When AI Decides Who Gets the Job (or the Loan)

Not all AI decisions are equal. A model suggesting email subject lines and a model screening job applicants live in different moral universes. Lawmakers have started drawing that line explicitly, using the idea of "consequential decisions": choices about employment, lending, housing, education, and health care.

Colorado passed the first comprehensive state law aimed at this category. Its core demand is refreshingly plain: if your AI helps make consequential decisions, assess it for discrimination, tell people it's involved, and give them a way to appeal. No exotic technology required — just diligence, documentation, and a human path for recourse.

Even if your state hasn't passed its version yet, the direction is obvious. Treat any AI that affects someone's livelihood as high-stakes: test it, document it, disclose it, and keep a human in reach.

I explain frameworks; I don't give legal counsel.