A state-level AI statute notable for the drafting choices that narrowed it substantially between introduction and enactment. The version passed turns on intent: it addresses development or deployment of a system with the intent to discriminate against a protected class, rather than reaching disparate outcomes irrespective of intent. That single choice determines whether compliance is a policy exercise or a testing programme.
Enforcement is reserved to the state Attorney General, with a cure period before penalties attach, and there is no private right of action. For most organisations this affects expected exposure more than any substantive provision.
The Act also provides that documented alignment with a recognised risk management framework can support a defence, which gives voluntary framework adoption a legal dimension it does not otherwise have.