SYNCLOGICSL-RC-001
How should AI-output reliability be classified?
Reliance Classes separate raw generation from increasingly governed uses. The required evidence, verification and human authority rise with consequence.
AuthorWalter Shepherd
Published26 July 2026
Updated26 July 2026
Version2.1.0
StatusPublic Research and Pilot Release
MethodSL-RC-001
Failure mode
What this method prevents
An exploratory AI answer is copied into a board, regulatory or safety decision without upgrading its governance burden.
Illustrative example
A broad research summary may be adequate for orientation but not for a submission, procurement decision or irreversible action.
How it works
A controlled sequence
- Declare intended use before analysis.
- Assign the target Reliance Class.
- Apply the required controls and evidence standard.
- Verify the complete governance chain.
- Confirm the achieved class, which may be lower than the target.
- State conditions and prohibited uses.
Reliance note
The method supports governance; it does not self-authorise a decision.
Application must be proportionate to the task, evidence and consequence level. The output should state what has been registered, what remains uncertain and which authorised human or organisation controls release.