What can remain unchanged
The component's objective
The system continues to pursue the same local goal or respond to the same underlying motive.
A dissociation
Alignability ≠ value alignment
A system can be highly steerable without wanting what the steerer wants.
Redirecting behavior does not require rewriting the component's values. Sometimes the surrounding environment supplies enough leverage to make the same underlying objective produce different behavior.
≠
What can still change
Its behavior
Changing prices, rules, constraints, permissions, or dependencies can make different actions locally advantageous or possible.
A simple example
A firm can oppose a tax and still respond to it.
Imagine a profit-seeking firm that would prefer not to face an emissions tax. If the tax is nevertheless imposed, reducing emissions can become the more profitable response.
The firm's environmental values need not have changed. What changed was the relationship between its existing objective and the available actions.
What this reveals
Control can live in dependencies rather than agreement.
If a component depends on resources, prices, permissions, contracts, or other environmental conditions, those dependencies can become handles for redirecting competent behavior.
This creates an alignment architecture quite different from trying to make every component internally want the system-level outcome.
Try this elsewhere