California’s most consequential AI-pricing bill cleared both chambers and still didn’t become law — a procedural death that tells you exactly where the algorithmic-pricing fight goes next.
AB 2564 — Legislative Record
2025–2026 Session / Earlier
AB 2564 (Assemblymember Ward) passes the California State Assembly; advances to the Senate with majority support in both chambers.
August 31, 2026 — Senate Floor
State Senate passes AB 2564 22 to 14; bill returns to Assembly for concurrence on Senate amendments.
August 31, 2026 — Assembly Floor
Assembly never takes the concurrence vote. Official leginfo terminal action: “In Assembly. Concurrence in Senate amendments pending.”
Midnight — Sine Die
Legislature adjourns. Bill dies pending final concurrence — procedurally, not on the merits, and never reaching Governor Newsom. KALW confirms the clock failure.
Per the official California legislative record at leginfo.legislature.ca.gov and confirmed by KALW, AB 2564 — Assemblymember Ward’s bill to ban AI-driven personalized pricing, sometimes called “surveillance pricing,” in which prices are set for individual consumers using data collected about them — died on the final night of California’s 2025–2026 legislative session without completing passage. The Senate passed it 22 to 14 on August 31. It returned to the Assembly. The Assembly never took the concurrence vote on the Senate’s amendments before the midnight sine die deadline. The bill’s last recorded action is “In Assembly. Concurrence in Senate amendments pending.” It did not reach the Governor.
Two facts need to be held precisely. First: this is a procedural death, not a defeat on the merits and not a veto. Because the Assembly never completed final passage, the bill never reached Governor Newsom — framing this as the legislature or the Governor rejecting a surveillance-pricing ban would be wrong. Second: “it had the votes” is a strong inference from the 22-to-14 Senate margin and the bill’s earlier Assembly passage, but a concurrence vote is a vote on the amended text and is not formally identical to the prior passage. The Assembly Daily Journal is the authoritative record and may post additional detail; the defensible statement, per leginfo and KALW, is that AB 2564 died pending concurrence. Nothing here is legal advice.
AB 2564 sat within a broader California AI legislative stack — a session that sent a substantial array of other AI bills to the Governor for signature or veto. It was the major consumer-facing pricing measure in that stack and the one that, alone among them, did not complete passage. Its failure was not a policy rejection; it was a calendar failure in the closing hours of a packed session. That distinction is the entire story.
The key insight: A bill that dies on the clock with apparent majority support in both chambers is not a defeat — it is a deferral with a running start. AB 2564 returns to the 2027 session not from scratch, but from a demonstrated majority, which changes the probability calculus for every company running ML-based personalized pricing in the largest state economy in the US.
The Structural Read
The shape of this failure matters more than the failure itself. When a bill dies on the merits — insufficient votes, committee block, floor defeat — it signals that the political coalition behind it is too weak to carry the measure. When a bill dies on the clock, it signals the opposite: the coalition exists, the votes are likely there, and the only missing ingredient was time. That is a fundamentally different regulatory signal, and it maps to a different set of near-term and medium-term risks for the industry.
Read through the Map of AI Redrawn framework, the Permission Layer — the stratum of government and regulatory action that determines which AI capabilities actually ship at scale — just demonstrated two things simultaneously. The categorical route (AB 2564) can win a legislative majority and still not become law, because categorical bans are politically heavy and procedurally fragile in a packed session. The conduct route (the FTC and 22-state suit against Amazon over its alleged secret ad surcharge, the live enforcement action on personalized pricing) is slower, narrower, and evidence-intensive — but it is the one actually moving. Both routes are now in play at once, and the week just made both visible.
Permission Layer — Map of AI Redrawn
“The Permission Layer doesn’t just block capabilities — it shapes which business models are viable. When the categorical route stalls on the calendar, the conduct route fills the vacuum. Both routes are constraints; they operate on different timescales and different levels of precision. Near-term, enforcement is the risk. Medium-term, the ban is.”
For anyone building pricing systems on machine learning, the practical read is this: the near-term enforcement risk is conduct-based — was your pricing disclosed, was it deceptive, does it mirror the alleged Amazon ad-surcharge pattern the FTC is litigating now? (Cross-reference: FTC v. Amazon Algorithmic Pricing, FourWeekMBA.) The medium-term risk is the categorical ban that now has a demonstrated majority behind it in California and reintroduces in 2027 from a position of legislative strength, not legislative aspiration. Those are two different compliance postures, operating on two different timelines, and conflating them is how companies get the risk calibration wrong in both directions.
The state-federal pincer is the structural frame that ties these together. With the fast state route stalled for now, the binding near-term constraint on personalized-pricing AI in the US is federal: the FTC case, the 22-state coalition, and the conduct-enforcement apparatus that doesn’t require new legislation to move. The categorical ban — faster to write, blunter in scope, and now with a majority on record — is the medium-term constraint, and it arrives in 2027 with the wind of demonstrated support behind it. AB 2564 also belongs to the same California session that sent the broader AI deployment stack to the Governor; it was the one consumer-pricing measure in that stack that did not make it across (see: California AI Regulation Stack, FourWeekMBA). Its isolation in failure is itself a signal about where legislative energy concentrates next session.
Three Implications
IMPLICATION 1 — The 2027 Reintroduction Starts from Strength
A bill that dies on the clock with a 22-to-14 Senate margin and prior Assembly passage does not restart from zero. Ward or a successor reintroduces in 2027 with a majority already on the record, a refined bill that incorporates the Senate amendments that caused the concurrence gap, and a political environment in which AI pricing has had another year of public salience. The categorical ban is not dead as a policy; it is deferred with compounding momentum. Every company running individualized ML pricing in California should be treating 2027 as the planning horizon for categorical risk, not 2028 or beyond.
IMPLICATION 2 — Federal Enforcement Is the Live Front, Not State Legislation
With the California categorical route stalled until 2027, the FTC’s suit against Amazon — alleging a secret ad surcharge embedded in algorithmic pricing — is the active constraint on personalized-pricing AI in the US right now. That case moves on conduct: was the pricing deceptive, was it disclosed, does it constitute an unfair trade practice? For ML-pricing builders, that means the compliance question today is not “is this practice legal in California” but “can we document that this pricing was disclosed and not deceptive under FTC standards.” Those are different legal frameworks, different evidentiary burdens, and different remediation playbooks.
IMPLICATION 3 — Categorical vs. Conduct Is the Axis That Defines the Regulatory Decade
This week put both routes on display at the same time. Categorical bans — AB 2564’s model — are fast to write, sweeping in scope, and politically legible, but they are heavy to pass and easy to lose on the calendar. Conduct-based enforcement — the FTC model — is slow, narrow, evidence-intensive, and years-long, but it requires no new legislation and is already in motion. The industry has spent most of its lobbying energy on the categorical route, because that’s the one that looks like an existential threat. The conduct route, which can reach any company whose pricing is found deceptive regardless of any ban, is the one that deserves equal weight in the risk register. Both routes are accelerating. Neither is going away.
Business Engineer Framework
The Permission Layer — Map of AI Redrawn
The Permission Layer is the stratum of the AI stack where government and regulatory action determines which capabilities actually ship — and at what scale. AB 2564’s clock death and the FTC’s Amazon case are both Permission Layer events, operating on different timescales and through different mechanisms. Understanding which layer a regulatory event sits in tells you how fast the constraint arrives and how blunt it is. The Map of AI Redrawn maps all nine layers — from compute to the permission boundary — and shows where the pressure points are accumulating in 2026 and into 2027.
AB 2564 did not lose — it ran out of time, with the votes likely in hand, and that is a materially different signal than a bill the legislature rejected. The categorical ban on AI surveillance pricing in California is deferred, not defeated, and it returns in 2027 from a position of demonstrated majority support. Until then, the conduct-based route — the FTC and 22-state enforcement action against Amazon’s alleged algorithmic pricing deception — is the live constraint, operating now, requiring no new legislation, and reaching any company whose personalized-pricing practices fail the disclosure and deception test. The state-federal pincer on algorithmic pricing is not a future scenario; it is the current structure of the market. The only question open is which route lands first.
This is businessanalysis, not legal advice. AB 2564 died pending final Assembly concurrence at adjournment per the official legislative record; it was not vetoed and did not reach the Governor.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
