Administered by the USPTO and governed by 35 U.S.C. §101, patent applicants must
prove their invention falls within four statutory categories (process, machine, manufacture,
composition of matter) and does not cover judicial exceptions: abstract ideas, laws of
nature, natural phenomena. The landmark Supreme Court Alice Corp. v. CLS Bank ruling
established the mandatory two-step eligibility framework, now codified in MPEP 2106
and updated with 2024 AI-specific examination guidance. Software, fintech algorithm
and artificial intelligence applicants frequently receive permanent §101 rejections due to
generic claim drafting that fails to integrate abstract logic into tangible technical improvements.
This case dissects a rejected AI anomaly detection patent, unpacks official examination
standards, and delivers drafting rules for global tech R&D teams.
Case Overview
A Singapore cybersecurity startup developed an AI neural network model to detect network
data anomalies and filed a utility patent application in late 2024. The independent claims only
recited generic algorithmic logic: collecting data points, running mathematical comparison
calculations, and outputting warning signals on a standard off-the-shelf computer. The
specification contained no description of custom hardware architecture, specialized neural
network training processes, or concrete improvements to network bandwidth or data processing
latency. During substantive examination, the USPTO examiner issued a final §101 subject-matter
eligibility rejection applying the Alice two-step test. Step 2A confirmed the claims were
solely directed to an abstract mathematical algorithm and mental data analysis
process, two statutorily excluded abstract idea groupings. The generic computer recited in the
claims was deemed conventional “off-the-shelf hardware” without inventive technical
limitations. Under Step 2B, the examiner ruled the application lacked an inventive concept
that transforms the abstract idea into a practical, improved technological application.
The startup’s attorney only submitted argumentative remarks without amending the
claims to add specific structural or process limitations. After two response rounds, the
application was abandoned, and the team lost exclusive U.S. protection for its core AI algorithm
design.
Core Legal & Procedural Insights
First, the full Alice two-step eligibility analysis is non-negotiable for all software/AI filings. Step 1
verifies whether the invention fits one of the four §101 statutory classes. Step 2A splits into two
prongs: identify whether claims recite a judicial exception (abstract idea, natural law), then
check if additional elements integrate the exception into a practical technical solution. If Step
2A fails, examiners proceed to Step 2B to evaluate whether the overall claim combination
supplies a unique inventive concept beyond routine generic computing. Second, three fixed
categories of abstract ideas trigger automatic high rejection risk: mathematical formulas/
calculations, human activity business logic, and standalone mental judgment processes.
Generic AI algorithm claims without custom hardware, optimized training pipelines or
measurable system improvements fall directly into this exclusion scope per the 2024
USPTO AI eligibility guidance. Third, generic computer hardware cannot cure abstract idea defects.
Merely reciting “a processor, memory, display” adds no patentable weight. Courts and
examiners uniformly hold that using ordinary computing equipment to execute abstract logic is a
conventional, well-understood step that cannot constitute a qualifying inventive concept under Step
2B. Fourth, specification disclosure must match claim technical limitations. To overcome
§101 rejections, the written description must detail concrete technical advantages:
reduced processing time, lower power consumption, customized chip circuit layouts, or
unique data filtering architectures. Vague “improved performance” language without
quantifiable technical metrics cannot support eligibility arguments. Fifth, post-rejection
remedy is limited to targeted claim amendments. Pure legal argument without rewriting claim
boundaries to add specific structural or process improvements will not overcome a final §101
office action; appeals to the PTAB rarely reverse eligibility findings if the original
specification lacks supporting technical disclosure.Practical Compliance Guidance for Global Tech
Enterprises
Draft independent claims around specific tangible technical improvements instead of standalone
algorithmic logic. Add limitations for custom ASIC chips, proprietary neural network training
sequences, unique data compression pipelines, or measurable latency reduction
thresholds to avoid pure abstract idea characterization. Align all specification content with claim
technical features: record quantitative performance benchmarks, hardware component structures,
and end-to-end technical workflow innovations to supply evidence for the inventive concept
required under Alice Step 2B. Classify claim elements clearly to avoid over-reliance on generic
computing terms. Replace vague references to “standard computer” with detailed specialized
hardware, dedicated sensor arrays, or modified network transmission modules that deliver unique
technological effects. Conduct a pre-filing §101 Alice test internal review following MPEP 2106
standards; separate abstract mathematical logic from the inventive technical application and ensure
the latter dominates claim scope. When receiving §101 eligibility rejections, prioritize claim
amendments that integrate physical machine limitations and concrete technical benefits, rather
than relying solely on written rebuttal arguments to examiners. Retain U.S. patent counsel specializing
in software and AI §101 prosecution to structure compliant claim sets before initial filing.Conclusion
35 U.S.C. §101 subject-matter eligibility compliance forms the absolute threshold for U.S. software
and AI patent allowance, governed rigidly by the Alice two-step judicial framework. This
abandoned cybersecurity AI patent case demonstrates that algorithm-only claims without integrated
tangible technical improvements will almost certainly face final eligibility rejection. For cross-border
artificial intelligence, fintech and software R&D teams, centering claims on measurable machine/
system innovations, avoiding isolated abstract algorithm recitations, and drafting detailed
supporting specification disclosure are mandatory steps to pass USPTO §101 examination and secure
enforceable exclusive patent rights within the U.S. market.
Hyperlink List：
● USPTO MPEP 2106 Full Subject-Matter Eligibility Guidelines (Alice Test Core Rules）：
https://www.uspto.gov/web/offices/pac/mpep/s2106.html
● USPTO 2024 Official AI & Software §101 Eligibility Guidance PDF:
https://www.uspto.gov/sites/default/files/documents/ai-sme-update-2024.pdf
