The landscape of enterprise artificial intelligence took a major step forward with the formal issuance of U.S. Patent No. 12,664,368. Titled “Dynamically generated LLM request package to generate a human-readable response comprising sensitive data to a user input,” this groundbreaking intellectual property was awarded to Playback Health, Inc., a leader in AI-driven ambient clinical documentation. The patent details an advanced methodology for leveraging the power of large language models while overcoming the rigorous security obstacles inherent in handling highly confidential data environments.
In recognition of its profound technical impact and immediate real-world utility, this invention has officially been named the Patent of the Month for New York State for July 2026. By solving the delicate tension between utilizing powerful cloud-based AI models and maintaining absolute data privacy, the patented framework provides healthcare systems and enterprise organizations with a validated blueprint for secure automation. This accolade highlights how the New York technology sector continues to drive compliant, high-stakes AI infrastructure solutions that protect consumer and patient privacy.
Why the Dynamic LLM Request Package Invention is So Innovative
The core innovation of this patent lies in its unique architectural approach to data protection during live AI processing. Instead of sending raw text or sensitive patient information directly to a large language model, the technology dynamically constructs a specialized request package. This package sanitizes and separates sensitive identifier fields from the core contextual input before transmission, utilizing tokenized placeholders or local preprocessing engines. Once the LLM generates a secure response, the local application seamlessly reconstructs a human-readable output containing the sensitive data. This completely mitigates the risk of protected health information (PHI) leaks or exposure to third-party model training sets, allowing clinicians to utilize lightning-fast ambient scribing and automated note compilation without sacrificing compliance or security standards.
U.S. R&D Tax Credit Eligibility for Practical Applications
To leverage the practical applications of this patent for the United States Research and Development (R&D) Tax Credit under Internal Revenue Code Section 41, an organization must demonstrate that its development activities satisfy the statutory four-part test. The engineering work required to design the dynamic packaging algorithms, integrate the framework seamlessly into Electronic Health Records (EHR) platforms, and optimize real-time tokenization mechanisms directly qualifies as a process of experimentation. This work addresses profound technical uncertainty using principles of computer science and software engineering. By systematically documenting the iterative cycles of design, prototyping, and performance testing for this secure LLM pipeline, a company can successfully qualify research expenses. These eligible expenses typically include internal developer wages, contractor fees, and cloud computing expenditures incurred during the testing phases of the technology.
