MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202641110406 A) filed by Vellore Institute Of Technology on September 15, 2026, for System And Method For Secure Authenticated Latent-State Communication Between Distributed Ai Models Using Cryptographic Pointer Exchange.
Inventors include Kalaiselvan N; and Shreejeet Chowdhury.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: ABSTRACT The present invention relates to a system and method for direct, secure, and efficient communication of machine-processable semantic states between heterogeneous artificial intelligence (AI) models. The system captures hidden-state representations from an originating AI model prior to text decoding, dynamically compresses the representations, and generates cryptographically authenticated structured pointers associated with the compressed representations. A Semantic Memory Broker stores and manages the representations, while the authenticated pointers are securely transmitted to a receiving AI model. The receiving AI model retrieves and verifies the corresponding representations and, where required, dynamically translates the representations into a compatible latent space using a Dynamic Latent Translation Engine. The aligned representations are directly injected into an execution-memory structure, including a key-value cache, thereby bypassing conventional text generation, tokenization, and embedding operations. The system further provides adaptive precision selection, multi-hop routing, replica-based fault recovery, semantic verification, and lifecycle cleanup, thereby reducing inference latency, computational overhead, memory utilization, and network bandwidth requirements in distributed AI systems.
Disclaimer: Curated by HT Syndication.
