ADAM: the controlled engine for decisions that change the system
ENTITY describes identities and claims. ADAM determines whether a requested action is permitted and commits an accountable change to the relevant system state. A state is simply the system's current agreed record: a licence is active, an asset version is qualified, a research budget is approved or a transaction request is on hold.
To understand the difference, imagine a business keeps a secure employee-access register. ENTITY identifies an employee and their job permissions. ADAM is the gate that checks a particular request and either permits a recorded change or refuses it. The fact that the employee once had permission does not mean that every later request is automatically allowed.
A NIKI agent might request that a completed experiment be retained. ADAM checks the appropriate authorization, the exact asset version, the owner's current rules and whether the record has changed since the request was prepared. If valid, the authorized domain makes the change and records its receipt. If stale, disputed or unauthorized, it rejects or holds the command. An attempted replay of a completed transaction must not create a second payment or second issuance merely because a network message was sent twice.
ADAM is central to autonomous systems participating in real business activity. Intelligence can recommend an action; a governed state engine decides whether that recommendation can produce a real side effect. That is safer than handing an AI agent unrestricted access to a company's databases, funds and legal records.
ADAM is not the registrar of securities, a bank, a judge or a substitute for the authorization of a human rights holder. It enforces rules backed by legitimate authority within its scope.