Episode Details

The Federal Trade Commission voted two to zero on August 19, 2026 to propose an enforcement policy statement targeting undisclosed personalized pricing under Section 5 of the Federal Trade Commission Act, the clearest signal yet that a startup adjusting prices by customer data now carries federal exposure independent of any state artificial intelligence statute. The proposal targets a business that shows one consumer a price built from that consumer’s own data while representing the price as standard. It entered a thirty day public comment period rather than taking immediate effect, and the Commission has conceded it cannot declare personalized pricing itself unlawful, only the deception surrounding it. A founder who reads no further than the headline will miss both the narrower federal opening and the far wider set of obligations sitting underneath it.

Most 2026 compliance calendars describe four states, Texas, California, Illinois, and Colorado, as having adopted artificial intelligence oversight effective January 1, 2026, with impact assessments and human review required for consequential automated decisions. That description was accurate for two of the four states on that date, incomplete for a third, and wrong about the fourth in a way that matters more than any of the others.

Texas did act on January 1, 2026, but the Texas Responsible Artificial Intelligence Governance Act, House Bill 149, is an antidiscrimination and government disclosure statute, not a private sector testing regime. It bars intentional discrimination against a protected class through an artificial intelligence system and requires a government agency, not a private startup, to disclose when a person is interacting with one. Illinois also acted on schedule, amending the Illinois Human Rights Act through House Bill 3773, but the statute reaches recruitment, hiring, promotion, discipline, and discharge alone. A startup pricing a subscription or scoring a customer for a service, rather than screening an employee, is not the party House Bill 3773 was written to reach.

California came closer to the popular description, and its rule is the one most likely to reach a consumer pricing decision, if that decision touches a defined trigger. The California Privacy Protection Agency’s regulations, codified beginning at title 11, section 7001 of the California Code of Regulations, took effect January 1, 2026 and require a risk assessment, a notice delivered before first use, an opt out where no exception applies, and a human reviewed appeal process for any automated decisionmaking technology used in a significant decision, defined as one touching financial or lending services, housing, education, employment, or health care. A dynamic discount tied to nothing more than browsing behavior sits outside that definition as written. A credit or financing decision embedded in the same checkout flow does not.

Colorado is where the popular description breaks down entirely. The Colorado Artificial Intelligence Act, Senate Bill 24-205, never took effect on any date. Colorado first postponed it from February 1, 2026 to June 30, 2026 through Senate Bill 25B-004, and then Colorado’s Governor signed Senate Bill 26-189 on May 14, 2026, repealing the original act before its postponed date arrived and replacing it with a narrower automated decision making technology statute that does not take effect until January 1, 2027. The new law drops the annual impact assessment entirely and instead gives a consumer a right to request meaningful human review of an adverse consequential decision, defined to cover education, employment, property leases or purchases, financial services, insurance, health care, and essential government services. General retail pricing sits outside that list as written, a gap the Colorado Attorney General has not yet been asked to close.

None of this touches the obligations that predate any AI specific statute. The Equal Credit Opportun