A safer model is not the same thing as a governed AI system.
Alignment can influence how a model tends to behave. It cannot tell an organization:
• Who used the model
• What data it accessed
• Which policy applied
• What tools it was allowed to call
• What decision was
