A parks IoT AI edge backhaul needs WORM or compliance tiers for evidence that will appear in model audits. Which storage outcome should the design deliver?

A parks IoT AI edge backhaul that must keep evidence for model audits needs capacity planning that coexists with WORM or compliance tiers. Auditable municipal evidence cannot be treated as disposable scratch; immutability/retention controls preserve trust for legal and model-governance reviews while still leaving room for growing IoT corpora. Capacity and compliance are joint outcomes, not opposing choices.

Archiving auditable AI evidence to rewritable cold volumes with WORM/compliance tiers fails because rewritable volumes cannot enforce write-once immutability; without retention locks that block overwrite, the evidence can be altered and cannot be proven compliant for a model audit. Mirroring WORM/compliance datasets to high-performance caches for auditable AI evidence fails the audit requirement because caches optimize for speed and are mutable and evictable; a cached copy without WORM/retention locks can be overwritten or aged out before an audit needs it. Using erasure-coded pools with mutable retention for auditable WORM/compliance evidence fails because erasure coding protects against drive or node loss, not against alteration; mutable retention allows deletes or rewrites, so the evidence cannot later be proven unaltered.

Exam caveat: Domain 2.3 capacity planning includes protection and compliance-capable tiers when the stem invokes WORM/audits. WORM does not replace backups by itself, but the item's focus is retaining auditable AI evidence on compliant media as capacity grows.

Operational check: classify IoT evidence needing WORM/compliance; size compliant tiers for retention years plus growth; verify write-once/retention locks and audit exports before production models train on parks evidence.