Privacy & Data Rights
State privacy laws, data-broker rules, deletion rights, and the systems built to collect and profile consumer data.
Start with the writing
The council passed the Fair Pricing and Transparency Act (CB 121267) 7 to 2, barring large grocers from using browsing history, location, or inferred income to change the price a shopper sees. If the mayor signs, Seattle becomes the first city to ban the practice, and the fight over personalized pricing moves from statehouses to city councils.
SB 1119 adds parental controls, crisis protocols, and first-in-the-nation independent child-safety audits and annual risk assessments for companion chatbots, while the wider package bans algorithmic feeds and autoplay for users under 16 and restricts targeted ads to kids. Useful as the accountability counterweight to the AI-companion adoption data already on the radar.
Enforcement Advisory 2026-01 warns data brokers that wrong answers on their annual registration carry a daily fine, and notes the agency has already brought actions over reporting errors. The registry is what makes DROP work, so accuracy in it is now being treated as a consumer-rights issue rather than paperwork.
SalesIntel Research, which markets a product that de-anonymizes website visitors into names, direct dials, and verified emails, pays a $36,400 fine and must route future deletion requests through DROP for failing to register on time. The product description is the interesting part, since it shows what registration is meant to make visible.
Internal testing reportedly exposed private data and produced unreliable behavior before launch. Read it alongside Meta's Secure VM and Sentinel claims in the announcement post.
Meta's free US personal agent browses, fills forms, negotiates and pays on a user's behalf, checking out through Stripe Link with single-use card numbers. Each user gets a dedicated Secure VM and a separate Sentinel agent that gates anything reaching the internet, and Meta says agent data stays out of its ad systems. Worth watching for how those accountability claims hold up once it is in millions of hands.
27% of U.S. adults online now use chatbots for social or emotional purposes, and 39% of those say they tell the bot things they tell no one else. That is a great deal of intimate data flowing into commercial models, with very little said about where it goes.
SB 923 closes the CCPA loophole that let a business keep personal data it bought from a third party even after a consumer asked for deletion. It now heads to the Governor, and would make DROP requests reach a good deal further than they do today.
500,000 signups since January and tens of millions of records reported deleted. It is the first real usage number for a one-stop deletion right, and a useful baseline for judging whether the other states' registries get any traction.
Researchers reviewed all 522 brokers on California's registry and found 9% fully compliant. 45% filed no rights-request metrics at all and 64% added friction to deletion requests. Read alongside the DROP signup numbers, it shows how far the demand side is running ahead of the supply side.
The first federal enforcement posture on personalized pricing. The FTC says it cannot ban the practice outright, but hiding that a price was set using personal data, or what data set it, can violate Section 5. Comments were extended to September 25.
83% of surveyed Washingtonians say they have little or no control over who accesses their data, and the report names data-broker opacity as a priority. A state AG setting out a privacy agenda in writing is usually a signal of enforcement to come.
Ranking Member Pallone sent 19-question letters to eight U.S. airlines asking what consumer data feeds their fares, with responses due August 25. Airlines were the industry that made surveillance pricing a household phrase, so the answers will be worth reading.
California's privacy agency brought its first case under both the CCPA and the Delete Act, against an Iowa data broker that failed to register and made consumers hand over unnecessary personal information to opt out. It lands less than two weeks after DROP deletion requests became mandatory.
As of August 1, every registered data broker in California must check the state's one-stop deletion platform every 45 days and actually delete what verified requesters ask for. Over 242,000 Californians signed up before the mandate kicked in, the first real test of a statewide deletion right at scale.
Meta's 'Muse Image' tool let anyone generate AI images of public accounts via @-mention — opt-out by default. SAG-AFTRA and CAA led the consent revolt, and Meta killed it in three days: a clean case study in how fast likeness-without-consent now backfires.
A.5328 took effect the moment it was signed June 30 — banning sensitive-data sales and uniquely requiring 'data collectors,' not just brokers, to register at fees up to $1.5M. On July 10 the AG blinked, delaying registration to 2027 after business outcry — the messiest data-broker rollout yet.
The FTC says the screening vendor duplicated criminal and eviction records, wrote off disputes as 'invalid,' and hid its data sources — making applicants look worse than they were. A clean case of bad consumer-report data costing people housing.
July 1 brought Arkansas's comprehensive privacy law online, Connecticut amendments that lower applicability thresholds and add neural data to sensitive categories, and Utah's new correction and social-graph access rights. A useful roundup of what changed and what's still coming in 2026.
The FTC's three-year case against data broker Kochava closes with a consent order barring sale of sensitive geolocation data without affirmative consent, plus deletion schedules and supplier-consent vetting. A capstone federal data-broker enforcement as state law takes over the beat.
Vermont enacts a comprehensive consumer-privacy law with opt-outs for data sale, targeted advertising, and profiling — enforced solely by the state attorney general and effective January 1, 2028.
Connecticut's omnibus privacy expansion, signed May 27, folds data-broker registration, geolocation-sale limits and pricing restrictions into one law, joining California, Texas, Vermont and Oregon.
Indiana, Kentucky and Rhode Island join the patchwork this year; California's DROP deletion platform is now compounding $200/day fines on brokers that ignore deletion requests.
EFF formally backs California’s AB 2564, which would bar retailers from setting customized prices off a shopper’s personal data and give consumers a private right to sue. The clearest privacy-first case yet against surveillance pricing as the bill heads to the state Senate.
The emotion-AI market nears $6B in 2026 with vendors claiming 90%+ accuracy reading faces and voices, even as the science stays contested and consent regimes lag. The next extraction frontier after behavioral tracking.
The definitive overview of the 20 state privacy regimes now in effect and California's data-broker DROP deletion platform phasing in, background to nearly every privacy story this year.
CalPrivacy's new enforcement unit, plus SB 361's expanded disclosure rules including sales to generative-AI developers, signals the data-broker crackdown has teeth ahead of the August 1 deadline.
From June 9, users receive a second cut of Meta's $725M Cambridge Analytica settlement (first round averaged $29.43) — the 'settle' stage of the playbook, priced at pennies on the data dollar.
Settlement over claims that an AI tool listened through phone microphones and fed targeting data into sold email lists. A clean example of the 'know, deny, settle' arc playing out in real time.
A first read on the federal bill that would set a national baseline for consumer-data rules — worth watching against the patchwork of state laws now in effect.
Bans the sale of precise cell-phone location data and keeps a private right of action despite heavy industry lobbying. The 'regulate' stage of the playbook, arriving on schedule.
A biometric-privacy suit testing whether face-scanning at the doorbell triggers the same liability as workplace fingerprint cases.
A study on how opt-out flows are engineered to fail — concrete evidence for the gap between privacy rights on paper and in practice.
The full chronological feed lives at On My Radar.
US20260278018A1 · Google Patents ↗
A two-stage generative pipeline for advertising: a text model takes a user's personalization data together with a merchant's brand assets and writes bridging terms, and a text-to-image model then renders an ad image from those terms, so the picture on screen is assembled fresh for that one person. The inputs are whatever the ad system already holds about the user — interests, demographics, and past behavior — now used not to choose which ad you see but to decide what the ad looks like. There is no step in the claims where a person reviews the generated image before it ships. It moves profiling from targeting into the creative itself, and a user has no way to tell that the image they are looking at was drawn from their file. Filed May 2026 as a continuation of a March 2024 application (now granted as US12645744) and published September 17, 2026.
In the news(1)
US20260270058A1 · Google Patents ↗
Describes identity plumbing for advertising after third-party cookies: an advertiser or publisher sends encrypted personal data — email addresses, phone numbers — into a trusted execution environment, where a generator assembles two split keys from independent key coordinators, briefly decrypts the data, and derives a consistent surrogate ID that goes back out for audience matching and attribution. No single party, Google included, holds enough key material to read the underlying data alone. The privacy gain is real but stops at the door: what leaves the enclave is a durable code that follows the same person across parties and over time, and the claims say nothing about how long it persists or what can be built on it. It is the cleanest recent example of a system engineered to be private on the way in and a tracking identifier on the way out. Filed September 24, 2025 and published September 10, 2026.
In the news(1)
US20260253102A1 · Google Patents ↗
An acquisition unit collects behavior information about customers moving through a defined physical space — a store, a mall corridor — and a generation unit feeds it through a model trained on which past content produced an advertising effect for people who behaved that way, then generates new ad content for the individual being estimated rather than selecting from a library. The data is where a shopper walks, what they stop near, and how long they linger; the applicant is the Sony division that makes the image sensors in most smartphones, which suggests how that data gets captured. The claims as published contain no consent mechanism, opt-out signal, or retention limit, and the filing's own drawings map traffic history to an inferred mental state. It brings website-style per-session personalization into a place people do not expect to be profiled. National-stage entry of a March 2024 PCT application, published August 27, 2026; the originally filed claims were canceled during prosecution and the application is under a non-final rejection.
In the news(1)
EFX-TBV-PII-RISK · Google Patents ↗
Isolates individual identity elements a consumer leaves behind in a transaction — email address, phone number, device ID — and scores each against modeled historical fraud distributions (chargebacks, declines) to rate the risk of card-not-present purchases. Equifax says the method is already deployed in its Consumer Insights, Signal Score for Email, and Payments Fraud products. It is data-broker infrastructure in its purest form: a credit bureau converting the contact details and devices of ordinary shoppers into behavioral risk scores that merchants act on and consumers never see, with no mechanism for a person to know why their transaction was flagged. NOTE: patent number to be verified — announced in Equifax's June 30, 2026 patent-portfolio release without a USPTO number; this entry links to the announcement until the number is confirmed.
In the news(1)
EFX-TBV-DATA-FABRIC · Google Patents ↗
Coordinates event data moving between multiple computing systems by holding records until defined triggers are met, so downstream systems receive complete, structured datasets rather than fragments — the patented plumbing of Equifax's 'data fabric.' That fabric ingests 20 billion records a month and orchestrates more than 250 billion keyed and linked consumer records unified from over 100 formerly siloed sources, feeding products like Account Protection, Identity Proofing, and Synthetic Identity Risk. Nominally back-office, it matters because it is the orchestration layer that makes bureau-scale profiling operate in real time: the more seamlessly consumer records flow between systems, the less friction stands between data collection and its use. NOTE: patent number to be verified — announced in Equifax's June 30, 2026 patent-portfolio release without a USPTO number; this entry links to the announcement until the number is confirmed.
In the news(1)
US12541610B2 · Google Patents ↗
A digital targeting system that scans a regulated consumer database, identifies the users who satisfy an advertiser's 'content factors,' and outputs a pre-validated list of those users to be matched against external digital-ID databases — so a display network can serve them ads. It operates over records governed by data-sharing restrictions (such as government-regulated or financial data), framing itself around compliance even as its purpose is to make otherwise-restricted consumer records addressable for targeting. Filed by Experian, one of the largest consumer data brokers, and granted February 2026. It is the identity-graph plumbing beneath behavioral advertising: the mechanism that turns regulated personal records into targetable ad audiences, with consent and accountability handled as access-control bookkeeping rather than as a consumer choice.
US12513102B2 · Google Patents ↗
Constructs a persistent AI persona from a user's behavioral, emotional, and interaction data — including post-mortem operation (posthumous digital presence). The system analyzes a lifetime of platform activity to generate a digital twin capable of posting, reacting, and interacting on the user's behalf, including after death. Originally filed by Meta CTO Andrew Bosworth in 2023; granted December 2025 and published February 2026. Meta has stated it has no current plans to ship this as a product, describing it as a protective filing — the same framing used across the tobacco-playbook 'deny' phase. The most-cited patent across all daily briefing editions.
In the news(8)
The full catalog lives at Patent Watch.