Contents5
On 16 September 2026, California’s governor signed SB 1050, which requires advertisements to disclose AI-generated performers; on 30 September, he signed SB 947, which bans dismissing an employee on the strength of an automated system alone. In Colorado, a law on automated decisions takes effect on 1 January 2027. None of these laws stops at the edge of American companies: they target whoever sells to, hires or makes decisions about the state’s residents.
This post in our Business AI section takes stock as of 20 October 2026. It is not legal advice: if you have employees or customers in the United States, a lawyer in the state concerned has the final say.
In short
There is no federal US law on AI; the states legislate, each with its own text and its own date.
A presidential executive order of December 2025 challenges those laws, but according to the status reports we have read (the most recent dates from September 2026), no court had struck any of them down.
A European business falls within scope through its customers, employees or candidates in the state, not through where its head office is.
The obligations look alike from one law to the next, and look like Europe’s: inform people, keep a human in the decision, disclose synthetic content.
No federal law, but state laws
The United States has no single AI regulation like the European AI Act. The states vote the laws, and a foreign company is subject to them as soon as it does business there: Colorado’s law, for instance, targets any business “doing business in Colorado” that uses automated decision-making technology, according to the law firm Skadden.
The federal government wants to take back control. Executive Order 14365 of 11 December 2025 set up a task force at the Department of Justice to challenge state AI laws, and asked the Department of Commerce to list the laws it considers too burdensome. The law firm Ropes & Gray sums it up: state laws remain in effect until Congress passes a law that overrides them or a legal challenge succeeds.
According to the status reports we have read, neither had happened. The federal bill that would suspend state laws for three years, released as a draft in June 2026, had not been introduced as legislation, according to a status report published by casrai.org around 20 September 2026. In Colorado, the Department of Justice joined an AI company’s lawsuit against the earlier law in April 2026, and the state agreed not to enforce it before the judge rules, reports the law firm McDermott in an analysis dated 27 May 2026; the replacement law is expected to face the same challenge. These findings are therefore dated: we found nothing to the contrary as of 20 October 2026, but a ruling or a filing in Congress can change everything within days, so check where the case stands on the day you decide.
Key takeaway
The clash between Washington and the states makes the timetable uncertain, not the obligations optional. Preparing for a law that may be suspended costs less than discovering that it applies.
The dates that matter, 2026 to 2027
- 1 January 2026
In California, SB 53 covers frontier model developers and SB 243 covers companion chatbots; in Illinois, the law on AI in employment takes effect.
- June 2026
New York State requires advertisements to disclose synthetic performers.
- 16 September 2026
California signs SB 1050, its equivalent for audio and video advertising.
- 30 September 2026
California signs SB 947 on discipline and dismissals decided by an automated system.
- 1 January 2027
SB 1050 applies in California; in Colorado, SB 26-189 applies if the attorney general’s rules are ready.
- 1 July 2027
SB 947 applies in California.
California: four laws, four audiences
- SB 53, in force since 1 January 2026, covers developers of models trained with more than 10²⁶ operations; those with revenue above 500 million dollars must also publish a safety framework. They report critical safety incidents within 15 days, and each violation can cost up to one million dollars, according to the law firm Morrison Foerster. A business that uses these models is not covered: its supplier is.
- SB 243, in force the same day, requires operators of companion chatbots to make clear that users are talking to a machine and to offer breaks to minors, TechCrunch reported when it was signed.
- SB 1050 applies from 1 January 2027: an audio or video advertisement shown in California that prominently features an AI-generated performer, voice included, must say so clearly. The governor states that a non-compliant advertisement may not continue to run, and Skadden notes that a breach falls under California’s false advertising law, which is open to private lawsuits.
- SB 947 applies from 1 July 2027: no discipline or dismissal based solely on an automated system. Its content is covered in the human role in the face of AI judgement.
Colorado: inform, explain, review
SB 26-189, signed on 14 May 2026, replaces the 2024 law before it ever applied. It covers consequential decisions (employment, housing, lending, insurance, health care, education) that automated technology materially influences. The deployer must notify the consumer beforehand, explain the tool’s role within 30 days after an adverse decision, and offer data correction and human review, according to Skadden. Only the attorney general can enforce it, with penalties of up to 20,000 dollars per violation and a 60-day period to cure.
The 1 January 2027 date assumes the attorney general has adopted the implementing rules by then. In casrai.org’s September 2026 status report those rules were still to be issued, and we found no adopted rules as of 20 October 2026; their progress is tracked in what changes on 1 January 2027 for digital.
Illinois: hiring first
Since 1 January 2026, HB 3773 bars employers from using AI to recruit, promote, discipline or dismiss when it results in discrimination against a protected class, and requires them to notify candidates and employees that they use it, summarises the law firm Epstein Becker Green. How that notice must be given depends on rules from the state’s human rights department.
Are you concerned? Four situations
- You run audio or video ads in California with a generated voice or face: SB 1050 from 1 January 2027. It follows the same logic as disclosing deepfakes in Europe, explained in our guide to marking AI-generated content.
- You hire or employ people in the United States with a tool that sorts, scores or proposes disciplinary action: Illinois since January 2026, Colorado in 2027, California in July 2027.
- You grant credit, insurance or housing to Colorado residents with the help of an automated score: SB 26-189.
- You offer a conversational chatbot to Americans, especially one that builds an ongoing relationship: SB 243 in California, and a chatbot law in Colorado on 1 January 2027. The issues are laid out in our article on AI companions and teenagers.
If none of these situations applies to you, these laws will mostly reach you through your suppliers, who will adapt their tools and contracts.
What it changes for a team of 20 to 500 people
The good news is that these laws ask for almost the same thing as European law: know where an automated tool takes part in a decision, tell the people concerned, and keep a human who can reverse it. Article 22 of the GDPR already requires this for fully automated decisions, as explained in automated decisions: can AI decide on its own?.
A business that keeps an inventory of its AI uses, a review procedure and an information notice therefore prepares for both sides of the Atlantic at once. The approach is set out step by step in our guide to the AI Act and GDPR for small businesses, and its place in an overall project in our guide to integrating AI in your business. Every deadline, European and American, is gathered in our AI regulation calendar.
From the field
This post comes out of the content engine we run for this blog: the dates of the six laws cited were checked on 20 October 2026 on the California governor’s website, in the press and in the analyses of five law firms, and the article stays in draft until the human review that alone can publish it.
To find out which of your tools take part in a decision about a person, and what to document before 2027, our digital and AI maturity audit starts from your actual uses.
Frequently asked questions
Does a European business have to comply with US AI laws?
Yes, with those of the states where it has customers, candidates or employees, within the scope of each law. California targets advertisements shown on its territory and employers of Californian staff; Colorado targets businesses that operate there and make consequential decisions about its residents. A European head office offers no protection, but a business with no activity in those states is not covered.
Did Donald Trump’s executive order cancel state AI laws?
No. The order of 11 December 2025 organises legal challenges against them and prepares a federal law, but it cannot repeal a state law on its own. According to the May and September 2026 status reports we have read (law firm McDermott, casrai.org), no judge had struck down a state AI law, and the federal preemption bill released as a draft in June 2026 had not been introduced in Congress.
When does Colorado’s AI law apply?
On 1 January 2027, provided Colorado’s attorney general has adopted the implementing rules by then. SB 26-189 requires informing the consumer before a consequential decision made with an automated tool, explaining the tool’s role within 30 days after an adverse decision, and offering human review. Only the attorney general can bring enforcement action.
Must an AI-generated voice be disclosed in an advertisement?
In California, yes from 1 January 2027: SB 1050 requires a clear disclosure when an audio or video advertisement prominently features an AI-generated performer, voice included. New York has required it since June 2026 for visuals, without covering voices. In Europe, the AI Act already requires businesses to disclose the deepfakes they publish.
Sources
- According to the law firm Skadden (new tab)skadden.com
- Remain in effect (new tab)ropesgray.com
- Had not been introduced as legislation (new tab)casrai.org
- Reports the law firm McDermott (new tab)mcdermottlaw.com
- According to the law firm Morrison Foerster (new tab)mofo.com
- TechCrunch reported (new tab)techcrunch.com
- Governor (new tab)gov.ca.gov
- Skadden (new tab)skadden.com
- Summarises the law firm Epstein Becker Green (new tab)ebglaw.com
Jim · founder of Brain Plus AI
Also in business ai
Understand9 min
AI Act and digital rules calendar: deadlines 2026 to 2028
AI Act, Data Act, Cyber Resilience Act, NIS2, e-invoicing in France and the first US state laws: every date that matters for a business, from autumn 2026 to August 2028, on one page kept up to date. For each deadline, who is covered and what to prepare.
Stepping back10 min
The human role alongside AI: knowing how to judge
AI excels at some tasks and is confidently wrong on neighbouring ones. The human's value shifts: knowing where the line runs, keeping your judgement, answering for the result.
Stepping back8 min
Automated decisions: can AI decide on its own?
An algorithm may be more consistent than a manager in a hurry, but a decision that affects a person's life has to be explainable and open to challenge. The law leans clearly one way: the two positions, our view, and the rule to adopt without waiting for 2027.