Feedback Report - 2023 Q3

Quarterly report for 2023 Q3 summarizing the ecosystem feedback received on Privacy Sandbox proposals and Chrome's response.

As part of its commitments to the CMA, Google has agreed to publicly provide quarterly reports on the stakeholder engagement process for its Privacy Sandbox proposals (refer to paragraphs 12 and 17(c)(ii) of the Commitments ). These Privacy Sandbox feedback summary reports are generated by aggregating feedback received by Chrome from the various sources as listed in the feedback overview , including but not limited to: GitHub Issues, the feedback form made available on privacysandbox.com , meetings with industry stakeholders, and web standards forums. Chrome welcomes the feedback received from the ecosystem and is actively exploring ways to integrate learnings into design decisions.

Feedback themes are ranked by prevalence per API. This is done by taking an aggregation of the amount of feedback that the Chrome team has received around a given theme and organizing in descending order of quantity. The common feedback themes were identified by reviewing topics of discussion from public meetings (W3C, PatCG, IETF), direct feedback, GitHub, and commonly asked questions surfacing through Google's internal teams and public forms.

More specifically, meeting minutes for web standard bodies meetings were reviewed and, for direct feedback, Google's records of 1:1 stakeholder meetings, emails received by individual engineers, the API mailing list, and the public feedback form were considered. Google then coordinated between the teams involved in these various outreach activities to determine the relative prevalence of the themes emerging in relation to each API.

The explanations of Chrome's responses to feedback were developed from published FAQs, actual responses made to issues raised by stakeholders, and determining a position specifically for the purposes of this public reporting exercise. Reflecting the current focus of development and testing, questions and feedback were received in particular with respect to Topics, Protected Audience, and Attribution Reporting APIs.

Feedback received after the end of the current reporting period may not yet have a considered Chrome response.

Glossary of acronyms

CHIPS: Cookies Having Independent Partitioned State
DSP: Demand-side Platform
FedCM: Federated Credential Management
FPS: First Party Sets
IAB: Interactive Advertising Bureau
IDP: Identity Provider
IETF: Internet Engineering Task Force
IP: Internet Protocol address
openRTB: Real-time bidding
OT: Origin Trial
PatCG: Private Advertising Technology Community Group
RP: Relying Party
SSP: Supply-side Platform
TEE: Trusted Execution Environment
UA: User Agent string
UA-CH: User-Agent Client Hints
W3C: World Wide Web Consortium
WIPB: Willful IP Blindness

General feedback, no specific API or Technology

Feedback Theme	Summary	Chrome Response
Ecosystem readiness	SSPs highlighted a concern with publishers not being ready and not doing the required deployment work.	Privacy Sandbox has outreach focused specifically on educating publishers, which includes dedicated webinars and meetings with both publishers and SSPs present to drive deployment work.
Third-party cookie deprecation	Concerns over third-party cookie deprecation (3PCD) ramp up in Q4 2023 due to industry tech blackout.	The timeline for the Privacy Sandbox has been discussed with the CMA, with sequencing leading to a second half of 2024 readiness. Privacy Sandbox will publish more detailed information on the sequencing of ramping up 3PCD. Under the Commitments, 3PCD is subject to the CMA's competition concerns being addressed.
Google Ad Manager	Google Ad Manager refuses to expose the API surface making testing difficult.	Response provided by Google Ad Manager:For the reasons explained in this response by Google Ad Manager , Google Ad Manager's plans for its Protected Audience API integration do not include supporting Google's publisher ad server without control of the top-level auction.
Google Ad Manager	Google Ad Manager has a secret floor price that is only exposed to AdX or Open Bidding SSPs.	Google Ad Manager's public documentation says that the winner of the contextual auction is passed to the top-level scoring logic and not to any component auction, including AdX or Open Bidding. Furthermore, that documentation says of the top-level scoring logic: "Ad Manager will compare the winning bid of each component auction, including Ad Manager's own component auction for interest group bids of its buyers, as well as the best contextual ad (which is selected via dynamic allocation), and will serve the ad with the highest bid."
Google Ad Manager	Google Ads products should be subject to the same rules as third-parties' ads products.	Google Ads products are already subjected to the same rules as third parties.
Chrome-facilitated testing	Add labels for browsers not in A or B.	We are not considering doing so at this time, as our investigation has found that adding non-experiment labels may complicate privacy concerns around traffic in incognito mode.
Advertising agency	Can agencies or companies without JavaScript on websites use Privacy Sandbox APIs?	Anyone can call the Privacy Sandbox APIs. If an agency or anyone else wants to build technologies directly on the APIs they can. Client-side APIs require integrating with the client, just as cookies do. Many of the APIs, like cookies, also have an HTTP header interface. We've already seen one ad industry framework, Prebid, build client-side integrations with the APIs. Other organizations could do the same.
Client-side Solutions	Why is Google adopting client-side solutions for Privacy Sandbox when an engineer has previously expressed concern on the scalability of such solutions in 2012?	Privacy-enhancing technology (PET) as a field of study has evolved significantly since 2012 and, with it, commercially viable applications. At the core of Privacy Sandbox are combinations of PETs which wouldn't have been feasible over a decade ago. In addition, personal computing power has increased, as have consumer expectations of browsers and regulatory expectations of privacy.
Machine Learning	What is Google's planned usage of Privacy Sandbox for machine learning purposes?	Much of the ad tech ecosystem uses machine learning today and we do not expect that to change. Privacy Sandbox does not prevent ad tech companies or anyone else from continuing to use machine learning. Nor does Privacy Sandbox require that companies integrating with its APIs use machine learning. It is reasonable to expect that companies will continue building products and services in ways that meet the needs of their customers, whether that includes machine learning or not. Any machine learning that Privacy Sandbox integrators do build will obviously be known to them and thus not be obscured to them.
Data verification	How can companies verify that the data they receive from using the Privacy Sandbox is accurate and is Google willing to be reviewed via an entity such as the Media Ratings Council (MRC)?	Privacy Sandbox APIs are built within the open-source platform that powers Chrome. The portions of the APIs meant to run in Trusted Execution Environments are also open source and auditable. Anyone who wants to inspect the code can, including MRC.
(Also reported in previous quarters) Production Support	What is the process in place for Chrome to support Privacy Sandbox technical issues and escalations affecting the ecosystem?	Google provides a range of channels to allow ad techs to report technical issues and enable any necessary escalations to resolve such issues. In addition, Chrome expects to further build and scale a process to resolve technical issues and escalations affecting the health of the ecosystem. Chrome is committed to ensuring resources for this effort. Please see our developer post for more information on the public and private forums for feedback and escalation.
Chrome-facilitated testing modes	More information about the timelines and exact implementations for the Chrome-facilitated testing modes.	We have shared a blog post about testing modes and are working to share more information soon. We are welcoming suggestions for what size the testing mode labels should be.
Integration with other industry standards	Will the Privacy Sandbox APIs connect to either or both TCF v2.* and Consent Mode?	We do not have plans to integrate Privacy Sandbox APIs directly with TCF v2 or Consent Mode. However, companies and industry trade groups are welcome to adapt their products and frameworks to work in conjunction with Privacy Sandbox APIs. For example, with frameworks like TCF, each participant must determine its own compliance approach based on the TCF signal it receives and the associated TCF policies. We expect companies to determine when and how to use various functionality our Privacy Sandbox building blocks offer.

Enrollment & Attestation

Feedback Theme	Summary	Chrome Response
Restriction	Enrollment process means Google can decide which company in the ecosystem is allowed to use Privacy Sandbox APIs.	The Enrollment and Attestation process essentially entails verification of the entity (for example, the entity has a DUNs number, can provide a link to a privacy policy, and so forth) and makes the public attestation a requirement for calling the APIs. Entities that can successfully fulfill the enrollment requirements will be validated. For companies that do not have a DUNs, we are providing an expedited, complimentary process with Dun & Bradstreet to acquire one. The objective is to enhance privacy protections of the APIs (by the measures just mentioned) and also to add a layer of transparency to the Privacy Sandbox APIs, so interested parties can better understand who is using which API and what attestations they are making. We are open to further industry feedback on this issue, which has already been used to shape the process.
Re-enrollment overhead	Attestation file expires every 12 months and requires websites to re-enroll.	We've heard feedback from the ecosystem and amended our approach accordingly. This means that files will no longer expire after 12 months or any set period of time. We are updating our enrollment developer guide with additional context.
Attestation file	How is the attestation file used?	All companies calling relevance and measurement APIs will be required by the enforcement deadline to upload the attestation file on their site and keep it for public view as long as you are intending to continue calling the APIs. Websites could expect approximately one request per hour from Privacy Sandbox, and other potential entities may query as well. This will be conducted via the enrollment system's own mechanism to query enrolled entities' servers and ensure the attestation file is valid. Attestations will be included in Transparency Reports and viewable by the general public. We expect companies to act in accordance with their stated attestations, as will the rest of the ecosystem and relevant regulatory bodies.
Enrollment	Is enrollment per site or per origin?	Enrollment is at the site level.

Show Relevant Content & Ads

Topics

Feedback Theme	Summary	Chrome Response
Performance	Performance concerns on the impact of Topics opt-in rate in the European Economic Area.	We would suggest to concerned stakeholders to contact your relevant Data Protection Authority about this issue. They are best placed to address such concerns and influence whether applications of privacy-enhancing technologies are incentivized by laws or instead treated like tracking, requiring the same approaches to consent. The latter could result in APIs like those in Privacy Sandbox not being available as often.
Enrollment	Do downstream bidders need to enroll in Topics API to use Topics signals from upstream SSPs?	The downstream receivers of topics beyond the initial Topics API caller do not need to be enrolled, though many are likely to be enrolled for other API usage. A list of Privacy Sandbox enrollees will be provided programmatically as part of the program's transparency efforts, which would allow an interested caller of the Topics API to check if the recipient they are sending a topic to is enrolled, if the caller should want to.
Topics filtering	Request to apply another caller's filtering to the topics that they retrieve on the page, in order to share only what buyers are eligible to retrieve.	We are considering this request and welcome additional feedback from the ecosystem.
Site exclusion	Exclude websites from contributing to a user's Topics.	Topics are not called by default. It's important to note that no page content is taken into account when topics are selected, and all topics are curated to make sure they are not sensitive. A website can also restrict their site from being included in topic calculation via the following permission policy header: `Permissions-Policy: browsing-topics=()`
Topics observation	Allow publishers to give permissions for Chrome to classify topics based on page content (for example, head or body).	We previously considered offering functionality to classify sites into topics based on page content, and made the decision not to move forward based on privacy and security concerns. This proposal may mitigate some of those concerns, but it's unclear as to what extent. Due to the upcoming CMA experiment period, we don't expect this change to occur before 3PCD. We welcome additional feedback here .
Topics observation	Provide more fine-grained permission policies for publishers.	Providing more fine-grained permission policies for publishers would enable publisher sites to negatively impact the utility of the Topics API for the ecosystem as a whole, without it negatively impacting the utility of the Topics API for the site itself. Refer to the Update permissions policy to support separate permissions for retrieve and observe GitHub issue for a more detailed discussion of the topic.
Medical and Health Topics	Why does the Topics taxonomy not cover topics in Medical or Health categories?	Medical and health categories are considered sensitive topics and thus excluded from the Topics taxonomy.
Topics retrieval	Faster way for DSPs to get Topics without fetching using headers.	The header methods are more performant and less costly than creating a cross-origin iframe and making a `document.browsingTopics()` call from it. (A cross-origin iframe must be used for the call, because the top-level context to observe a topic must match the context from which topics are accessed.) This was discussed in detail here .
Topics retrieval	Requests to support passing Topics via headers on cross-origin script tag requests.	From a security perspective, this isn't possible. Each document and its execution environment are associated with a single origin–that of the document. Third-party subresources loaded and executed within that same environment are considered to be owned by the origin of the document. This is to prevent unconsented data leakage from one origin to another. An alternative is to provide a `browsingTopics` attribute on `<script>` tags. This should be clean from a security perspective, and not add additional latency. We are open to feedback from interested parties.
Awareness	Improve public awareness of Topics API and how the API will be used.	We've engaged with the stakeholder who provided this feedback and this issue was resolved on GitHub . Going forward, we'll continue supporting ecosystem understanding of the API and we look forward to hearing views from stakeholders. In the meantime, we suggest stakeholders wanting to know more about the Topics API familiarize themselves with the documentation in the Chrome developer guide .
Notification	Notification to alert user when their Topics are being observed by a website.	We addressed this feedback on GitHub . Users can learn more about Topics controls in the Chrome help center .
Machine Learning	How ML can be used to infer user Topics?	We are discussing this issue and welcome additional feedback .
Usefulness for different types of stakeholders	Smaller ad tech companies may not be able to observe Topics due to the way browsers calculate them.	Only ad techs that observed the user visit a page about the topic in question within the past three weeks will receive a topic. If the ad tech did not call the API in the previous three weeks for that user on a site about that topic, then the returned value will be empty. This feature means that ad techs whose services are used by a larger number of site owners, and therefore have more opportunities to observe a site visit by a given user, may receive more topics than other ad techs. This feature is essential for the privacy protections of the API as it limits the availability of information about a user to only those parties who are already able to observe the same underlying information (currently via third-party cookies).
XHR Request	When will Topics inclusion in `XMLHttpRequest` (XHR) requests be deprecated?	As Chrome announced in August 2023 , Chrome began deprecating support for XHR when transitioning from Origin Trial to General Availability. As the ramp-up of Topics progressed, XHR support was only included for users for whom the OT features were enabled and was fully deprecated when the individual OT experiment groups were merged. If you were using Topics with XHR, your sites will not break. The topics just won't be added to your XHR request headers. We recommend that you either transition to `fetch` for your request, use the iframe attribute, or use the JavaScript API to retrieve topics. Fetch is supported by all modern browsers, but not Internet Explorer or Opera Mini.
Taxonomy and classifier update process	More information on the Topics taxonomy and classifier release cadence and how companies can prepare for such updates.	Our response remains unchanged from Q2: As shared in the recent blog post , we expect the taxonomy to evolve over time, and for governance of the taxonomy to eventually transition to an external party representing stakeholders from across the industry. We also shared the ramp-up plan in the topics-announce group .
Abuse	Potential attack via redirect chain.	We are considering this issue and welcome additional feedback.
Publisher Inventory Types	What types of publisher inventory will Protected Audience and Topics testing support?	Neither Protected Audience nor Topics are inherently restrictive in terms of the types of inventory they can be used on.
Ramp-up time	Recommend no ramp-up time for new taxonomies to get to 100%.	Following this feedback request from the ecosystem and discussion during PATCG meetings, we have announced our plan for the rollout of the new taxonomy .

Protected Audience API (formerly FLEDGE)

Feedback Theme

Summary

Chrome Response

Top-Level Auctions

Ability to use Google's publisher ad server without also giving Google Ad Manager control of the top-level Protected Audience API auction.

Response provided by Google Ad Manager:
Google Ad Manager's plans for the Protected Audience API do not include supporting Google's publisher ad server without the control of the top-level Protected Audience auction, for the following reasons.

In order to properly serve our customers in the publisher ad serving market, Google's publisher ad server needs to retain control of the top-level Protected Audience auction. As a publisher ad server, our role is to provide publishers forecasting so they can negotiate direct sold campaigns without overbooking, and to pace and deliver their direct reservations optimally. Doing this requires running the final auction to compare all eligible direct and indirect demand.

Forecasting and pacing are core functionalities that publishers expect from an ad server. Without accurate forecasting, publishers may end up overselling their inventory, which puts their business reputation at risk. Pacing is also critical, as being unable to fulfill reservation contracts with advertisers also risks damage to the publisher-advertiser direct relationship, which could result in significant impact to a publishers business.

In short, therefore, we do not view a publisher ad server's activity of running the top-level Protected Audience auction as distinct from the other activities of the publisher ad server.

directFrom
SellerSignals

directFrom
SellerSignals

allows Google Ad Manager to prevent the publisher from seeing the price of its contextual auction.

Chrome response:
Information passed into runAdAuction() is not known to come from the seller unless the seller calls runAdAuction() from its own iframe. In a multi-seller auction it becomes impossible to have all sellers create the frame calling runAdAuction() . directFromSellerSignals addressed this issue by loading content from a subresource bundle loaded from a seller's origin. This ensures that the authenticity and integrity of information passed into an auction from the seller-auctions configurations cannot be manipulated. If publishers want to use Protected Audience API to understand any of the information their technology providers are passing into Protected Audience auctions, they can ask those technology providers for this functionality.

Response provided by Google Ad Manager:
We have maintained a strong focus on auction fairness for years, including our promise that no price from any of a publisher's non-guaranteed advertising sources, including non-guaranteed line item prices, will be shared with another buyer before they bid in the auction, which we then later reaffirmed in our commitments to the French Competition Authority .

For Protected Audience auctions, we intend to keep our promise by leveraging directFromSellerSignals , and not share the bid of any auction participant with any other auction participant prior to completion of the auction in multi-seller auctions. To be clear, we won't share the price of the contextual auction with our own component auction either, as explained in the Further clarify top-level auction dynamics update.

Information Exposure

Sensitive business logic and contractual details may be exposed by the browser.

The person using a web browser can see everything that is happening in the browser. When an ad auction happens inside the browser, it is true that the person whose browser it is could watch that auction take place, including seeing how much different parties choose to bid. Since a browser is the user's agent, we do not think it is possible or desirable to try to change this. Only the person using the browser has visibility into these operations, however. An on-device auction run using the Protected Audience API is not observable to any servers, including Google's.

PerBuyerExperiment
GroupId

Current value range of

PerBuyerExperiment
GroupId

could allow buyers to correlate the contextual data with the trusted server request.

Using the Protected Audience API in this way is inconsistent with Privacy Sandbox's mandatory attestation that API users will not try to circumvent the Privacy Sandbox protections. In the future, the requirement that key-value servers run in trusted execution environments (TEEs) will provide technical protection against this attack.

Same-origin policy

Relax the same-origin policy to allow for subdomains.

We are considering this request and welcome additional feedback from the ecosystem.

API versioning

Request for versioning and release notes for changes to the Protected Audience API.

We are considering this request and welcome additional feedback from the ecosystem.

Multi-SSP Auctions

Allow top-level auction signals to perform JSON merges with component signal auctionSignals .

We are considering this request and welcome additional feedback from the ecosystem.

Bid limit

Increase the limit on the number of ad components entering the bid from 20 to 40.

We are considering this request and welcome additional feedback from the ecosystem on why this would be useful.

(Also reported in previous quarters)
Performance of Protected Audience Auctions

Report from testers that Protected Audience auctions have high latency.

On questions of latency, the Protected Audience API has generally followed the existing standard paradigm of building controls that let sellers decide how much time and resources the bidders can consume, and building tools that let buyers decide how to best use the resources available to them. These controls and tools are generally available today, but their full benefit will only be realized after adoption by buyers and sellers. In addition, Chrome continues to work on a variety of infrastructure improvements to auction speed ( crrev.com/1190815 , crrev.com/1199839 , crrev.com/1201837 , crrev.com/1198339 , crrev.com/1197323 ).

We invite feedback on both halves of this latency effort: new tools that buyers and sellers would find useful, and reports of observed bottlenecks that Chrome engineers should investigate.

Buy-side filtering

Add support for buy-side filtering based on interest groups.

We have suggested several ways in which SSPs and DSPs could change their designs to handle this:

Moving some work into the DSP's Key/Value server.

SSPs creating some contextual signals and giving those to DSPs.

SSPs caching contextual signals for DSPs.

Publisher Interest Group Control

Support for publishers seeking to delegate the use of publisher-created interest groups.

We have engaged in discussions with many parties about the request. We believe that all such use cases involved in "delegating" the publisher-created interest groups can be accommodated now, and furthermore that we should build additional support to make some use cases flow more smoothly in the future.

(Also reported in Q2) Trusted Execution Environments

Support for Trusted Execution Environments (TEE) in non-public cloud environments.

Our response is similar to previous quarters:

While we are continuing to explore support for options beyond public cloud-based solutions, we have no current plans to support on-premise TEEs. At this stage, given Privacy Sandbox security requirements and the significant challenges presented by on-premise deployments, we believe that continuing to expand and improve cloud-based deployments (for example, supporting Google Cloud in addition to AWS) is the most beneficial for the ecosystem. However, we welcome additional feedback on why such a requirement is necessary and feasible given the privacy and security constraints.

Trusted Execution Environment

Components in the TEE serving path, such as the load balancer, can observe all the traffic and have information of the IP address of each request.

Currently IP address is passed as a metadata in request headers to untrusted seller's ad service in the case of both Bidding and Auction and on-device Protected Audience auctions. Refer to Metadata forwarding for more information. In the long term, we plan to proxy ad tech and tracker traffic through an IP Proxy, which will prevent components from observing all the traffic in the serving path.

Time-to-Live (TTL)

Will the time-to-live (TTL) before services have to request new keys be set or is it intended to be flexible (or dynamic)?

The TTL is generally static. Currently, the TTL for the public is 8 days, and the rotation happens every 7 days; the TTL is also the same for private keys in the case of the Aggregation Service. In case of Bidding and Auction services, private and public keys are fetched every N hours in the non-request path and cached in-memory, so that there is no more than an N-hour delay between keys rotating and servers picking up these keys. The 1-day buffer between key rotation and expiry is to ensure that even if the key generation fails, the services can continue operating. We are considering extending the TTL to be more resilient for outages. In case of a key leak, we plan to manually force key generation and invalidate keys sooner. Note that public keys are cached on the clients, currently for 24 hours, again to ensure that in case of coordinator outage, the services can still operate.

Traffic Shaping

Traffic Shaping support for Bidding and Auction Services.

Buyers can indicate, based on Publisher first party data or contextual data, demand for Protected Audience auctions. Sellers can do similar determinations as well in the seller's ad server or Ad Exchange server. The models can be trained on 1P data and any aggregate reports from Protected Audience auctions. Sellers can use this information to avoid sending requests to Bidding and Auction servers when there is no demand for Protected Audience auctions. We believe this can be an effective way to shape traffic.

Component Auction

What top level auctionSignals are shared with Component sellers?

Buyers in a component auction only receive signals from the component seller. We are looking to share documentation around the overall sequence of a combined auction with header bidding and Protected Audience auction soon.

Video Rendering

Support for video rendering using Protected Audience and Fenced Frames.

Protected Audience API supports video rendering using a mechanism that relies on iframes. However, we haven't yet designed a solution that is compatible with Fenced Frames, and this is one of the reasons we had decided to push back Fenced Frames enforcement to 2026. That means if a partner does decide to enforce Fenced Frames now, the support for video would be lacking for that partner.

Frequency capping

(Also reported in previous quarters)
Per-user frequency controls within a campaign and ad group.

Our response is unchanged from the previous reports:

Protected Audience will support frequency capping for on-device auctions and contextual and branding campaigns as well. Shared storage and site-specific caps can also be used for additional frequency capping controls.

Ad Preferences

Does Protected Audience provide a way to opt-out or blocklist by advertiser sites or a way to leave all interest groups from the same owner?

There are several ways for users to block access to the Protected Audience API and other Privacy Sandbox features.

Same-origin policy for source URL of bidding and auction scripts

Relax the requirement that all fields that specify URLs for loading scripts or JSON must be same-origin with the owner.

We are currently considering this request and welcome additional feedback from the ecosystem.

forDebuggingOnly

Potential for

forDebuggingOnly
.reportAdAuctionWin

to be misused if it remains post 3PCD.

Over the past years we have been receiving feedback from the ecosystem regarding functionality gaps in Protected Audience once third-party cookies are deprecated, and we are working to formulate a plan to support them post 3PCD without compromising on the goals of Privacy Sandbox. We welcome any additional suggestions and feedback on missing functionality that the ecosystem would like to see.

Multiple Interest Groups

Use multiple interest groups in the same bid.

This is not supported in Protected Audience API today, as it would result in a change to the underlying privacy model. We welcome additional discussion here .

On-device auctions

Will Chrome on Android support on-device Protected Audience auctions?

Yes, on-device auctions will be supported in Chrome on Android.

(Reported in Q2 2023) Click-related data

Add click-related data to browserSignals.

We continue to evaluate this feature request and welcome additional feedback on why this should be prioritized.

Trusted Execution Environment providers

Are there material differences in the Trusted Execution Environment offerings of different cloud providers?

We are not aware of any major differences, but we recommend the ecosystem review the public deployment guides to see which solution best suits their needs.

Google Cloud .
AWS .

(Reported in previous quarters )

Support for negative Interest Group targeting

An API to support negative interest group targeting: showing ads only if a user does not belong to an interest group.

We are looking into implementing this feature and are discussing the request .

Content Violation

Support features that allow users to report bad ads served by Protected Audience API in Fenced Frames.

We believe that the existing Fenced Frame Ads Reporting mechanism offers good options for ad techs who want a user-generated "Bad Ads" reporting flow. This would allow bad ads reporting in a way essentially unchanged from the industry standard today. We welcome additional feature requests if any gaps remain, including during the time after third-party cookie removal but before Fenced Frame rendering becomes widespread.

Private Aggregation API Reporting

How can we calculate time the user has spent in that interest group?

In Chrome M116+ you should be able to use recency as defined pull/639 .

K-Anonymity server

More information on K-Anonymity server.

We shared more information on K-Anonymity servers here and welcome additional feedback.

Dynamic Creative URLs

Support for creative URLs without pre-declaration while still respecting k-anonymity.

We are discussing this feature request and welcome additional feedback on why this should be prioritized.

K-anonymity requirement

Will k-anonymity requirement on Interest Group updates be re-introduced?

We don't anticipate changes to the position stated in this GitHub post . As announced in that post, we decided to remove the k-anonymity requirement on Protected Audience interest group updates, which does not have a significant impact on the API's overall privacy protections, and we plan to consider other potential more direct protections (such as IP address privacy or a trusted update server) at a later date when the related technologies are more developed, deployed and adopted.

Bidding & Auction Services Beta Testing

When will Bidding & Auction Services Beta testing begin?

As stated in Timeline and roadmap , the first phase of Bidding and Auction Services testing begins in November 2023.

Roadblocking

Request to support Creative coordination for Ad Networks (SSP and DSP are in the same company or properties).

We appreciate the feedback for this use case and we're looking to understand whether more ad techs are interested in seeing this supported. We welcome additional feedback .

Native Advertising

Fenced Frame support for Native Advertising.

We are considering supporting the use case and are discussing possible workarounds and solutions .

K-anonymity

How can I maximize interest group ads that meet k-anon thresholds?

We have shared some tactical guidance on this topic .

POST support

Support for sending auction data via POST requests.

We are evaluating this feature request and welcome additional GitHub issue submissions on why this should be prioritized.

Reporting granularity

What is the reporting granularity of Fenced Frame ad reporting with Ads Composed of Multiple Pieces?

The current design does not allow capturing product ID or position as this may compromise user privacy. Only the reserved.top_navigation can be invoked, which would be sent when there is a user activation (such as a click) on the ad component fenced frame, which results in a top-level navigation.

Ad auction

Can an SSP participating in a component auction trigger another component auction itself?

A componentSeller cannot also include componentAuctions .
The multi-seller auction only has two levels:
1. The component auctions in parallel.
2. The top-level auction (where the winning ad from each componentAuction competes).

Bidding & Auction Services availability

Will Bidding & Auction be available during the Chrome facilitated testing phase?

Bidding and Auction Server will not be available during the Chrome facilitated testing phase.

Bidding signals

Allow browsers to request and delete bidding signals.

We are discussing this request and welcome additional feedback on why this should be prioritized.

generateBid()

Ability to update interestGroup's userBiddingSignals through updateURL .

We are considering this proposal and welcome additional feedback and discussion.

Publisher Inventory Types

What types of publisher inventory will be supported by Protected Audience and TOPICS testing?

Neither Protected Audience nor Topics are inherently restrictive in terms of the types of inventory they can be used on.

Server-to-Server integration

Is direct integration between the SSP and DSP required for Protected Audience?

Direct integration between the SSP and DSP is not required if DSP does not need to process contextual signals in its own server in order to pass that processed information into its on-device bidding function.

A bid_currency field in B&A

Support for bid_currency field in Bidding and Auction Service.

B&A doesn't support a bid_currency yet, although we plan to support that by the end of January 2024. Refer to the timeline here .

perBuyerSignals

Is there a size limit for perBuyerSignals ?

There is no limit on the number of per-buyer signals, but sending too much data may have detrimental effects on the browser's performance.

Cross-site use cases

Can we use Protected Audience API interest groups across multiple websites?

Protected Audience is not designed for such use cases, as explained in turtledove/issues/282 .

Interest Group HTTP Requests

Include Interest Group Blob in the HTTP headers.

We are considering this request and welcome more feedback on this request.

Ad quality control

Loss of ad quality control related on cross-site information.

We are considering this feedback and welcome additional feedback .

Chrome DevTools

Outgoing Protected Audience network requests should be visible in the Chrome Developer Tools Network Tab.

We are working on enabling this functionality in the network tab and welcome additional feedback on why this should be prioritized.

Trusted Execution Environment

When will the details on which metrics are privacy-impacting (and their degree) be added to the explainer on Trusted Execution Environment monitoring?

We are in the process of updating the explainer with this information. The updated explainer will be available by November 2023.

directFrom
SellerSignals

Why is

directFrom
SellerSignals

not packaged as a web bundle?

We shared the rationale for this decision here .

Impression delegation

Is there any viable way to do impression delegation where the outcome of an interest group being selected is yet another targeting action?

Multiple nested auctions are not compatible with our privacy goals for two reasons. First, when the winner of an auction renders inside a Fenced Frame, our privacy goals for Protected Audience include the resulting creative rendering without knowledge of the context: the surrounding page's URL or first-party cookie are a privacy violation. In that environment, a nested auction is not viable. Second, the Protected Audience model says that each auction's winner should be based on data from just one additional site. Nested auctions would be a way to compound that, resulting in the possibility of choosing ads based on a many-site profile.

Data at Rest criterion

Explain further the Data at Rest criterion in the Key/Value service trust model.

Data in the Key Value Service is loaded into memory and served from there rather than doing any read-through caching.

Buyer Data Signal

Is there a defined size limit for the buyer_data signals received from the DSPs?

There are currently no browser imposed limits for buyer_data signals received from DSPs.

Measure Digital Ads

Attribution Reporting (and other APIs)

Feedback Theme	Summary	Chrome Response
Cross-device	Plan for cross-device support for Attribution Reporting API.	Cross-device presents new privacy challenges on top of 3PC and also adds technology distribution challenges given the range of devices and platforms a user might use. We are exploring potential solutions, but we are focused on the critical use cases currently supported by Attribution Reporting and do not have plans to introduce cross-device support before the removal of third-party cookies.
(Also reported in previous quarters) Trigger Data Size	Why is the trigger data size limited to 3 bits?	The size is limited to 3 bits and 8 distinct values to ensure that the amount of cross-site and cross-context information about a user is limited. We welcome ecosystem players to submit feedback on whether the current parametrization for event-level reporting is sufficient.
Conversion funnel	Report multiple domains that were used in conversion.	This use case is possible since the addition of multiple destinations . We welcome additional feedback .
Same domain in different country support	Does Attribution Reporting work with websites that have the same Domain but multiple country TLDs?	This issue has been discussed and resolved with the stakeholder that raised the question. If an ad tech needs to use multiple country TLDs they will need to have multiple enrollments, with one for each country TLD.
Protected Audience and Attribution Reporting	Can ad techs access both view-through conversions for Protected Audience auctions as well as click-through conversions for Attribution Reporting?	Yes, Privacy Sandbox should support both VTCs and CTCs within Protected Audience.
Agaggregatable report delays	Reduce aggregatable report delays further.	We have heard recent feedback regarding this and have shared ideas here . We welcome additional feedback from the ecosystem.
Agaggregatable report delays	Reducing delays via introducing server mediation.	We are considering this proposal and welcome additional feedback .
Event-level report delays	Reduce event-level report delays.	The full flexible event-level proposal, described in Flexible event-level configurations , can reduce event-level reporting delays down to 1 hour with a noise tradeoff.
Source reporting origin per source	Limitation of max source reporting origins per source reporting site prevents ad techs from registering sources from different reporting origins for a single publisher origin.	This has been discussed with the stakeholder that raised the issue and a potential solution of using 1 reporting origin per source-reporting site is being tested before trying other potential solutions involving redirects. We are open to any additional ecosystem feedback regarding this limit as well.
Issue reporting	How can we report errors or issues with the Attribution Reporting API to Chrome?	Currently we recommend ad techs report any Attribution Reporting API errors they may be facing as an Issue on GitHub. If they are facing a Chrome-related issue we recommend creating a Chromium bug. Links for how and where to flag any issues can be found in Engage and share feedback .
Deduplication	How can we deduplicate conversions across different pipelines and devices?	Deduplicating across devices and measurement pipelines is a known and current challenge that ad techs also face today with 3PCs. With the Attribution Reporting API, ad techs can decide when to register specific conversions and add specific metadata to indicate which measurement pipelines they have used to track the conversions (in other words, part of the aggregation key), which can be compared against other measurement pipelines. We are open to any additional ecosystem feedback regarding this.
Deduplication and Priority	Request to have priority first before deduplication.	We are considering this request and welcome additional feedback .
Anti-fraud	Risk of malicious user tampering the event-level data.	Report verification does not work for event-level reporting for the reasons described in Why doesn't this support event-level reports? .
Conversion type	How can we differentiate between view through and navigation in Attribution Reporting?	We have the following built-in filtering option: `source_type` . Additional details are available here .

Aggregation Service

Feedback Theme	Summary	Chrome Response
Budget recovery	Some ad techs have requested the ability to reprocess reports in cases where there are failures, errors, or deletions of their reports.	The team is exploring ways to address this in a privacy-preserving way.
Site enrollment	Multiple ad techs have requested support for processing multiple origins in the same account for use cases such as splitting data by Geo, advertiser. This behavior is also expected by ad techs given that the client API enrollment is now site-based (and not origin based). Migration from origin to site enrollment streamlines the ad tech onboarding process via consistency with the client enrollment process.	We will be launching migration from origin enrollment to site enrollment for the Aggregation Service soon and welcome feedback from the ecosystem .
Release & Deprecation Plan	Release and depreciation schedule for Aggregation Service features and patches published. The goal of the plan is to give ad techs visibility into our release policies to enable them to prepare for upcoming releases and deprecations, and ensure they run stable and secure versions of services.	We have recently published a proposal for the Aggregation Service release and deprecation plan and welcome additional feedback .
Coordinators	What happens if the coordinators go down on aggregation service?	Both coordinators need to be fully available for the system to function correctly. Short unavailability is accommodated with retries in our client libraries; longer unavailability of either of the two coordinators will have aggregation jobs fail. Jobs can be rerun if the budget for privacy isn't consumed yet. In the case where any service failure led to budget consumption without a summary report written to ad tech storage, we currently recommend they use debug reports to retrieve results using the local testing tool . We are also working on features to allow for budget recovery in the case of failures so ad techs can rerun their jobs.

Private Aggregation API

Feedback Theme	Summary	Chrome Response
Blob Url	Request to support Blob Url in Shared Storage.	Support for Blob Url has been added in Chrome M116.

Limit Covert Tracking

User Agent Reduction and User Agent Client Hints

Feedback Theme	Summary	Chrome Response
JavaScript API	Availability of the User Agent Client Hints JavaScript API.	There are no plans to remove this functionality as it is our core solution for partners who want to actively access the high-entropy data beyond what is available by default in the frozen and reduced UA.
Device and Form Factor information	Ability for websites to understand input, output, and other information the device visiting the website can support.	We have added support for this request following feedback from the ecosystem.

IP Protection (formerly Gnatcatcher)

Feedback Theme	Summary	Chrome Response
Eligible Third Party Traffic	What is "eligible third-party traffic" referring to in the explainer?	We understand the importance of this question and are actively working to identify which third-party traffic will be eligible and which will not. We welcome feedback on this topic.
Network Traffic Audits	Support for enterprises to perform network traffic audits for their networks.	Only third-party traffic embedded in first-party sites will be affected, which should limit the amount of traffic that requires filtering. Additionally, we plan to give users the option of whether or not to use IP Protection, and for enterprise-controlled Chrome, there will be enterprise policies to disable IP Protection. Finally, we're exploring what controls (if any) will be provided to network operators to disable IP Protection. We welcome feedback on this topic.
Access control	IP Protection may impact web services that use IP addresses for access control.	We understand the importance of anti-fraud use cases and the possible impact to those use cases. We are seeking ecosystem feedback on how we can better support anti-fraud use cases that typically have relied on IP addresses.
Communication between the 2-Hop proxies	How to ensure there is no information between proxies.	We are in the process of designing the proxy interactions. Our goal is to minimize the chances for such information sharing via business, process, and technical means.
Non-Google Authentications	Support for Non-Google Authentications.	We plan to publish more details about account authentication in the future, though we have shared some initial considerations .
Tracker classification	How will IP Protection determine what constitutes a tracker and its variants?	We understand the importance of this question and are actively working to identify which third-party traffic will be eligible and which will not. We welcome feedback on this topic.
Analytics	IP Protection may impact the accuracy of analytics services.	We are looking to understand the impact of IP Protection further and welcome additional feedback and examples from the ecosystem.
Proxy	If a user is using proxy or has manually defined a proxy, how will IP Mask work in this case?	We are looking to understand the impact that IP Protection may have on other proxies. We do not have any plans to share at the moment. We welcome feedback on this topic.
Premium offering	Will IP Protection be a paid feature?	IP Protection will be available to Chrome users as part of the core browser experience. It will not be a paid feature.
Proxy server	Will the same proxy servers be used during user sessions?	An HTTP/S connection will use a single pair of proxies and will present a single masked IP address to the origin. Beyond that, there are no hard constraints on different HTTP/S connections having to use the same servers.
Platform support	On which platform will IP Protection be supported?	IP Protection will initially be available on Chrome for Android and Desktop. We continue to evaluate how to expand the protection to other platforms.
Opt-Out	Will users be able to disable IP Protection?	We plan to provide users the choice on whether they want to use IP Protection or not.
Anonymization	What kinds of requests will be anonymized under IP Protection?	HTTP/S and DNS requests to eligible third-party domains are anonymized via the privacy proxies. We will provide additional details in an upcoming explainer on how we will determine which domains will be included. The rest of the traffic (for example, the rest of the DNS requests or other HTTP/S traffic) is unaffected.
Data Visibility	Network addresses may be accessed during the first hop in IP Protection.	In the two-hop proxy model, the first hop (controlled by Google) only sees the source client IP and a request to connect to the second hop, while the second hop (controlled by an external CDN) only sees a tuple on the first hop (proxy IP + port) and the destination IP. For the response back from the origin, the second hop is able to forward the response to the first hop proxy+port associated with the request and doesn't need to learn anything about the original client IP (and the first hop just returns the response to the client, without learning anything about the destination IP). In this way, the first hop only learns the client IP and the second hop, while the second hop only learns the destination IP.
WebView	Will IP Protection be available to Android WebView in the future?	We do not have any plans to share at the moment, but our vision is to provide this protection as broadly as possible.

Bounce Tracking Mitigation

Feedback Theme

Summary

Chrome Response

Interaction Tracking

How are user interactions tracked?

Bounce tracking mitigations track two types of user interactions:

User activations as defined by the html spec . These are basically clicks, key presses, touch screen taps, etc.
Successful webauth assertions. These are cases where a user taps a security key or uses a passkey as form of authentication

These interactions are associated with the top-level site on pages where they occur. For example, if a user clicks in an embedded iframe the interaction is associated with the top-level site and not the embedded site.

The interactions are stored in a database containing the schemeless etld+1 and the time of the interaction.

Interactions protect the associated domain from bounce tracking mitigation state deletion for 45 days.

Allowlisted Exemptions

Can domains be exempted?

We are considering this request and we welcome additional feedback from the ecosystem .

Privacy Budget

No feedback received this quarter.

Strengthen cross-site privacy boundaries

Related Website Sets (formerly First-Party Sets)

Feedback Theme	Summary	Chrome Response
Centralized Approach	Concern over the centralized repository approach for managing Related Website Sets.	A public, easily accessible repository is key to the design of RWS as it provides accountability for submissions. Third-party cookie functionality is ultimately provided by the use of the Storage Access API or the rSAFor API, with RWS membership providing auto-granted access (as opposed to through prompts with the Storage Access API). We believe that an approach like the RWS submission process is an appropriate requirement for auto-granted third-party cookie access.
Renaming json file	With the change in API name, does the hosted JSON file name need to be changed?	Yes, the submission guidelines have been changed, and the primary domain must serve a JSON file at `/.well-known/related-website-set.json` . Existing sets in the RWS list do not need to be changed, but if there are modifications submitted to existing sets, the JSON file must be changed.
(Also reported in previous quarters) Domain Limit	Request to expand the number of associated domains	As announced in a blogpost on August 31, we have raised the associated domain limit to five domains following feedback from the ecosystem. We have decided to increase the associated domain limit to five domains (plus one primary domain) which best matches the most comparable implementation offered by another major browser.
Third Party Cookies	Will Related Website Sets only work with third-party cookies disabled?	Related Website Sets will work even when a user has not blocked third-party cookies; but there will be no observable effect since the relevant cookies are available without any need for Related Website Sets and Storage Access API.
Legitimate edits	How does the Related Website Sets repository prevent non-owners from modifying sets?	Per the submission guides , anyone can submit a PR on GitHub to edit the `first_party_sets.JSON` file. However, if the PR is approved (passes technical validations, and so forth), it will be manually merged in batches to the canonical FPS list once per week (Tuesdays at 12pm Eastern Time) by Google. If a bad actor tries to modify a set they don't own, it shouldn't be a problem since they won't be able to modify the `.well-known` files and therefore the validations will fail.
Domain hijacking	Domain hijacking may expose related domain data to unauthorized parties.	This is not possible, as discussed in this Protected Audience GitHub issue .

Fenced Frames API

Feedback Theme	Summary	Chrome Response
Content Violation	Allow users to report suspicious ads.	Suspicious ad reporting is not prevented by Fenced Frames. Users can still interact with the ad and report suspicious ads to the ad tech in the usual way.
Interaction with surrounding sites	Allow interaction with the surrounding or top-level website.	We are looking to understand why this request is necessary and welcome additional feedback from the ecosystem.
Native Advertising	Fenced Frame support for Native Advertising.	We are considering supporting the use case and are discussing possible workarounds and solutions .

Shared Storage API

Feedback Theme	Summary	Chrome Response
Cross domain	Allow communication across domains for local storage.	This use case is currently not in line with Shared Storage's privacy-preserving output gates but we welcome additional context as we evolve proposals for non-partitioned storage.
Blob Url	Request to support Blob Url in Shared Storage.	Support for Blob Url has been added in Chrome M116.

CHIPs

No feedback received this quarter.

FedCM

Feedback Theme	Summary	Chrome Response
Third-party cookies	Is FedCM currently disabled if users enable "Block third-party cookies" in the Chrome settings"?	Yes, FedCM is currently disabled. For testing, we recommend that you additionally enable `chrome://flags/#fedcm- without-third-party-cookies` . We are looking to support FedCM without third-party cookies in the future.

Fight spam and fraud

Private State Token API (and other APIs)

Feedback Theme	Summary	Chrome Response
Token expiration	Once Google Chrome is uninstalled, will the Token be lost or will it be cached?	The token will be lost if the user uninstalls Google Chrome.
Token Information	How can issuers keep issued information within the Private State Token private?	Information is always kept private in the token and cannot be unencrypted by external parties that do not have the keys.
Error in demo	Error when trying to run the Private State Token demo.	We have updated the demo and it should be working correctly now.