The honest view from inside technical leadership
AI News

AI News

Daily AI news with an engineering leadership take, each story credited to its original AI News report.

AI Governance · AI News

US States Enacted 280 AI Bills in 2026, and the Only Two Broad Laws Both Require Telling People When AI Is Involved

The National Conference of State Legislatures published the counts on October 21, 2026. The volume is the headline. The pattern underneath it is a single engineering requirement that keeps showing up in different statutes: notice.

Line-art illustration of a stack of state AI bills beside a speech bubble with an AI notice, representing 2026 state AI laws and their disclosure rules

US state legislatures introduced 1,539 AI bills in 2026 and enacted 280 of them by early October, according to figures the National Conference of State Legislatures (NCSL) presented at its Base Camp 2026 event on October 21. That compares with 1,262 state AI bills introduced in 2025. Only two of this year’s enacted state AI laws count as broad frameworks, Colorado SB 189 and Connecticut SB 5, and both are built around telling people when AI is part of what happens to them.

State AI Laws Grew in Number but Stayed Narrow in Scope

Most of the 280 enacted bills target one subject each. NCSL fellow Will Clark said deepfake bills were the most common type enacted in 2026, covering explicit content, political ads and unauthorised digital likenesses, as AI News reported. Healthcare and education also drew legislation. Clark identified notification bills, which require a platform to tell users they are dealing with AI, as the category with the sharpest rise, at close to three times the previous year’s level. NCSL’s article frames the states as the place where AI rules are actually being written. Connecticut Sen. James Maroney told the session that Congress has passed no major tech law since the Children’s Online Privacy Protection Act in 1998.

Colorado’s Revised AI Act Replaced Assessments With Notice and Human Review

Colorado SB 26-189, signed on May 14, 2026, repeals and reenacts the 2024 Colorado Artificial Intelligence Act, which never took effect. According to a June 2 analysis by the law firm Wiley, the new version drops mandatory risk management programs, annual impact assessments and the broad duty of care against algorithmic discrimination. What remains is disclosure. Deployers must tell people when AI is used in a consequential decision, explain how automated tools affect outcomes, let people seek human review or correction, and keep records. The law takes effect on January 1, 2027, and the Colorado Attorney General enforces it, with a 60-day cure period before enforcement for most violations.

Connecticut SB 5 Puts Disclosure Duties on Chatbots, Hiring Tools and Generated Content

Connecticut’s Artificial Intelligence Responsibility and Transparency Act, enacted as SB 5 and signed in late May 2026, spreads its obligations across several product types. Wiley’s analysis lists the main ones. Operators of AI companions must disclose that users are talking to an AI. Employers deploying automated decision tools must notify applicants and employees and give written notice before decisions. Generative AI providers with more than one million monthly users must attach provenance data to content they create or materially alter. Providers must get written consent before charging for AI subscriptions. The law also creates a regulatory sandbox and a third-party verification pilot, and its provisions take effect in stages from October 2026 to January 2028.

Notice Is the Requirement That Recurs Across the State AI Laws

Read side by side, the two broad state laws and the fastest-growing category of narrow bills ask for the same thing from a product: a clear statement that AI is involved, delivered at the right moment, with a path to a human when the outcome matters. The details differ by state. Colorado ties notice to consequential decisions. Connecticut ties it to companions, employment and synthetic media. Utah, which NCSL holds up as the example of a lighter approach, runs an AI Learning Laboratory created in 2024 that lets companies work through rules with the state before launch. The trigger conditions vary, and the capability a company needs underneath them is the same.

Why This Matters for Engineering Leadership

Treat AI notice as a platform service with one owning team. When each product squad writes its own disclosure string, nobody can say which surfaces are covered. Start with an inventory of every surface where a model talks to a person or shapes a decision about them, and tag each one with the decision type: hiring, credit, housing, health, companionship, generated media. Build one notice and human review service that those surfaces call. Keep the jurisdiction rules in configuration, so that when a state passes the next notification bill the change is a rules update in one place. Log every notice shown, every review requested and every override, because Colorado expects records and a regulator’s first question will be whether you can prove the notice appeared. The nearest deadline is concrete. Colorado’s law starts on January 1, 2027, about six weeks from now. If your product makes or supports consequential decisions about Colorado residents, the notice path, the explanation of how the tool affected the outcome and the human review route all need to work in production by then.

FAQ

How many AI bills did US states pass in 2026?

According to the National Conference of State Legislatures, states introduced 1,539 AI bills in 2026 and enacted 280 of them by early October. In 2025 they introduced 1,262.

Which states passed comprehensive AI laws in 2026?

NCSL counts two broad frameworks: Colorado SB 189 and Connecticut SB 5. Most other enacted bills cover narrower subjects such as deepfakes, notification and frontier models.

When does the revised Colorado AI Act take effect?

Colorado SB 26-189, signed on May 14, 2026, takes effect on January 1, 2027. It requires deployers to notify people when AI is used in consequential decisions and to offer human review.

What does Connecticut SB 5 require on AI disclosure?

It requires AI companion operators to tell users they are talking to an AI and requires employers using automated decision tools to notify applicants and employees. Its provisions take effect between October 2026 and January 2028.

Why are states regulating AI instead of Congress?

Congress has not passed major tech legislation since the Children’s Online Privacy Protection Act in 1998, according to Connecticut Sen. James Maroney. States have filled the gap with their own bills.

What should engineering teams build in response?

A single AI notice and human review service that every product surface calls, with jurisdiction rules held as configuration and every notice shown logged as evidence.

Originally reported by Ryan Daws for AI News, October 22, 2026: State legislatures become de facto AI regulators as Congress stalls. Primary sources: National Conference of State Legislatures, “As Congress Deliberates, States Lead the Way on AI Regulation” (October 21, 2026); Wiley, “Connecticut Enacts AI Framework While Colorado Scales Back Landmark AI Law” (June 2, 2026).

Read more AI News