What question did this study set out to answer?

To determine if AI decisions can be reliably re-run for audit purposes.

July 13, 2026Open Access

Can You Re-Run an AI Decision and Get the Same Result for an Audit?

MIMicky IronsMaryland Institute College of Art

Key Points

To determine if AI decisions can be reliably re-run for audit purposes.
Examined AI models for their ability to reproduce past decisions under various conditions.
Discussed the importance of pinned versions, recorded inputs, and fixed seeds.
Utilized a Sovereign Intelligence Operating System for audit verification.
Only systems with specific features (pinned versions, fixed seeds) can reproduce past results reliably.
Public cloud AI services face challenges with model and routing drift, affecting decision reproducibility.
The Open Audit Record system ensures every action is securely signed and verifiable.

Abstract

Only if the system was built for it. A public cloud AI service cannot reproduce a past decision because the model and routing drift. Pinned versions, recorded inputs, fixed seeds and a sealed build let an operator re-run and verify.Originally published at https://mickai.co.uk/articles/deterministic-inference-can-you-re-run-an-ai-decision-and-get-the-same-result. Mickai is a Sovereign Intelligence Operating System; the Open Audit Record signs every artificial intelligence action before it executes, post-quantum and offline-verifiable. 104 filed UK patent applications, Mickai LTD (Companies House 17166618).

Ask AI

Helpful

Bookmark

View Full Paper