Abstract
Generative artificial intelligence (genAI) technology enables users to specify natural language commands and obtain executable code and/or database queries as AI output. While such code can help users accomplish goals quickly, in enterprise computing environments, it can introduce operational issues/vulnerabilities, waste computational resources, lead to platform failures, and/or administrative friction. This disclosure describes a pre-execution orchestration architecture and dynamic user interface for user authorization of code prior to execution. A user can use genAI to generate code. Prior to execution, such code (executable code, database queries, or infrastructural commands) is subjected to analysis to generate pre-execution telemetry that includes factors such as computational cost and resources accessed/modified by the code. A user interface is provided with the pre-execution telemetry and is requested to authorize execution. Until the user approves, a rigid backend operational lock is placed to prevent execution. The user approval can be confirmed as an explicit, tokenized cryptographic authorization. The described mechanism provides verifiable, effective, and resource-efficient AI tools, and can boost trust in autonomous AI agents operating within a distributed computing infrastructure.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Hoyer, Timo, "User Authorization of AI Generated Code Based on Pre-execution Telemetry", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/11378