·  ·  · 

Gmail and Gemini AI Controversy Explained: Is Google Training on Your Emails?

November 24, 2025
by
Gmail and Gemini AI Controversy Explained: Is Google Training on Your Emails?

Updated for November 24, 2025 – what’s really happening inside Gmail, and how to take control of your data.

1. What happened – and why everyone is suddenly worried about Gmail

Over the past few days, millions of Gmail users have been told that “Google’s AI is reading your emails and attachments to train Gemini – unless you turn it off.”

The panic started when:

  • Security firm Malwarebytes published a blog post on November 20 claiming Gmail could read your messages and attachments to train Google’s AI models, apparently with users auto‑opted in. The post went viral on X and Reddit. [1]
  • Tech outlets and commentators amplified the claim, tying it to a class-action lawsuit in California that accuses Google of giving its Gemini AI access to Gmail, Chat and Meet without proper consent. [2]
  • Earlier ZDNet coverage framed the change as Google “snooping on your emails without your permission,” before following up with a clarification piece titled “Google denies analyzing your emails for AI training – here’s what happened.” [3]

On social media, that all condensed into a scary, oversimplified message: “Gmail is now training AI on all your emails.”

Today (November 24, 2025), multiple outlets – and Google itself – are trying to put that genie back in the bottle.

2. Malwarebytes walks back its claim – but keeps a warning

Malwarebytes has now corrected its original article and explicitly acknowledges that it helped create a “perfect storm of misunderstanding.” [4]

Key points from its updated piece:

  • The underlying Gmail “smart features” settings are not new – but Google recently rewrote the wording and surfaced them more prominently in the UI.
  • That vague new language, plus the timing of the lawsuit, led Malwarebytes (and many others) to believe Gmail content might be used to train Google’s generative AI models and that people were being auto‑opted in.
  • After rereading Google’s documentation and reviewing other reporting, Malwarebytes says that doesn’t appear to be the case.
  • However, Gmail does scan your emails and attachments to power things like spam filtering, categorisation, and writing suggestions – and those settings can be turned off if you’re uncomfortable.

So Malwarebytes’ revised view matches what Google is saying publicly: Gmail scans your emails for features, but not (according to Google) to train Gemini’s core AI models. [5]

3. Google’s official line: “Misleading reports” and no Gmail content in Gemini training

Google has now responded through:

  • Statements to The Verge [6]
  • A widely shared post on X from the official Gmail account, repeated by outlets like NDTV and Business Standard [7]
  • Follow‑up coverage across tech media, including TechRadar, BGR and others. [8]

Google’s core message can be summarised as:

  • It has not changed anyone’s Gmail settings in secret.
  • Gmail Smart Features have existed for many years.
  • The company says it does not use Gmail content to train its Gemini AI model. [9]

Instead, Google insists that:

  • When Smart Features are on, Gmail analyses your messages to provide conveniences like:
    • Spam filtering
    • Tabbed inbox categories (Primary, Social, Promotions)
    • Smart Compose and Smart Reply
    • Automatic extraction of travel bookings and order confirmations into Calendar or other Workspace tools [10]
  • That content analysis is used to power those specific features, not to feed Gemini’s training data.

From a user perspective, that means Gmail is definitely scanning your email – but Google’s position is that this scanning is for product features, not for general AI model training.

4. The lawsuit: why the stakes feel higher this time

Separate from the media misunderstanding, Google faces a serious legal challenge.

A proposed class‑action lawsuit, Thele v. Google LLC, filed November 11 in federal court in San Jose, alleges that: [11]

  • Around October 10, 2025, Google allegedly “secretly turned on Gemini” for all users’ Gmail, Chat and Meet accounts.
  • The complaint claims this enabled Gemini to “track private communications” — including every email, attachment and chat — without users’ knowledge or consent.
  • Plaintiffs argue this violates California’s Invasion of Privacy Act, which restricts unauthorised recording or interception of confidential communications.

News outlets including Bloomberg, Yahoo and TechInAsia have all covered the lawsuit, which has helped fuel the public sense that “something changed” in October, even as Google denies altering user settings. [12]

It’s important to stress:

These are allegations, not findings of fact. The case is just beginning, and no court has ruled on whether Google did anything unlawful.

Still, the lawsuit dovetails with renewed scrutiny of how tech companies use private data to train AI. A joint Al Jazeera/PolitiFact explainer today notes that the US still lacks a comprehensive federal privacy law for AI training, making opt‑out rights patchy and confusing compared to Europe. [13]

5. Smart Features vs Gemini vs AI training: three different things

A lot of the panic comes from blending three separate mechanisms:

5.1 Gmail Smart Features (built‑in automation)

  • These are the classic Gmail conveniences:
    • spam filtering
    • category tabs
    • autocomplete / Smart Compose
    • Smart Reply suggestions
    • trip cards, order tracking, etc. [14]
  • To do any of that, Gmail must read and analyse your messages – that’s been true for years.
  • Depending on your region and account, these features may be on by default (opt‑out). Several reports, including Politifact/Al Jazeera and NEWS.am Tech, say this is the default for many US users, even though Google’s help pages frame them as opt‑in. [15]

5.2 Gemini in Gmail and Gemini Deep Research (AI helpers)

Google now offers Gemini‑powered tools inside Gmail and across Workspace:

  • Gemini in Gmail can draft replies, summarise threads, and pull context from Drive files. [16]
  • Gemini Deep Research, announced November 5, can connect to Gmail, Docs, Drive and Chat to perform long‑form research, summarise content and generate reports. [17]

When you explicitly use these tools, Gemini has to read the relevant messages or files to respond to your prompts. How that data is later used depends on your Gemini privacy settings (see Section 7).

5.3 Training Google’s AI models

According to the Gemini Apps Privacy Hub, Google uses data from Gemini apps – including user prompts, generated content and information from connected apps – to maintain, improve and develop its AI models, unless you restrict that via the “Keep Activity” and related controls. [18]

Crucially:

  • Google consistently says “we do not use your Gmail content to train our Gemini AI model” when people simply have Smart Features turned on. [19]
  • However, when you use Gemini apps or connect Gmail as a data source, information flowing through Gemini is treated under the Gemini privacy policy – which explicitly allows some data to be used to improve Google’s AI unless you opt out in Gemini settings. [20]

So the picture looks like this:

Gmail Smart Features ≠ Gemini training dataset
But Gemini features that read Gmail may use that data under Gemini’s own privacy rules unless you limit it.

That nuance is exactly what’s getting lost in panicked social posts.

6. What today’s coverage (24 November) is saying

A wave of articles published today tries to clarify the situation:

  • TechRadar runs with “Panic over – Google says your Gmails aren’t being used to train Gemini,” explaining that the earlier scare came from misreading Smart Features settings. [21]
  • Business Standard and NDTV Profit both report Google’s statement that stories about Gmail emails being used to train Gemini are “misleading” and that Smart Features do not feed model training. [22]
  • NEWS.am Tech in Armenia emphasises that Gmail still scans emails for Smart Features, that these are usually enabled by default, and that scanning now overlaps more with Gemini‑powered experiences – prompting privacy experts to still recommend turning the options off if you handle sensitive data. [23]
  • Other outlets (BGR, Tom’s Guide, local explainers and TV segments) largely echo Google’s stance while still advising users to review their privacy settings. [24]

Taken together, today’s consensus is:

The most viral claim – that Google secretly flipped a switch to use all Gmail content for Gemini training – is not supported by current documentation or Google’s public statements. But Gmail does scan your emails extensively for Smart Features, and Gemini has new ways to tap into that data when you opt into its tools.

7. How to audit your Gmail Smart Features and Gemini settings

If you only remember one thing from this article, make it this:

You can and should review what Gmail and Gemini are allowed to do with your data.

7.1 Turn off Gmail Smart Features (desktop)

Based on Google’s help pages, Malwarebytes’ walkthrough and today’s news explainers, here’s how to reduce Gmail’s automated scanning for Smart Features on a personal account: [25]

  1. Open Gmail in a desktop browser and sign in.
  2. Click the gear icon (Settings) in the top‑right corner.
  3. Click “See all settings.”
  4. In the General tab, scroll until you find a section labelled something like “Smart features and personalisation” or “Smart features in Gmail, Chat and Meet.”
  5. Untick the option that allows smart features to use your email content.
  6. Scroll to the bottom and click “Save changes.”

What this does:

  • Disables things like Smart Compose, Smart Reply, automatic categorisation, and some card‑style extras (trip summaries, order tracking, etc.).
  • Reduces the amount of ongoing automated analysis Google performs on your messages for convenience features.

7.2 Disable Workspace smart features (for business / Workspace users)

If you’re on Google Workspace (for work or school), there is a second layer: [26]

  1. After turning off Smart Features as above, look for a link or button like “Manage Google Workspace smart feature settings”.
  2. In the pop‑up, you’ll typically see toggles such as:
    • Smart features in Google Workspace (Gmail, Calendar, Drive, etc.)
    • Smart features in other Google products (e.g., Maps, Google Assistant surfaces tied to Workspace data).
  3. Turn these off if you don’t want your Workspace content used for cross‑product personalisation.
  4. Save your changes.

Your organisation’s admin might restrict some options, so not every toggle will be available to every user.

7.3 Adjust Smart Features on mobile (Android and iOS)

On mobile Gmail, the labels vary slightly, but the principle is the same. ZDNet’s summary and other guides point to this general flow: [27]

  1. Open the Gmail app.
  2. Tap the three‑line menu (☰) and select Settings.
  3. Choose your account if prompted.
  4. Look for a Data privacy or Smart features section.
  5. Turn off Smart Features and, where available, the related Workspace smart features and “smart features in other Google products” toggles.

Again, wording can differ slightly by region and app version, but anything labelled “smart features” or “personalisation based on Gmail, Chat and Meet” is worth reviewing.

8. Locking down Gemini: controlling how AI uses your data

Even if Gmail content isn’t feeding Gemini training by default, Gemini itself has its own data‑collection and training behaviour.

According to the Gemini Apps Privacy Hub: [28]

  • Gemini collects:
    • Your prompts and uploads
    • The content it generates
    • Info from connected apps (like Gmail, Drive, Calendar, Photos, etc.)
  • Google uses this data to:
    • Provide and improve Gemini
    • Develop new services and machine‑learning models
    • Maintain security and reliability

A few concrete steps you can take:

8.1 Review “Gemini Apps Activity”

  1. Visit Gemini Apps Activity in your Google Account (you’ll find a link from the Gemini Privacy Hub). [29]
  2. Here you can:
    • See previous Gemini interactions
    • Delete activity
    • Change the auto‑delete period
    • Decide whether your activity is used to improve Google AI

If you turn off activity saving, Google says future chats are not used to improve its AI models (with some narrow exceptions when you explicitly submit feedback). [30]

8.2 Manage connected apps (Gmail, Drive, Chat, etc.)

Google’s connected‑apps help pages say you can manage what apps Gemini can see, and use this access to: [31]

  • Summarise emails from Gmail
  • Pull content from Drive or Docs
  • Interact with Calendar, Photos, communication apps and more

Inside the Gemini app or web experience, look for “Apps”, “Connected apps” or similar settings and:

  • Check whether Gmail and other Workspace apps are connected.
  • Disconnect anything you don’t want Gemini to read or act on.

This doesn’t change how Gmail itself works, but it does limit how much of your email Gemini can see when you use AI features.

9. Why privacy advocates still aren’t relaxed

Even with Google’s clarification and Malwarebytes’ correction, digital‑rights experts and privacy‑focused communities remain uneasy. Their concerns include:

  • Default‑on scanning: For many users, Smart Features appear enabled by default and can be easy to miss, especially when labels and wording change. [32]
  • Blurry consent: It’s often unclear where “product feature” ends and “AI training” begins, especially as Gemini tools are more tightly integrated into Gmail, Drive and Android Auto. [33]
  • Legal grey areas: The California lawsuit argues that quietly enabling Gemini‑related options in Workspace could violate long‑standing wiretap and privacy laws. [34]
  • Security risk of more data processing: Even if your emails never become part of a global training set, the more systems that read and link them, the more attractive they become as a target – and the more damage a breach or misconfiguration could cause. [35]

Al Jazeera/PolitiFact’s analysis today makes a broader point: when transparency is poor and settings are scattered, confusion and fear‑mongering fill the gap – whether or not the worst social‑media claims are accurate. [36]

10. So… is Gmail “safe” to use now?

There’s no one‑size‑fits‑all answer, but here’s the most balanced way to see it as of November 24, 2025:

  1. The viral claim that Gmail is universally training Gemini on all your emails appears overstated. Public documentation and Google’s statements do not support that specific scenario. [37]
  2. Gmail absolutely does scan your emails for Smart Features and personalisation, and in many regions these are effectively opt‑out rather than opt‑in. [38]
  3. Gemini tools can read your emails and files when you connect them, and that data may be used to improve AI models unless you change Gemini’s activity and personalisation settings. [39]
  4. A lawsuit is actively challenging Google’s approach, and it could set meaningful precedents about how clearly companies must explain AI access to private communications. [40]

If you:

  • Handle highly sensitive information (health, legal, trade secrets), or
  • Simply don’t want any automated system reading more than absolutely necessary,

then the most cautious approach today is to:

  • Turn off Gmail Smart Features (Section 7.1 and 7.2).
  • Limit or disconnect Gemini’s access to Gmail and other apps (Section 8).
  • Disable Gemini Apps Activity for training uses, if you’re comfortable losing some personalisation.

If, on the other hand, you’re comfortable with some scanning in exchange for convenience, you may decide to leave Smart Features and Gemini on – but it’s still worth checking your settings so that decision is informed, not accidental.

Bottom line

  • The story that “Gmail secretly flipped a switch to feed all your emails to Gemini” is more complicated – and less apocalyptic – than the memes suggest.
  • But the underlying trend is real: AI is being woven into your inbox, and your privacy increasingly depends on settings screens that most people never look at.
  • Taking five minutes today to review both Gmail and Gemini privacy controls is the most practical thing you can do in response to this week’s controversy.
How To Use Gemini in Gmail the Right Way (Gemini AI Assistant) Tutorial
Watch this video on YouTube.

References

1. www.malwarebytes.com, 2. www.classaction.org, 3. muckrack.com, 4. www.malwarebytes.com, 5. www.malwarebytes.com, 6. www.theverge.com, 7. www.ndtvprofit.com, 8. www.techradar.com, 9. www.ndtvprofit.com, 10. www.theverge.com, 11. www.classaction.org, 12. www.bloomberg.com, 13. www.aljazeera.com, 14. www.theverge.com, 15. www.aljazeera.com, 16. workspace.google.com, 17. blog.google, 18. support.google.com, 19. www.ndtvprofit.com, 20. support.google.com, 21. www.techradar.com, 22. www.business-standard.com, 23. tech.news.am, 24. www.bgr.com, 25. www.malwarebytes.com, 26. www.malwarebytes.com, 27. www.reddit.com, 28. support.google.com, 29. support.google.com, 30. support.google.com, 31. support.google.com, 32. tech.news.am, 33. www.tomsguide.com, 34. www.classaction.org, 35. tech.news.am, 36. www.aljazeera.com, 37. www.theverge.com, 38. www.malwarebytes.com, 39. support.google.com, 40. www.classaction.org

Mateusz Brzeziński

Technology News

  • Tesla Cybertruck Program Head Quits Amid Underperformance
    November 24, 2025, 10:50 AM EST. Tesla's eight-year veteran Siddhant Awasthi, who led the Cybertruck program and previously headed the Model 3, is leaving the company, per CNBC. The departure comes as the Cybertruck has struggled, with about 16,000 units sold through the first three quarters of 2025. Despite a tough year for Tesla on sales, the company posted a 12% revenue rise in Q3, and the Model Y set a quarterly record of 114,897. Elon Musk has signaled a broader shift toward a technology focus, with observers calling Tesla increasingly a tech company. The piece notes investor chatter and contrasts with traditional automakers like Volkswagen, Hyundai, GM, and Kia ramping EV models.
  • China launches commercial trial of satellite Internet of Things services
    November 24, 2025, 10:48 AM EST. China has kicked off a commercial trial of satellite Internet of Things (IoT) services, signaling a push to broaden global IoT coverage beyond terrestrial networks. The pilot aims to connect remote and maritime regions, support logistics and industrial applications, and demonstrate the viability of satellite-enabled connectivity for sensors and devices. If successful, the program could reduce dead zones, improve asset tracking, and accelerate adoption of industrial IoT in sectors like transportation, energy, agriculture, and smart cities. Regulatory progress, satellite constellations, ground stations, and device ecosystems will shape the scale and reliability of the service. The move reflects China's broader strategy to diversify communications infrastructure and strengthen competitiveness in the global tech landscape.
  • Boulder to Begin Citywide Fiber Buildout in 2026: Municipal Broadband Expansion
    November 24, 2025, 10:46 AM EST. Boulder is in the final planning stages of a municipal broadband project, with citywide fiber construction slated to start in 2026 and extend through 2030. The effort follows a 2016 feasibility study and a 55-mile dark fiber backbone completed in 2018 (Phase 1). Construction has begun in Chautauqua and will move to last mile links, targeting roughly 80% completion by 2027 and 97% by 2029. The project covers about 300 miles of fiber serving residents, mobile home communities, and foothills properties. Boulder has signed a 20-year lease with ALLO to deliver service, with discounts for income-qualified residents, free upgrades for college students, and free service for 25 nonprofits. The plan aims to boost competition and broaden choices for consumers and businesses.
  • DXYZ: Despite Owning SpaceX and OpenAI, This Fund Is A Sell
    November 24, 2025, 10:44 AM EST. This analysis argues that DXYZ is a sell even though the fund owns SpaceX and OpenAI. The author concedes the marquee stakes look attractive, but the thesis rests on weak upside catalysts and outsized concentration. The piece points to high-fee structure, liquidity concerns, and the risk of a single-asset downturn harming performance. With limited diversification beyond the two tech juggernauts, the fund may underperform broader tech or market indices in a volatile environment. Investors are urged to scrutinize holdings, valuation, and the manager's ability to manage risk, rather than rely on prestige of SpaceX and OpenAI alone.
  • Enshittification: How the internet became what no one asked for
    November 24, 2025, 10:42 AM EST. Tech critic Corey Doctorow coins the term 'enshittification' to describe a three-step process pushing major platforms toward worse user experiences. In an interview with Nosheen Iqbal, he explains how giants from Amazon to Google to Instagram squeeze more value from users and creators, tightening control, nudging toward paid features, and privileging engagement over utility. The piece outlines the three-step theory, shows how these shifts affect everyday browsing, shopping, and social connection, and considers what can be done to stop it. Possible remedies include policy changes, greater transparency, competition, and user-empowered tools. The goal is to spark a debate about preserving a more open, user-centric internet.