Abstract
Autonomous artificial intelligence security agents assist security operations centers by evaluating security alerts. Unstructured memory mechanisms can cause agents to reuse outdated decisions when underlying conditions change. The present publication provides a memory framework that grants permission based on risk. The framework uses live data to verify conditions before acting. Memory entries include explicit conditions, direct pointers to live authoritative systems, permission limits, and temporal validity bounds. The system applies asymmetric verification logic. Memory operations that increase investigative scope proceed immediately. Operations that reduce work undergo live, non-cached verification of specified conditions against authoritative sources. A fail-closed decision model and independent behavioral invariant overrides safely revert anomalous contexts to standard triage.
KEYWORDS: Artificial Intelligence, Autonomous Agents, Security Operations Centers, Precondition Gating, Memory Authority, Risk-Asymmetric Verification, Fail-Closed Decision Logic, Zero-Cache Live Reads.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Raja, Vinayak; Agrawal, Sunil; Desai, Nitin; and , "Risk-Asymmetric Memory Authority and Deterministic Precondition Gating for Autonomous Artificial Intelligence Security Agents", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/12981