Three Strategies Help Overcome Rejected AI Patent Applications

Navigating patent eligibility under 35 USC § 101 is challenging for artificial intelligence inventions, which are often rejected as “abstract ideas.”

A recent executive order prompted the US Patent and Trademark Office to publish a new guideline in July 2024 that provides eligibility examples. Overcoming Section 101 rejections often involves amending the claims. Legal arguments alone are rarely sufficient.

Drawing from recent examiner interviews and experience since July, there are strategies for overcoming Section 101 rejections by demonstrating practical applications, technical improvements, or tasks impractical for humans.

Demonstrating Integration

Demonstrating integration into practical application is usually the most effective in overcoming Section 101 rejections.

For instance, Example 47 from the USPTO’s July 2024 guidance describes an AI model for network intrusion detection. Using an artificial neural network, or ANN, the model not only identifies network anomalies but also blocks malicious traffic in real-time without manual intervention.

This practical application strengthens eligibility by demonstrating a practical, immediate application of AI that enhances network functionality and contributes directly to system security.

The case of Ex Parte Jere Armas also provides a useful example. In this case, the applicant used a reinforcement learning model to generate optimized product recommendations. Initially, the claims were rejected as abstract; however, on appeal, the applicant argued that the claim limitations worked together to create a practical application, a process that enhanced the likelihood of successful product recommendations.

The Patent Trial and Appeal Board found this integration sufficient to overcome the Section 101 rejection, recognizing that the claim elements interacted meaningfully to achieve a technical outcome beyond data analysis.

This case illustrates the benefit of presenting claims as solutions to defined, real-world issues rather than as standalone computational models.

Technical Improvements

In addition to showing practical applications, applicants should demonstrate specific technical improvements achieved through the claimed invention. This approach can be effective when applicants can articulate measurable advances over existing methods.

In Example 48 from the USPTO guidance, a deep neural network improves speech separation for real-time transcription. The DNN’s clustering and masking techniques enhance audio clarity by distinguishing voices, reducing background noise without speaker-specific training, and advancing speech recognition technology beyond generic transcription.

Similarly, the decision in Ex Parte Andrew B. Covit was decided on technical improvement arguments in AI patent eligibility. This case involved a system that automatically assigned medical codes to unstructured data, a task traditionally handled by coders. The applicant successfully argued that the system’s improved accuracy and efficiency directly benefited the medical coding process, advancing healthcare data processing, thereby meeting the Section 101 requirements.

In Ex Parte Murray A. Reicher, Flora Gilboa-Solomon, and Guy Amit, the claims were directed to an automated image review and diagnosis system. The PTAB found that the claims lacked a clear technical improvement in computer functionality and merely recited abstract steps.

Further, the impracticality argument is most compelling when the AI model’s application involves large data sets, complex pattern recognition, or rapid processing beyond human capability and execution. An examiner in early 2024 suggested amending the claims under this rationale, but the success rate of this approach has been low.

Patent Drafting Tips

In AI patent claims, specifying hardware elements can strengthen eligibility under Section 101 by grounding the invention in tangible technology. Rather than vague terms such as “computing devices,” claims should recite the algorithm interacts with hardware to improve efficiency or enable unique functions.

For instance, an algorithm might optimize hardware by reducing computational load or speeding up data processing. Detailing specialized hardware, such as ASICs or GPUs, and explaining claim element interactions further supports the practical application. The specification should describe specific hardware (e.g., NPU, HBM, etc.) for implementing algorithms.

Long legal arguments without claim amendment rarely work. When responding to office actions involving Section 101 rejections, applicants should amend the claim to demonstrate integration into practical applications, technology improvement, or performance of tasks impractical for humans.

For amendments relying on integration into practical applications, start by citing MPEP 2106.04(d). Applicants can emphasize how the invention applies abstract AI functions within a specific, real-world context. Adding claim limitations that directly connect the algorithm’s output to hardware functionality is often effective.

Technical improvement arguments should focus on specific, measurable enhancements brought by the AI model. Examiners are more receptive to claims that show the AI’s real-world impact, such as faster processing times or improved computer efficiency.
Highlighting an AI model’s ability to handle tasks impractical for humans can be an effective support strategy, particularly in time-sensitive applications.

The path to Section 101 compliance for AI patents can be complex, but the USPTO’s recent guidance provides a framework for strengthening AI claims. As AI technology continues to evolve, so too will interpretations of Section 101, making it essential for applicants to stay informed on case law and revisit their approaches accordingly.

This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law and Bloomberg Tax, or its owners.

Author Information

Lu Yin is partner at Quarles & Brady and helps US businesses navigate China’s patent landscape.

Write for Us: Author Guidelines