MUMBAI, India, Oct. 5 -- Intellectual Property India has published a patent application (202641114230 A) filed by Kalyan Kumar Duggineni on September 24, 2026, for Ai-Driven System And Method For Causal-Provenance-Based Trust Evaluation, Action-Specific Dynamic Authorization, And Selective Recovery In Autonomous Enterprise Integration Systems.
Inventor includes Kalyan Kumar Duggineni.
The application for the patent was published on October 02, 2026, under issue no. 40/2026.
Abstract: A computer-implemented system and method are disclosed for governing autonomous artificial-intelligence operations across multi-agent systems, Model Context Protocol (MCP) services, software tools, application programming interfaces (APIs), and distributed enterprise resources, using causal execution provenance, action- specific dynamic authorization, and selective recovery. An adaptive causal authorization engine constructs and dynamically updates a causal execution provenance graph whose nodes represent artificial-intelligence agents, delegated agents, MCP services, tools, APIs, enterprise services, and data resources, and whose edges represent invocation, delegation, dependency, data-flow, or causal relationships. For a proposed action, the engine identifies the causative provenance path that produced the action, derives a provenance-weighted effective trust state by propagating and attenuating trust conditions along that path, determines semantic compatibility between the proposed action and an authorized intent representation, and predicts a downstream consequence, or blast radius, using enterprise dependency relationships. These factors are combined into a multi-factor authorization vector from which the engine selects a progressive enforcement action — allow, allow with constraints, escalate for human review, isolate, or block — rather than a binary permit-or-deny decision, for the specific tuple of agent, action, transaction, and causal path. When a provenance path is found to carry contaminated or untrusted state, the engine identifies the contamination boundary and selectively holds, isolates, compensates, restores, reroutes, or replays only the causally affected branches, while unaffected branches continue unimpeded. Verified recovery outcomes and enforcement decisions are fed back into a continuous learning module that refines future trust propagation, authorization vector weightings, and enforcement thresholds.
Disclaimer: Curated by HT Syndication.
