In the continued development of the next patent eligibility policy EX PART DESJARDINSthe Patent Trial and Appeal Board (PTAB) reversed its denial of subject matter eligibility under 35 USC § 101 to Microsoft’s inventions directed to AI-assisted code editing tools such as GitHub Copilot. Here, Expert ZieglerAppeal No. 2025-003643, decided on May 29, 2026, the Board reversed the denial of eligibility in a 2-1 split decision, emphasizing its clear view of what constitutes an eligible claim for an invention in AI. Specifications made the difference in this decision.
Rejection based on the examiner’s typical “feasible to perform in the human mind” argument
Microsoft’s claims concerned a system for improving code prediction in programming editors, Application No. 17/740,164, filed May 9, 2022. Rather than restricting input to the code before the cursor as in traditional code language models, the claimed system assembles “additional context” from the text after the cursor, open editor tabs, associated code files, and metadata. Additionally, it is specified that the prompt is constructed by identifying the tree structure from its context before inputting it into the code language model. The examiners rejected the system claimed under § 101 as covering steps that can be taken in the human mind, and the board reversed, with Administrative Patent Judge Jeffrey S. Smith writing the majority opinion and Administrative Patent Judge Sharon Fennick dissenting.
The majority found support in the spec to claim improvements in code editor technology
Judge Smith found that the specification sufficiently described the limitations of prior art code editors and explained how the claimed invention overcomes those limitations by leveraging additional technical context to improve the predictions displayed to the programmer. The majority opinion concluded that the specification “provides sufficient detail to enable one skilled in the art to recognize that the claimed invention provides an improvement” to code editor technology, and then verified that the claim itself “reflects the disclosed improvement” through specific technical steps, including tree structure identification, sibling relationships, and object rearrangement, rather than simply invoking machine learning models common in code editing environments.
Opponents found the claim was simply applying general machine learning to new input data.
Judge Fennick applied. Recent Analytics, Inc. v. Fox Corp.134 F.4th 1205 (Fed. Cir. 2025), characterizes this claim as simply the application of “machine learning to code editing using apparently new input.” Judge Fennick also cited the lack of precedent. Rensselaer Polytechnic Institute v. Amazon.com, Inc. (Fed. Cir. 2026)[g]The holistic use of AI without other parameters, such as improving mathematical algorithms or improving machine learning, is abstract. ” The dissent read Microsoft’s argument that using “additional context” as input amounted to an improvement in code language modeling and was an “admission that no improvement to the machine learning model was claimed,” but merely “the use of a different input data source to a general-purpose code language model.” Judge Fennick would therefore have found that the claimed improvement did not incorporate an abstract idea into practical application and would have upheld the rejection.
Doctrinal disagreement: Improving model operation and general application to new data
While the majority opinion establishes that simply improving the technology provided by the machine learning model, in this case the code editor, is sufficient to integrate an abstract idea into a real-world application, the dissent takes the position that unless the claim improves the machine learning model itself, the claim fails. recently. This division reflects a broader framework that has emerged across the PTAB following the precedent framework. desjardins Decision No. 2024-000567, decided on September 26, 2025, found that claims that reflect “improvements in the way machine learning models themselves operate” may be eligible, but claims that “merely advocate the application of general-purpose machine learning to new data environments” are not.
This decision is Ziegler The generated code editing system that Microsoft claims is located at the exact boundary between these two locations. While the majority accepts eligibility as long as the claimed improvement to the way a machine learning model operates concerns a specific application of machine learning input in a relatively new data environment, as long as the specification sufficiently supports the improvement, the opposition remains critical of claims that the mathematical algorithm is not improved or the machine learning model itself is not improved.
Highlights of Litigation Regarding Obtaining Patent Eligible Inventions Under 35 USC § 101
1: Limitations of prior art in specifications
The majority opinion faulted the examiners for failing to address why “the specification’s disclosure of the use of additional context to overcome the limitations of prior art code editors does not account for improvements in the code editor art field.” This specification identifies prior art limitations on precursor text, explains how those limitations reduce the usefulness of predictions, and describes the claimed structure as a solution. This may have provided an important basis for the majority to find improvements in technology.
2: How the model behaves in context
Opponents argued that this claim merely changes “what is used as input to the code language model.” The majority opinion held that the claimed system overcomes this characterization because the claims further specify how the system structures the context, such as tree identification, sibling analysis, and object relocation, rather than just listing the data that is input to the model.
3: Technology improvements and business results
The majority opinion framed this improvement as being for “code editing technology” rather than a general invocation of hardware applied to abstract processes. This difference reflects: desjardins In this framework, the claimed improvement must be technical in nature, and any claim of improvement that can only be explained by reference to business benefits rather than technical advances is recently side of the watershed.
conclusion
of recently and desjardins This division will define patent examination and filing strategies at the U.S. Patent and Trademark Office for the foreseeable future. in view of Ziegler and other procedures thereafter. desjardinsThere are clear strategic advantages to (1) writing specifications that identify technical deficiencies in existing systems and describing the claimed invention as a solution to those deficiencies; (2) claiming how the system structures and processes context rather than simply listing new input data to a general model; and (3) framing claimed improvements as technological advances in related fields rather than as business results.
