Colorado repealed and rewrote its AI Act; Connecticut wrote one anyway
Senate Bill 26-189 replaced the 2024 Colorado Artificial Intelligence Act, stripping mandatory risk management programmes, annual impact assessments and the broad duty of care. Two weeks later Connecticut enacted a framework covering chatbots, synthetic media and automated decision-making. The states are diverging in real time.
Colorado's original act was the most demanding state AI law in the country and the model everyone else was measured against. The revision removes precisely the obligations that made it demanding — the risk management programme, the annual impact assessment, the general duty of care.
Read that as a data point about enforceability rather than politics. A state that wrote the strictest regime in the union looked at implementation and pulled back before it took effect, which is the same conclusion Brussels reached when it pushed high-risk conformity assessment out to December 2027.
Connecticut moved the other way, but narrower: targeted requirements on specific use cases rather than a horizontal risk framework. Use-case regulation is easier to enforce and easier to comply with, and it is emerging as the shape that actually survives contact with a legislature.
The courts are meanwhile deciding a prior question. In Garcia v. Character Technologies, a judge declined to hold that model output is speech, letting product liability and negligence claims proceed — a doctrinal opening that matters more than most statutes.
Wiley — Connecticut enacts AI framework while Colorado scales back landmark AI law → · IAPP — US state AI governance legislation tracker → · Lawfare — If AI outputs aren't speech, who has to prove they're human? →