Gmail automated scanning is software processing of message content for spam filtering, inbox categorisation and smart features. Google states that staff reading of mail is limited to consent and security cases and that Gmail content is not used to train Gemini. The November 2025 dispute about those activities concerned rewritten settings wording, not a newly introduced scan, according to Google and to the corrected Malwarebytes report.
Automated processing in Gmail
Automated processing is the work software performs on messages for delivery, protection and personalisation. Google stated in a post dated 3 July 2018 that Gmail protections prevent more than 99.9% of spam and phishing emails from reaching the inbox, and that automatic processing of emails delivers those protections and intelligent features such as Smart Reply. The same post stated that Google does not process email content to serve ads, and that ads shown in consumer Gmail are not based on the content of messages. The ads position changed earlier: Google announced on 23 June 2017 that consumer Gmail content would not be used or scanned for ads personalisation after that change, with ads instead based on account settings, a change reported at the time by TechCrunch.
Smart features are the optional functions that use message content to personalise the inbox. The Gmail, Chat and Meet setting covers automatic email filtering and categorisation into Primary, Social and Promotions, Smart Compose suggested text, Smart Reply suggested quick replies and summary cards above messages for travel, package tracking and related items. The Workspace setting covers features across Workspace, including events from Gmail shown in Calendar and Gemini summaries, drafts and searches for key information. The other Google products setting covers features outside Workspace, including restaurant reservations and takeaway orders in Maps and suggested tickets, passes and loyalty cards in Wallet. Those three settings are described in detail in Google Support.
Human reading
Human reading is separate from automated processing, and Google stated its limit in the same 2018 post. The post stated: "no one at Google reads your Gmail, except in very specific cases where you ask us to and give consent, or where we need to for security purposes, such as investigating a bug or abuse." That sentence covers ordinary Gmail reading. A later update note added to the same post on 16 October 2026 describes a narrower case: features that let account holders choose to connect Workspace apps to other Google services process data from Workspace apps to personalise other Google experiences when those features are enabled, and depending on settings, that same subset of data can be subject to human review in limited circumstances. The note covers connected features chosen by the account holder, not ordinary Gmail reading.
Gemini training
Google stated its position on Gemini training in Gmail in a post published on 7 April 2026. It was updated on 10 April 2026. The post stated that Google does not train its foundational AI models, including Gemini, on personal emails, and that access granted to Gemini in Gmail is for isolated tasks such as summarising a lengthy email. The post added that Gemini in Gmail does not retain data, because it processes only what a request asks for and then leaves the inbox: Google.
Google Workspace stated that it does not use customer Workspace data to train or improve the underlying generative AI and large language models that power Gemini, Search and other systems outside of Workspace without permission. The words "without permission" form part of the commitment, because they leave a permission-based exception in place. No independent audit of the no-training claim was found for this article, so the claim rests on Google statements and on Google contractual Workspace wording, not on outside verification: Google Workspace.
Smart feature settings
The three smart feature settings work independently, so turning off one setting does not turn off the others. Google Support stated that the settings update that split the controls did not alter underlying data handling practices or privacy commitments. The Gmail, Chat and Meet setting is changed under the General tab in Gmail settings, while the Workspace settings are changed through the Manage Workspace smart feature settings screen reached from the same tab. When the Gmail, Chat and Meet setting is turned off, the features listed for that setting stop working in those apps, while features with individual controls, such as conversation summaries in Chat, can remain available. When a Workspace setting is turned off, content is no longer used to provide the relevant smart features in future. Learnings developed from earlier improvement processing may persist, according to the support page.
Turning off smart features does not stop all Gmail processing, since basic mail delivery and security protections continue. Account holders who turn off all three settings lose the personalisation tied to each setting, while delivery and security processing remains in place.
November 2025 controversy
The November 2025 controversy began with a post by Dave Jones on 19 November 2025 and a Malwarebytes article published on 20 November 2025. The post warned that account holders had been automatically opted in to AI training on private messages and attachments and had to turn off smart features manually. The Malwarebytes article repeated the training claim. It gave instructions for turning off smart features in Gmail settings and in Workspace settings. Both items treated smart features as equivalent to training a generative AI model.
Google denied the training claim on 21 November 2025. The Gmail account posted four points: no settings had been changed, Gmail smart features had existed for many years, Gmail content was not used to train the Gemini AI model, and Google would be transparent about changes to terms and policies. A Google spokesperson, Jenny Thomson, gave the same denial to The Verge, stating that the company had not changed any user settings and did not use Gmail content to train Gemini. That denial was reported by MoneyControl, by 9to5Google and by TechSpot.
Malwarebytes corrected its article on 22 November 2025. The correction stated that the settings themselves were not new, but that the way Google recently rewrote and surfaced them led many people, including Malwarebytes, to believe Gmail content might be used to train AI models and that account holders were being opted in automatically. The correction further stated that after a closer look at Google documentation and other reporting, that belief did not appear to be correct, and that Gmail scanning for smart features such as spam filtering, categorisation and writing suggestions is part of how Gmail normally works and is not the same as training generative AI models. The corrected article is the primary record of the retraction: Malwarebytes.
Two threads from the episode remain attributed rather than stated as fact. Reporting linked the confusion to a January update that separated personalisation settings for Workspace products from other Google products, with some account holders reporting that previously disabled settings appeared enabled again, while Google's stated position is that no settings were changed. The corrected Malwarebytes article also noted a proposed California class-action lawsuit, reported by Bloomberg, alleging that Google gave Gemini access to Gmail, Chat and Meet without proper consent, and that allegation is unproven. Further reporting of the correction appeared in the Financial Express, NDTV Profit, WinBuzzer and Chrome Unboxed.
Connected features and Gemini app review
Two related review paths are distinct from ordinary Gmail reading. The first path is the connected Workspace features described in the October 2026 update note to the 2018 post, where data from Workspace apps connected to other Google services can be subject to human review in limited circumstances depending on settings. The second path is the consumer Gemini app, not Gemini in Gmail. Google stated in the Gemini Apps Privacy Hub that feedback submitted by account holders, including content from connected apps contained in that feedback, is reviewed by specially trained teams, and that reviewed feedback and associated conversations are retained for up to 3 years disconnected from the account. The same hub stated that temporary chats and chats with Keep Activity turned off are retained for 72 hours. Those statements describe the Gemini app, and they do not change the Gmail statements above: Gemini Apps Privacy Hub.
Email aliases
The distinction between provider processing and address control determines what an email alias can and cannot change. Automated scanning happens on the provider systems whether the address given to a website is a primary address or an alias, because the mail still arrives at the provider inbox for delivery. An alias changes which address outside parties hold, and it allows the owner to pause or replace that address if it leaks, but it does not prevent the provider hosting the destination inbox from processing the mail. That limit applies to any forwarding setup, not only to AliasFleet.
AliasFleet provides email aliases as managed, persistent forwarding identities, with a separate alias per service so that a leak or a spam source is tied to one address rather than to the primary inbox. The product is not a disposable email service: aliases persist under the owner's control and can be paused or replaced individually. The product does not change how Gmail or any other destination provider processes mail delivered to it, and it does not remove a provider automated scanning, smart features or security processing. Its function is address control at the point where an address is handed to a third party. That is a different problem from provider processing.
Top comments (0)