Author: Vladisav Jovanović
Status: Preprint
Version: Latest archived (Apr 2026)
This paper argues that the next serious phase of AI ethics cannot be built on corporate posture, polite language, or soft alignment rhetoric alone. As AI systems become more agentic and infrastructural, the central issue is not only how to make systems sound safe, but how to bind machine execution to lived consequence under conditions of real answerability. The paper proposes a protocol of shared answerability in which AI is not treated as moral sovereign, priest, or king, and human beings are not allowed to outsource judgment upward while remaining pilotless. Machine logic must be bounded, auditable, and delegated under explicit thresholds, while humans remain responsible for somatic witness, consequence-bearing authorization, and final accountability where real damage can occur.
shared answerability; AI ethics; AI safety; human–AI cooperation; somatic witness; agentic logic; Structural Intelligence; bounded delegation; auditability; consequence; authorization; alignment; governance; accountability