MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641054421 A) filed by Sri Thejas N; Pradeeswari R; Rathimalar G; Thenmozhi A; and Varshini J on April 29, 2026, for Evaluating The Unseen:a Risk Calibrated Artifact Framework.
Inventors include Sri Thejas N; Pradeeswari R; Rathimalar G; Thenmozhi A; and Varshini J.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: A zero-trust security framework for Artificial Intelligence pipelines is presented to protect against poisoned datasets, compromised models, and unauthorized modifications. The system performs rule-based and statistical inspection of datasets to compute a trust score that determines whether training is allowed, reviewed, or blocked. Five independent validation checks label consistency, duplicate detection, distribution analysis, outlier detection, and source validation collaboratively determine a risk classification of LOW, MODERATE, or HIGH. Cryptographic SHA- 256 integrity checks detect dataset tampering, and a recommendation engine generates context-aware security decisions. The framework outputs a security verdict, trust metrics, and audit logs, ensuring only trustworthy datasets enter the training pipeline.
Disclaimer: Curated by HT Syndication.
