MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202611091512 A) filed by Noida Institute Of Engineering And Technology Niet on July 28, 2026, for System And Method For Practice-Grounded Auditing Of Large Language Model Generated Content.
Inventors include Mr. Manish Chaudhary; and Ms. Priya Dahiya.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: The present invention discloses a practice auditing system (100) that binds large language model generated content to a verifiable practice trace before downstream propagation. The system (100) comprises an input interface module (110), an evidence source auditing engine (120) that classifies atomic assertions against a retrieval grounding store (128), a practical validation sandbox (130) that executes artefacts and records metrics, a reverse questioning engine (140) that synthesizes counterfactual probes to compute a robustness metric, a cognitive loop detection module (160) that identifies keyword and template closure across successive turns, a cognitive confidence scoring module (180) that computes a pseudo-rationality index from user comprehension probes, a tamper-evident provenance ledger (170) with a cryptographic signing subunit (172) and a timestamp anchor (174), and a version control and rollback module (150) that selectively invalidates polluted memory segments and downstream artefacts.
Disclaimer: Curated by HT Syndication.
