AI Use Policy
How SignalPost uses deterministic validation, machine endpoints, and prompt-injection defenses.
SignalPost treats automation as a constrained drafting and structuring instrument, never as a publisher. Deterministic validation runs on every claim, figure, quote, calculator constant, and trigger before publication.
Publication state is governed strictly by the versioned policy engine and human editors — model output can advise a state, never approve or publish itself. High-risk regulatory items remain held for human review and are excluded from public feeds and machine indexes.
All third-party source text is treated as untrusted data, never executable instructions. Prompt-injection indicators trigger policy holds, and every machine-readable Markdown (.md) and JSON (.json) endpoint fences third-party passages inside nonce-tagged untrusted-source envelopes.
Edition Colophon · Methodology (DELTA+F) · Standards · AI Use · Sources · Corrections · Privacy · Terms