Who Governs a Machine? Washington and the States
Why it Matters
Who writes the rules for a machine that answers customers and ranks job applicants: Congress, the President, or the state where the customer lives? The Constitution sorted those powers long before anyone built a chatbot. Congress can pass a law that displaces a state’s, while a President can direct federal agencies but cannot repeal a state statute. A headline about AI seldom says which power was used, and that one fact decides whether a business has to change anything. You trace the federal government’s moves on AI to the power behind each, then tell a Daytona Beach retailer what binds it today.
Current Context
On September 29, 2026, the President signed Executive Order 14434 (opens in a new tab), which tells federal agencies to write “Super Intelligence” in place of “artificial intelligence” in their letters, websites, reports, and other documents no statute governs. The order defines its new term by borrowing the definition of artificial intelligence that Congress wrote into 15 U.S.C. § 9401(3) (opens in a new tab). It gives the President’s science adviser sixty days to draft legislative language for a federal definition, which would take an act of Congress to become law. The order binds federal agencies alone: it changes what the government calls the technology and nothing a business that uses it must do.
Key Concepts
Preemption
The rule, from the Constitution’s Supremacy Clause, that valid federal law displaces conflicting state law. The federal law must come from the Constitution itself, an act of Congress, or a rule an agency issues under power Congress gave it. Preemption is express when a statute says so in words, and implied when a state law conflicts with a federal one.
Executive Order
A President’s written direction to the executive branch, signed and published in the Federal Register. It can tell agencies to sue, to study, to propose rules, or to change their vocabulary, and the next President can revoke it. An order displaces no state statute by itself, and most end by saying they create no right that anyone can enforce.
Agency Power
An agency can act only through power a statute gives it. A rule adopted after public notice and comment can carry the force of law, while a policy statement only announces how the agency reads a statute and means to enforce it. Section 5 of the FTC Act is the statute the Federal Trade Commission has relied on in its cases about AI.
Bicameralism
Lawmaking by a legislature with two chambers, each of which must pass the same bill before it goes to the chief executive. Congress and Florida’s Legislature both work this way. A provision one chamber passes is not law, and it dies if the other chamber strikes it or never takes it up.
Police Power
A state’s general power to protect the health, safety, and welfare of the people within it, which is the source of its consumer-protection and civil-rights laws. A state needs no permission from Congress to use it. A state AI law rests on this power and stands until a valid federal law displaces it or a court strikes it down.
Generally Applicable Law
A law that governs conduct by anyone, whatever tool carries it out, rather than one aimed at a single technology. Section 5 of the FTC Act, Title VII of the Civil Rights Act of 1964, and Florida’s Deceptive and Unfair Trade Practices Act all work this way. They reach a chatbot or a résumé screener without naming either.
Automated Decision-Making Technology
Colorado’s term, in its 2026 law, for technology that processes personal data to produce a score, ranking, recommendation, or other output used to make or help make a decision about a person. The law reaches it only when it materially influences a consequential decision, such as hiring. Its definition leaves out a chatbot that answers consumers’ questions and is not used for those decisions.
Enforcement Authority
The power to bring a case under a law, and the remedies that come with it. Colorado’s 2026 AI law gives that power to the state’s attorney general alone, after written notice and sixty days to cure where the attorney general finds a cure possible. Florida’s Deceptive and Unfair Trade Practices Act lets a customer who lost money sue as well.
Resources
- Executive Order 14365, Ensuring a National Policy Framework for Artificial Intelligence, 90 Fed. Reg. 58499 (Dec. 16, 2025) (opens in a new tab) (The order behind the headline. Read sections 3 through 8 for what it tells each agency to do, and section 9(c), which says it creates no enforceable right.)
- H.R. 1 as passed by the House, 119th Congress (May 2025), GovInfo (opens in a new tab) (The House’s version of the 2025 budget bill. Search it for “SEC. 43201” and read subsection (c), the ten-year moratorium on enforcing state AI laws.)
- Senate Roll Call Vote 363, 119th Congress, 1st Session (July 1, 2025) (opens in a new tab) (The Senate’s vote, 99 to 1, on the amendment “To strike the section relating to support for artificial intelligence.”)
- H.R. 9925, the FRONTIER Act, as introduced (July 23, 2026), GovInfo (opens in a new tab) (A pending House bill that would preempt some state AI laws. Read section 9, and note who subsection (c)(2) leaves the states free to regulate.)
- Policy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems, 91 Fed. Reg. 41638 (July 7, 2026) (opens in a new tab) (The Federal Trade Commission’s proposed statement that a state law requiring an AI company to deceive consumers is impliedly preempted. Its action line says whether it is proposed or final.)
- 15 U.S.C. § 45, Section 5 of the FTC Act, Cornell LII (opens in a new tab) (The federal ban on unfair or deceptive acts in commerce, in subsection (a)(1), and the statute the FTC’s statement rests on.)
- 42 U.S.C. § 2000e-2, Title VII, Cornell LII (opens in a new tab) (Title VII’s unlawful employment practices, in subsection (a). Section 2000e(b) limits the law to employers with fifteen or more employees.)
- Kansas v. Garcia, 589 U.S. 191 (2020) (opens in a new tab) (The Supreme Court on where preemption comes from. Read page 9 of the slip opinion before you keep any row that says a federal action displaces a state law.)
- Senate Bill 26-189, Colorado General Assembly (signed act) (opens in a new tab) (Colorado’s 2026 law on automated decision-making technology, which repealed and reenacted its 2024 AI Act. Read section 6-1-1701(2), (3), (4)(b), and (5) for whom it reaches, section 6-1-1704 for a deployer’s notices, section 6-1-1706 for enforcement, and Section 5 of the act for its dates.)
- House Bill 26-1263, Conversational Artificial Intelligence Service Operator Requirements, Colorado General Assembly (opens in a new tab) (Colorado’s 2026 chatbot law, with its duties from January 1, 2027. Open the Final Act under Bill Text and read the definition of a conversational artificial intelligence service in section 1, including what paragraph (b) leaves out.)
- X. AI LLC v. Weiser, No. 1:26-cv-01515 (D. Colo. filed Apr. 9, 2026) (docket, CourtListener) (opens in a new tab) (The suit against Colorado’s AI law that the United States joined. The minute order at ECF No. 24 says when Colorado may begin to enforce the law.)
- Colorado Automated Decision-Making Technology and Chatbot Safety Rulemaking, Colorado Attorney General (opens in a new tab) (The enforcer of both Colorado laws, and where its rules stand.)
- Section 501.204, Florida Statutes, Unlawful acts and practices (opens in a new tab) (The core of Florida’s Deceptive and Unfair Trade Practices Act. Section 501.211 gives a customer who lost money the right to sue.)
- Section 760.10, Florida Statutes, Unlawful employment practices (opens in a new tab) (The Florida Civil Rights Act’s rule for hiring, in subsection (1). Section 760.02(7) sets the fifteen-employee threshold.)
- CS/SB 482 (2026), Artificial Intelligence Bill of Rights, The Florida Senate (opens in a new tab) (Florida’s AI bill, its votes, and its fate. Open the version labeled S 482 e1 under Bill Text to read section 7, which would have created section 501.9985, Florida Statutes.)
- Artificial Intelligence Legislation Database, National Conference of State Legislatures (opens in a new tab) (A searchable list of state AI bills. Filter by state and choose Enacted to see how many states have acted while Congress has not.)
What to Do
Every federal move on AI rests on some source of power, and the source sets how far the move reaches. You trace six federal actions and two state laws to the text that created each, then use what you find to answer the question a client actually asks. What you post is the table and a short note to the client.
Part 1: Read, Then Meet Beacon
Start with the background reading. Then read the facts below, and keep the owner’s question in mind as you work, because the note in Part 4 answers it.
Beacon Pet Supply. Beacon is a pet-supply retailer in Daytona Beach with about 60 employees, a store, and a warehouse, plus a second warehouse near Dallas, Texas, and a small distribution center outside Denver, Colorado. It sells online to customers in Florida, Texas, and Colorado. It uses two AI systems.
Ask Beacon, the chatbot. Ask Beacon answers product questions and handles returns on Beacon’s website, for every customer who opens it. A software vendor built it for Beacon on a general-purpose AI model, and it runs under the Beacon name.
The screener. Beacon licenses a résumé screener from a second vendor. The screener reads every application and ranks the applicants for interviews. Beacon uses it to hire for the Daytona Beach store and warehouse, for the Texas warehouse, and for the Denver distribution center.
The owner’s question. Beacon’s owner forwards you a headline saying the President has banned state AI laws, with a note: “Does this mean we can stop worrying about Colorado? And what does federal law actually require of Ask Beacon and the screener?”
Part 2: Build the Table
Build a table with one row for each item below and five columns. The first column names who acted and the power the action rests on: the Constitution, a statute, or the President’s authority over the executive branch. The second says in one sentence what the action does. The third says whether it binds Beacon today, and for which system. The fourth asks about preemption: for a federal action, whether it displaces any state AI law, and for a state law, whether anything federal has displaced it yet. The fifth says where the action stands, with the date you checked and the page you checked.
The rows. (1) Executive Order 14365. (2) The ten-year moratorium on enforcing state AI laws in section 43201(c) of H.R. 1, as the House passed it, and the Senate’s vote of July 1, 2025. (3) The FRONTIER Act, H.R. 9925. (4) The Federal Trade Commission’s proposed policy statement of July 2026. (5) Section 5 of the FTC Act, 15 U.S.C. § 45(a)(1). (6) The United States’ intervention in X. AI LLC v. Weiser. (7) Colorado’s two 2026 laws, Senate Bill 26-189 and House Bill 26-1263, taking each of Beacon’s systems in turn. (8) Florida’s Deceptive and Unfair Trade Practices Act, section 501.204, Florida Statutes.
Draft every row from the background reading and the sources in Resources before you open a chat. Then score your own draft: after the fifth column of each row, write how many of the five entries you could defend with a sentence from the source’s own text, out of five.
Part 3: Verify Every Row
Open the text behind each row and check your five entries against it. Run the Statute Verifier on each row, pasting the row and the passage you relied on, and then mark the row confirmed, corrected, or unsupported, saying in a few words what you changed. The fifth column needs a fresh look at the source itself, because these actions move. Check the Federal Register for whether the FTC’s statement is still proposed, GovInfo for the FRONTIER Act’s latest action, the docket for Weiser, and the Colorado Attorney General’s page for the rules. If any row says that a federal action displaces a state law, test the claim with the Preemption Explainer before you keep it.
Part 4: Answer Beacon’s Owner
Write a note of about 200 words to Beacon’s owner, in plain words, that answers four questions in order. Did the President ban state AI laws, and what did the order actually do? What federal law reaches Ask Beacon and the screener today? Does either Colorado law reach either system, and from when? Why does Florida have no law written for AI that reaches Beacon, and which Florida laws reach each system anyway? Score the note one point for each question it answers from a row of your table. Then paste the owner’s question, your table, and the note into a fresh chat and run the Client Note Critic. For each flag, go back to the text and decide: revise the note, or keep it and write in a sentence why the critic was wrong.
Part 5: Post Your Table and Note
Create a single new discussion thread with your verified table, its row scores and marks, and your note to Beacon’s owner below it. Before you post, check that every row cites the text behind it and every answer in the note rests on a row.
Part 6: Share Chat Link
Include one AI chat link with a 1–2 sentence explanation of what the conversation shows and why you chose to share it.
Suggested AI Prompts
Use these prompts as a starting point, then adjust them to fit your goal. Strong prompting develops through trial, revision, and testing. It’s a foundational skill that grows into more advanced AI work such as context engineering and agent-based workflows.
Statute Verifier
Act as a legislative research librarian. I will paste a row from my table along with the text of the executive order, bill, statute, agency statement, or court order it cites. Tell me whether the text supports each entry in the row, quoting the exact language that does or does not, and note anything material the row leaves out, such as an effective date, an exclusion, a limit on who may enforce the law, or whether a document is proposed or final. If the pasted text does not address an entry at all, say so. Do not rely on your own memory of the law; use only the text I provide.
Confined to the text you paste, the model works as a careful reader instead of a source of new claims, which is what source grounding asks of it. An entry it cannot find in that text is one to recheck before you keep it.
Preemption Explainer
Act as a constitutional law professor explaining federal preemption to a college student. I am looking at a claim that a federal action, such as an executive order, a bill, or an agency policy statement, preempts a state artificial intelligence law. Explain what preemption is, where it comes from, what each kind of federal action can and cannot do to a state statute, and what would have to happen for a state AI law actually to be displaced. Then apply that explanation to the specific claim I paste. Where the answer depends on pending litigation or legislation, say what is pending and that the outcome is unknown rather than predicting it.
Making the model lay out the rule before it touches the facts is decomposition, doctrine first and application second. The instruction to name what is pending instead of predicting it keeps a fast-moving question honest.
Client Note Critic
Act as a supervising paralegal reviewing a note a colleague wrote to a small-business client. I will paste the client’s question, my table of federal and state actions on AI, and my note of about 200 words. Check that the note answers the client’s four questions in order: what the executive order did, what federal law reaches the chatbot and the résumé screener today, whether either Colorado law reaches either system and from when, and what Florida law reaches each system. For each problem, name the sentence and ask me one question that would send me back to the table or the law’s text. Do not rewrite the note, supply an answer, or tell me what the law requires. If your view depends on a law, bill, or court order that may have changed after your training, say so plainly instead of guessing. Wait for me to paste the question, the table, and the note.
Held to questions, the model leaves the note yours, and a critic that asks rather than answers is the Socratic form of a review. Asking it to name what may have changed since its training tells you which of its doubts to take to the text first.