Governance describes the structures through which an organisation constrains, authorises, controls and remains accountable for how automated AI systems operate. In the context of AI and automated decision systems, governance is not the technology itself and not general policy statements about technology.
It is the presence of enforceable constraints: who has authority to permit or refuse system actions, what the system is authorised to do within a defined scope, how those actions are controlled during operation and what accountability exists for outcomes. These structures must produce evidence showing that actions were constrained at the point they occurred. Without demonstrable constraints on system behaviour, governance cannot be said to exist.