MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202641100815 A) filed by Sisa Information Security Private Limited on August 20, 2026, for System And Method For Multi-Layer Detection Of System Prompt Drift In Artificial-Intelligence Models.
Inventors include Anjan Krishnamurthy; Renju Varghese Jolly; and Bharath M.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: The present invention discloses a computer-implemented system and a method for multi-layer detection of system prompt drift in an artificial-intelligence model. A prompt canonicalization subsystem (206) standardizes a system prompt into a canonical prompt representation. A structural rule extraction subsystem (208) extracts structural rules. A policy profiling subsystem (210) derives behavioral policies. An output contract extraction subsystem (212) determines an output contract representation. A baseline profile assembly subsystem (214) assembles a baseline profile. A prompt integrity evaluation subsystem (216) determines an integrity evaluation result from an active system prompt. An output contract validation subsystem (218) determines an output contract validation result from an artificial intelligence model response. An intent alignment evaluation subsystem (220) derives an intent alignment result based on behavioral compliance. A drift correlation subsystem (222) produces a correlated drift assessment. A drift severity classification subsystem (224) determines a system prompt drift classification according to severity categories. FIG. 2
Disclaimer: Curated by HT Syndication.
