On 30 September, California Governor Gavin Newsom signed a package of laws requiring human review of AI-based dismissal and disciplinary decisions, notice when AI causes mass layoffs, and limits on workplace surveillance. Employers will also be barred from using workers' biometric data to infer their emotional state. Although framed as labour law, these rules address a classic constitutional problem: how can power exercised through automated systems remain accountable to the individuals it affects?

The mandatory human review of algorithmic dismissals can be read as a due process guarantee, securing the right to be heard by a human decision-maker before a decision with serious consequences is taken. The ban on emotion inference touches on human dignity and mental privacy. Both ideas are well established in European law:

Article 22 GDPR protects against solely automated decisions and was interpreted broadly by the CJEU in the SCHUFA case, while the AI Act prohibits emotion recognition in the workplace. Underlying them are the rights to dignity, data protection, fair working conditions and an effective remedy in the EU Charter of Fundamental Rights.

The case also highlights two structural questions. First, these safeguards apply between private parties, illustrating the growing horizontal effect of fundamental rights in the algorithmic society. Second, Newsom justified the laws by pointing to the absence of federal leadership, while Congress has considered proposals to pre-empt state AI laws. Who holds the power to regulate AI is thus becoming a question of federalism. The limits of the package are telling too: the Governor vetoed bills on AI in health care and mental health care, and signed only part of the union-backed slate.