A halt is not a verdict

A halt is an operational event. Execution stops when a condition is detected: a guardrail trips, a monitoring threshold is crossed, a human intervenes. The halt produces a record of its occurrence: a log entry, an alert, a timestamp. It does not produce a decision over a defined space of permitted actions. A halt says "this stopped." It does not say "this was not permitted, for these reasons, under this policy, in this context."

A verdict is a decision. It is reconstructable from policy, context, and proposed action specification. It has bounded scope, pertaining to a specific proposed action evaluated against a specific policy version in a specific context. It produces an authorization artifact: an evidence package sufficient for a third party to verify the decision without access to the system that made it. The verdict takes a value in the canonical verdict space {ALLOW, DENY, ABSTAIN}, with ABSTAIN as the third value.

These properties (reproducible logic, bounded scope, verifiable proof) are not optional features of an authorization verdict. They are what makes the verdict an authorization artifact in the ex-ante sense. A halt cannot be replayed deterministically from a defined governed state because its trigger is observational rather than structural. A halt cannot be appealed because there is no defined decision an appellant can contest. A halt cannot be audited as authorization because there is no policy version, no decision space, no authorization artifact to examine.

The structural claim follows. The ability to halt, block, interrupt, or escalate execution does not establish that the action was permitted before execution. Intervention may prevent an action from completing. It does not produce the artifact an ex-ante authorization regime requires.

This doctrine distinguishes FERZ from the categories most commonly conflated with authorization through their ability to interrupt execution: content guardrails, alerting systems with auto-block, runtime application security with intervention capability, and human-in-the-loop review queues. Each can stop an action. None produces a verdict over a defined decision space. They report that something stopped. Authorization produces evidence of whether the action was permitted: under what policy, in what context, and with what proof, verifiable independent of the system that proposed it.