20 Jan 2026 Search

For many people in web advertising in 2026, the Google “Privacy Sandbox” story of the past six years or so was an embarrassment, like having your brand name on NFTs or something. And a lot of people would rather forget it. (I’m not including links to all the opinion pieces by advertising thought leaders who go on about how important it is for everyone to test “Privacy Sandbox” because it’s the future. You’re welcome.)

But somehow I still think this mess is worth remembering. Google tried to do a bunch of crimes in the browser at once and then (mostly) had to stop. So, in my humble opinion, that counts as a win. And it’s worth celebrating, and posting some notes and links on. As scams and slop proliferate, we’re going to need more wins like it.

We did get lucky, though, because Google got too greedy, too fast. Even in its early days, Privacy Sandbox failed to deliver the kind of ad revenue that old-fashioned cookies could. If Google had been smart enough to throw a bone to ad-supported web sites and to small-time adtech players in the first release, just give them some sweet cash to brag on for a couple of quarters and then plan to enshittify later, like every other Big Tech scheme, this story could have come out a lot differently. A tweak to “Performance Max” to throw a little more of Google’s incomprehensible flow of ad money into “Privacy Sandbox” would have kept it going. But sometimes different parts of big companies go off in different directions, and that saved us.

The “Privacy Sandbox” problem is still not completely over. The attribution cartel is still having meetings at W3C (now focusing on a descendant of a Meta/Mozilla tracking system, not the “Privacy Sandbox” one). But at least this whole saga shows that a win is possible.

Regulators and an NGO in the UK took the lead on this, so, as an Anglophile, I’m going to put on Red, White & Blue and celebrate.

2019

August 22: Initial commit to FLoC (which became Topics API). Google posts Building a more private web, stating that large scale blocking of cookies undermine people’s privacy by encouraging opaque techniques such as fingerprinting and offering “Privacy Sandbox” as an alternative. (See December 18, 2024 for more on fingerprinting.)

August 25: First comments on the Privacy Budget repository. The Privacy Budget proposal suggests a limit to the amount of individual user data that can be exposed to sites, so that in total it is insufficient to track and identify individuals. (The “privacy” measured by “Privacy Budget” is defined by some math concepts, and is unrelated to “privacy” as experienced in real life. All of the information about yourself that you most want to keep private is shared by at least k other people.)

September 4: According to an early bug report, This proposal makes false claims about the privacy properties provided by the anonymization techniques used (The “privacy” that people want is not the kind of “privacy” that “Privacy Sandbox” can provide. What if the right people had paid attention?)

November 20: Initial commit to the “ip-protection” project. We propose introducing a signed attestation (perhaps in the form of an HTTP header) that advertises the fact that a server masks IP addresses and other identifying network information from the application layer of the services that it hosts. A CDN could offer this as a feature to the services they host such that the hosted service can access more APIs than would otherwise be available because their Privacy Budget hasn’t been exhausted on the IP address.

2020

January 14: Google to ‘phase out’ third-party cookies in Chrome, but not for two years In those cookies’ place, Google is hoping that it can institute a new set of technical solutions for various things that cookies are currently used for. To that end, it has proposed a bunch of new technologies (as have other browser makers) that may be less invasive and annoying than tracking cookies have become….If it all came to pass, it would radically shift the way ad tracking and privacy work on the web. It could also open up entirely new vectors of tracking we have yet to imagine.

January 16: Initial commit to TURTLEDOVE (which became FLEDGE, then Protected Audience API)

February 12: 51Degrees, a UK-based device detection and location data firm, sends a letter to the UK’s Competition and Markets Authority, pointing out that When a Chromium-based web browser accesses a Google service additional information not available to any other service is transmitted to Google by Chromium.

November 23: Marketers for an Open Web calls on UK Competition and Market Authority to block Google’s ‘Privacy Sandbox’. The organization (founded by James Rosewell of 51Degress, and which later changed its name to Movement for an Open Web) has today written to the CMA asking them to impose a legal block on Google to delay the launch of its so-called ‘Privacy Sandbox’ technology, which will remove login, advertising and other features from the open web and place them under Google’s control. If the move succeeds it will be the first time the CMA has used its wide-ranging powers to challenge the online platforms’ ever-growing control of the internet. TechCrunch coverage: Digital marketing firms file UK competition complaint against Google’s Privacy Sandbox

2021

January 8: The Competition and Markets Authority in the UK begins an investigation. CMA to investigate Google’s ‘Privacy Sandbox’ browser changes

January 22: FLEDGE released. This document describes an early prototype for ads serving in the TURTLEDOVE family, appropriate for experimentation before a fully-featured system is ready. (FLEDGE later became Protected Audience API, and the fully-featured system was never released.)

January 26: The Industry Reacts To Google’s Bold Claim That FLoCs Are 95% As Effective As Cookies “Although Google shared information in October about its process for assessing clustering algorithms, there seems to be little available on the baseline or control group it used to compare the efficacy of FLoCs with third-party cookies.”

February 22: Securely Propagating Auction Signals feature request. This lack of exposure also allows the GAM team to extend their market position into the fledge auction. By not exposing the price of the contextual auction to publishers, publishers will have to use GAM to call runadauction if they want to use GAM at all, a choice they don’t have because AdWords won’t transact elsewhere. This later became one of the key competition issues with the FLEDGE/PAAPI auction system, and was unresolved when “Privacy Sandbox” ended.

March 3: Google’s FLoC Is a Terrible Idea from EFF.

March 3: Google Clarifies Its Stance On Post-Cookie Advertising IDs – And It’s Not Good For Ad Tech “Third-party cookies and other proposed identifiers that some in the industry are advocating for do not meet the rising expectations that consumers have when it comes to privacy. They will not stand up to rapidly evolving regulatory restrictions. They simply cannot be trusted in the long term…” (covered in Stop Letting Google Get Away With It)

June 11: The CMA accepts Google’s commitments, which leave out many of the substantial issues raised in the consultation process. (But Google will have trouble sticking to even this limited set of commitments anyway).

June 14: Notes on Google’s Commitments to the CMA by Robin Berjon. Instead of listing data sources that are excluded for specific purposes, at the very real risk of omitting an important source of distortion as Google competes with publishers in the ad market, I would like to suggest relying on a principled approach, which is essentially a return to the Drummond Doctrine: Google commits to not using publisher data for any purpose other than those explicitly requested by the publisher as part of a service agreement. On the assumption that everyone’s intentions here are to strike the best deal possible, this should not be a difficult change to agree to. It’s hard to understand how anything else would be fair.

June 24: Google Pushes Back Plan to Ditch Third-Party Cookies in Chrome to 2023

2022

January 25: Google kills FLoC, introduces Topics API as its next-gen targeting tech

February 8: CMA secures final Privacy Sandbox commitments from Google – Competition and Markets Authority

March 30: Google starts global tests of Privacy Sandbox ad targeting. A Chrome origin trial of Topics, FLEDGE, and Attribution Reporting APIs begins. Users in the EU are not automatically opted in but can turn on the “Privacy Sandbox” features using browser settings.

July 8: Google’s Topics API Picks On Smaller Publishers

July 27: Google to Delay Third-Party Cookie Phase Out Until 2024

July 28: First of the quarterly reports from the UK CMA. In making this assessment, we recognise that there is still a lot of work to be done.

August 4: Topics API Review by Aram Zucker-Scharff.

November 10: Topics API: Criteo’s First Look at Google’s Interest-Based Advertising Solution As we have seen in this article, the Topics API on its own and in its current form likely does not adequately address marketers’ needs for interest-based advertising.

December 20: Apple Webkit releases a position on Google Topics API, listing several reasons why We do not think Topics API is a good addition to the web platform.

2023

January 18: Google Topics API for targeted ads gets shot down. W3C Technical Architecture Group says Topics API is not appropriate behavior for any software purporting to be an agent of a web user. According to Robin Berjon, I guess that maintaining the revenue stream from disinformation—which is a key part of what Topics does—is more important to Google in a downturn than keeping the standards process credible.

February 2: Opting Out Of Google’s Topics API Won’t Affect Search (for what it’s worth, they never did get around to saying how usage of other “Privacy Sandbox” features by sites will or won’t affect SEO.)

February 7: User-Agent Reduction is fully deployed. Adtech firms can still pass device information that can help identify a user, but are required to maintain additional code and consume more bandwidth. (A 2021 article explains a library for reconstructing a User-Agent string from the new “client hints” system.)

March 24: Attack on privacy budgets using fake sites and malicious Match Key Provider.

April 13: Google publishes Developer enrollment for the Privacy Sandbox, which states that sites using the “Privacy Sandbox relevance and measurement APIs,” including Attribution Reporting, FLEDGE, Topics, Private Aggregation and Shared Storage, must enroll and provide a D-U-N-S Number to Google. This requirement may have been an attempt to address the “malicious match key provider” attack.

April 17: FLEDGE renamed to Protected Audience API. (FLEDGE was the experimental subset of the more ambitious TURTLEDOVE in-browser ad auction.)

June 1: Privacy Sandbox Developer Enrollment and Attestations begin.

June 30: 3 Billion Chrome Users Are About to See This Privacy Sandbox Pop-Up. Some users begin seeing a dialog claiming “Enhanced ad privacy in Chrome” with a roll-out planned to take several weeks.

December 21: In Shipping the Privacy Sandbox relevance and measurement APIs, Google states that the enrollment process will soon become a mandatory requirement for using the relevance and measurement APIs.

2024

Some time during 2024: According to one source, one or more of Google’s own consumer brands (Pixel? Nest? Chromebook?) runs a test of “Privacy Sandbox” ads. The results were never reported publicly. FIXME: add details if/when they come out.

February 6: IAB Tech Lab releases results of Privacy Sandbox gap analysis. There’s little debate that The Privacy Sandbox APIs are poised to usher in a transformative era in digital advertising fundamentals.

March 4: Privacy Sandbox’s Latency Issues Will Cost Publishers. According to data from advertising services firm Mediavine, the Privacy Sandbox’s Protected Audiences API (PAAPI) causes an average increase of 1500 milliseconds of additional latency, and lowers ad visibility to 39%.

March 28: Yohan Beugin and Patrick McDaniel of the University of Wisconsin publish A Public and Reproducible Assessment of the Topics API on Real Data. They find that 46%, 55%, and 60% of the 1207 users in the dataset are uniquely re-identified across websites after only 1, 2, and 3 observations of their topics by advertisers, respectively. This paper shows on real data that Topics does not provide the same privacy guarantees to all users, further highlighting the need for public and reproducible evaluations of the claims made by new web proposals.

April 19: Google’s New Privacy Plan Has Flaws, Regulator Says in Internal Documents. Google’s proposed technology, dubbed Privacy Sandbox, leaves gaps that can be exploited to undermine privacy and identify users who should be kept anonymous, the Information Commissioner’s Office wrote in a draft report. Based on what the ICO called systemic industry noncompliance, it is likely companies will use the tech to continue tracking users across different sites, the ICO said.

April 23: Google Won’t Pull Cookies In 2024

April 26: The April 2024 CMA report states, Google does not provide sufficient clarity to individuals regarding how their data is used by the Topics API. To address this concern, Google has agreed to update the Topics API consent interface and to strengthen developer guidance to highlight the requirement to obtain purpose-specific consent prior to calling the API. We are awaiting the results of these updates. The report also mentions that Topics API may have a negative impact on advertisers, as the SSPs they work with could selectively use topics to misrepresent user cohorts that are lower value (e.g. those that use Made For Advertising (MFA) and pirate sites).

June 13: noyb.eu files a complaint with the Austrian data protection authority. Max Schrems, Honorary Chairman of noyb, says, Google has simply lied to its users. People thought they were agreeing to a privacy feature, but were tricked into accepting Google’s first-party ad tracking. Consent has to be informed, transparent and fair to be legal. Google has done the exact opposite.

June 27: Criteo releases Privacy Sandbox Testing Results Show Shortfalls to Meet CMA Requirements. Second, the current version of the Privacy Sandbox also creates an advantage for Google’s advertising business. Our testing showed that Google Ad Manager (GAM) captured the majority of spend in the treatment population, an increase in market share of 360%: from 23% to 83%. This demonstrates a significant increase in publisher reliance on Google for ad revenue.

July 2: Index Exchange releases Insights From Our Privacy Sandbox Testing. Most worrisome, throughout our testing, we’ve observed a 33% decline in CPMs on Sandbox-enabled impressions compared to impressions where a third-party cookie was present. This compares to 36% lower CPMs in impressions without both Sandbox and cookies. So, while the Sandbox APIs did help (+3%), they’re not closing the gap enough at the current scale, available feature set, and level of adoption to prevent significant revenue consequences for publishers.

July 15: Ad execs sound the alarm over Google’s risky Privacy Sandbox terms, ‘It’s in Google’s best interest’: Sources urge more formal Privacy Sandbox legal terms

July 19: Some ad tech vendors are pulling back from Google’s Privacy Sandbox amid uncertainty

July 22: Third-party cookies will be a user choice, Google says. In A new path for Privacy Sandbox on the web, Google announces that Instead of deprecating third-party cookies, we would introduce a new experience in Chrome that lets people make an informed choice that applies across their web browsing, and they’d be able to adjust that choice at any time. Google’s latest Privacy Sandbox gambit could pit user choice against tracking

July 26: W3C TAG releases Third Party Cookies Must Be Removed.

November 11: Final CMA report released. At 114 pages, many privacy and competition issues were unresolved, or substantially unchanged from previous reports. (Even a very basic one, search. “We consider that Google’s assurance that a site’s decision to support (or not to support) the Topics API will not influence its Google Search ranking should also extend to the other Privacy Sandbox tools.”)

December 6: First privacy lawsuit to cover a “Privacy Sandbox” technology? A class action case about user tracking by Kaiser Permanente included Google’s Topics API along with other tracking methods.

December 18: Google announces changes in advertising policies to allow fingerprinting, starting in February 2025. Covered in Biggest Privacy Erosion in 10 Years? On Google’s Policy Change Towards Fingerprinting by Lukasz Olejnik.

December 19: The Information Commissioner’s Office in the UK publishes Our response to Google’s policy change on fingerprinting the following day.

2025

February 19: Google Admits Privacy Sandbox Doesn’t Work. According to their own numbers, as many as 85% of advertising conversions reported by the system were inaccurate by 60-100% when compared to the same conversions as measured by cookies.

April 22: Google Chrome will now continue to use third-party cookies. In a shocking development, Google won’t roll out a new standalone prompt for third-party cookies in Chrome, meaning ad tech companies can still use the targeting technology in the world’s most popular web browser. From Google: Next steps for Privacy Sandbox and tracking protections in Chrome

June 25: Can Privacy Technologies Replace Cookies? Ad Revenue in a Field Experiment by Zhengrong Gu, Garrett Johnson, Shunto Kobayashi “We find that removing third-party cookies reduces publisher revenue by 29.1%, while Privacy Sandbox preserves just 4.2% of this lost revenue. We further document that Privacy Sandbox increases ad latency and reduces impression delivery by 2.9%.” Covered in No Shade, But Does Anyone Still Care About The Privacy Sandbox?

October 17: Google’s Privacy Sandbox Is Officially Dead, Google ‘Released’ From CMA Privacy Sandbox Commitments (even as Google’s in-browser adtech work continues: a Terminator ending for Google “Privacy Sandbox”?)

2026

January 14: The class action against Kaiser Permanente for “Privacy Sandbox” and other tracking was settled for $46 million. (This may mean that the only companies to make any money from “Privacy Sandbox” were two law firms: Kessler Topaz Meltzer & Check LLP, and Carella Byrne Cecchi Brody & Agnello, PC.)

April 13-17: Inferring Users’ Demographics and Sensitive Interests Using the Topics API by Athicha Srivirote, Muhammad Abu Bakar Aziz, Jeffrey Gleason, Desheng Hu, and Christo Wilson. In this study, we use browsing history data from 782 US residents to examine whether Chrome’s Topics API leaked six demographic traits and 13 sensitive interests. We use a high-fidelity simulator to generate the topics that would have been assigned to these participants. Our study is the first to show that ML models trained on topics data can predict unseen demographics and sensitive interests.

We’re not quite done. A similar proposal that would add privacy risks to browsers is still under discussion at W3C. To be continued: a Terminator ending for Google Privacy Sandbox?

Thanks to Brian May for reviewing an earlier version of this timeline and providing helpful suggestions and corrections.