/Pre-execution Governance Control Layer For Managing Artificial Intelligence System Actions
Abstract

A method, system, and non-transitory computer-readable medium govern execution requests generated by an artificial intelligence system. The method includes intercepting, by a control layer positioned between an artificial intelligence system and an execution system, an execution request generated from an output of the artificial intelligence system, and storing the execution request in a holding state inaccessible to the execution system, wherein the execution system is configured with a default-deny execution posture rejecting execution requests lacking a valid authorization artifact. The method includes classifying the execution request based on governance-relevant attributes to identify a governance handling category, assigning a discrete execution state by applying governance logic, generating an immutable governance record prior to authorization, and generating or withholding the authorization artifact based on the discrete execution state. The method routes deferred requests to an oversight system and enforces machine-enforceable modification constraints.

Full Text

What is claimed is:

A method, system, and non-transitory computer-readable medium govern execution requests generated by an artificial intelligence system. The method includes intercepting, by a control layer positioned between an artificial intelligence system and an execution system, an execution request generated from an output of the artificial intelligence system, and storing the execution request in a holding state inaccessible to the execution system, wherein the execution system is configured with a default-deny execution posture rejecting execution requests lacking a valid authorization artifact. The method includes classifying the execution request based on governance-relevant attributes to identify a governance handling category, assigning a discrete execution state by applying governance logic, generating an immutable governance record prior to authorization, and generating or withholding the authorization artifact based on the discrete execution state. The method routes deferred requests to an oversight system and enforces machine-enforceable modification constraints.
Timeline
Filed
06/08/2026
Published
09/24/2026
Granted
Not Available
IPC Codes(1)
G06N 20/00:Machine learning