A public boundary statement for authority, responsibility, refusal, and execution-time control across Moral Clarity AI systems.
The system is a tool. Humans retain decision authority. Organizations retain accountability. Liability does not migrate into software.
If authority, responsibility, or liability cannot be clearly located for a use case, the system must not be deployed in that context.
Moral Clarity AI systems are tools. They are not legal subjects, autonomous agents, or bearers of rights, duties, interests, or liability. Governance obligations attach to natural persons and legal entities involved in design, deployment, and use.
Authority to make binding decisions remains with identified human actors and the deploying organization. The system provides assistance, does not exercise final discretion, and does not initiate irreversible actions independently.
Responsibility and liability for outputs, recommendations, and downstream effects remain with the deploying organization and designated human decision-makers. Liability is not transferred to software.
Refusal and non-action are mandatory capabilities. The system must halt, deny, or escalate when authority, state, or consequence boundaries are insufficient for legitimate continuation.
Operational constraints are defined externally by humans and enforced at execution time. No learned behavior, optimization objective, or internal system state may supersede these controls.
Material system actions should be logged to show who authorized use, who reviewed or approved outcomes, when refusals occurred, and why execution was permitted or denied.
Legitimate execution requires valid state, real authority, and a constrained action path. Where that basis fails, execution must halt, deny, or escalate.