MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641075847 A) filed by Vellore Institute Of Technology on June 19, 2026, for Judicial Entropy Engine For Detecting Hallucinated Legal Citations And Jurisdictional Doctrine Drift In Ai-Generated Text.
Inventors include A. Sivaranjani; P Priya Ponnuswamy; Rakshana B; Suvathi S; and Sujitha T.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: The present invention relates to a computer-implemented system and method for detecting hallucinated and untrustworthy legal citations in textual documents, particularly those generated or assisted by artificial intelligence systems. The system extracts legal citations from unstructured text using a citation extraction engine (102) applied to user input (101), and validates the extracted citations against a structured multi-jurisdictional legal precedent database (107) using a citation validation engine (103) with exact and similarity-based matching and adaptive confidence thresholds. Unverified citations are processed by a multi-signal legal analysis engine (104) configured to evaluate structural compliance, jurisdiction compatibility, temporal consistency, and doctrinal alignment, including Z-score based temporal anomaly detection and jurisdictional drift assessment. A topological quarantine engine (106) classifies citations into risk states including unverified, weak citation, and hallucination based on weighted features. A trust scoring engine (105) generates a normalized document-level trust score using machine learning-based anomaly detection and sigmoid scaling. The system output (108) provides citation verification status and hallucination risk assessment.
Disclaimer: Curated by HT Syndication.
