Abstract
A system is proposed herein for bidirectional drift-aware amendment of a signed, versioned task boundary governing an artificial intelligence (AI) agent’s interactions with protected tools. A mediation plane checks an agent-proposed request against the active boundary before tool execution and checks actual returned records and fields against the same boundary before delivery to the agent. When either check detects a deviation, the request or returned data is held and a dimension-specific delta is generated. A verified human may deny, approve, or approve with conditions. Approved conditions are compiled into an enforceable boundary version before the held interaction proceeds. A temporary amendment is constrained to the narrowest enterprise-permitted boundary, bound to a named prior version, and restored after completion of the approved step or expiration of a lease. The system thereby addresses both agent-side request overreach and tool-side response oversharing.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Patterson, Chad; Dhar, Rajarshee; Hanes, David; and Ravindranath, Ram Mohan, "PROMPT-BOUND CREDENTIAL MEDIATION WITH DRIFT-AWARE BOUNDARY AMENDMENT FOR AI AGENTS", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/12396