CTI · Thesis research chapter

Trust in Programmable and Agentic Systems

OPEN RESEARCH · COMMUNITY TRUST INFRASTRUCTURE

AI as interpreter, not authority

AI may organize observations, identify inconsistencies, summarize evidence and prepare human review. It must not silently replace community confirmation or acquire governance authority by virtue of technical capability.

Agent execution rights must be explicitly delegated, bounded, auditable and revocable, and enforced in code. At small scale, manual stewardship may be more reliable and less costly than automation.

Return to the thesis overview

Related research