Execution Gates
Answers the question: what stops an AI output being used too early?
What this shows
Reliance and destination, not the prompt alone, decide whether output proceeds. Firms can encode separation between generation and reliance at the point of action.
Where this would sit in a firm
Risk and supervision design, last-mile liability discussions, and any workflow where client-facing or filing-bound work needs explicit approval paths.
Demo only
Gate outcomes use fixed rules in this site’s code, not your policies or live systems. Not legal advice.
What the rules illustrate
- Risk changes when an output is used, not only when it is generated.
- A summary in a workspace is lower risk than advice sent to a client; gates encode that distinction.
- Governance belongs at the point of use: destination and reliance, not the prompt alone.