are not patent-eligible. Beauregard claims survived Alice in the sense that they are a valid claim format; however, they still must pass the Alice two-step test for subject matter eligibility. A Beauregard claim that merely encodes an abstract idea (a mathematical algorithm) without additional inventive concept does not become eligible simply by claiming it on a computer-readable medium. The inquiry is the same: are the instructions that implement a specific technical improvement, or are they just implementing an abstract idea on a storage medium? The name 'Beauregard' comes from In re Beauregard (Fed. Cir. 1995), in which the USPTO rescinded its position that computer programs are per se non-statutory, and began allowing such claims.
What kinds of AI and machine learning inventions can be patented?
AI and machine learning inventions can be patented, but the claims must be drafted carefully to avoid Alice rejection. What tends to survive: (1) Specific AI architectures — a novel neural network structure with specific layers, connections, or activation functions that solve a defined technical problem; (2) Novel training methods — a specific technique for training a model more efficiently, with fewer parameters, or with better generalization that is described at the algorithmic level; (3) Application-specific AI systems — AI applied to a specific technical domain (medical imaging analysis using a defined algorithm, real-time control systems, autonomous vehicle perception pipelines) where the technical improvement is specific; (4) AI that improves the functioning of a computer system — e.g., an AI-based memory management system that specifically reduces cache misses in a defined way. What tends to fail: (1) AI applied to an abstract business outcome (AI for better customer engagement, AI for financial forecasting) without a specific technical mechanism; (2) Using a standard neural network to do something humans have always done (classify objects, recognize speech) — this may be anticipated or obvious, or treated as an abstract mental process. Drafting tip: for AI patents, describe the specific model architecture, training process, data representation, or inference mechanism in detail. The more the claim is anchored to a specific technical implementation rather than the general idea of using AI, the stronger the position.