MUMBAI, India, Feb. 27 -- Intellectual Property India has published a patent application (202641018726 A) filed by Vellore Institute Of Technology, Vellore, Tamil Nadu, on Feb. 19, for 'a system and method for detecting prompt injection in multilingual inputs to large language models.'
Inventor(s) include Dr. Jagadeesan S; and Silla Ann Regi.
The application for the patent was published on Feb. 27, under issue no. 09/2026.
According to the abstract released by the Intellectual Property India: "The present invention relates to a system (100) for detecting prompt injection attacks in multilingual textual inputs provided to large language model-based interfaces. The system (100) comprises a computing device (102) having a processor (104) and memory (106) configured to execute plurality of modules (108). An input processing module (110) receives and preprocesses multilingual or code-mixed textual inputs. A rule-based detection module (112) analyzes the inputs using predefined linguistic pattern libraries to identify explicit injection indicators, while a model-based analysis module (114) evaluates contextual characteristics. A decision fusion module (118) combines outputs from the rule-based and model-based modules (112, 114) to generate a classification outcome indicating a prompt injection attempt. The classification outcome is used to permit, restrict, or flag inputs prior to execution by the large language model, thereby enhancing security, reliability, and real-time robustness in multilingual application environments."
Disclaimer: Curated by HT Syndication.
