Google Has Gmail AI Scanning Turned On by Default. Here Is How to Turn It Off.

What Google Is Doing

Google's current Gmail interface ships with AI mail scanning and analysis enabled by default. That means, unless you have changed your settings, Google's AI systems are reading and analyzing your emails automatically.

This isn't a secret program. Google makes the settings accessible. But default-on is a meaningful choice: most users never change defaults, and Google knows that.

The disclosure comes via a piece by Thomas Neuburger, published through Naked Capitalism, which aggregated a Twitter thread walking through the opt-out steps. ZeroHedge republished Neuburger's piece. The underlying opt-out instructions are verifiable independently through Google's own settings interface.

The Broader Context

Neuburger's piece frames Gmail's AI features as one piece of a much larger pattern. Doctor's offices are now using AI to record and transcribe patient visits, with retention periods that vary and can be changed. AI systems are screening job applicants and loan applications before a human ever sees a résumé or credit file. The IRS is reportedly using AI to decide audit targets. The Social Security Administration has moved AI into constituent-service workflows.

AI-assisted policing is already operational in multiple U.S. jurisdictions. And military use of AI in targeting decisions, what Neuburger calls "AI battlefield murder," is an active and largely unresolved policy debate.

The common thread: AI is replacing the human in the loop, often before the public has consented to or fully understood that substitution.

Two Legitimate Concerns

Neuburger specifically names two distinct problems. First, government and corporate surveillance expanding through AI infrastructure. Second, AI companies training their models on user-generated content, including private email, without explicit consent or payment to the people who created that content.

Those are separate issues that often get conflated. Surveillance is about who can see your data and under what legal authority. The training-data question is about intellectual property and contract: did you agree, in Google's Terms of Service, to allow your emails to feed a training dataset? Most users couldn't tell you.

The Strongest Counter-Argument

Google's position is coherent: AI-assisted features make Gmail genuinely more useful. Smart Reply, spam filtering, phishing detection, and meeting-scheduling suggestions are all AI-powered, and most users benefit from them daily. Turning off AI features isn't cost-free: you lose the defensive tools alongside the data-collection ones. Google also argues its on-device and cloud processing complies with its published privacy policy, which users agreed to when creating an account.

Blanket anti-AI framing ignores that AI spam filters alone have materially reduced the volume of malicious emails reaching ordinary inboxes. The question isn't whether AI in email has any value. It's whether users should have to opt out of data collection rather than opt in.

How to Reduce AI Access to Your Gmail

If you want to limit how much AI interacts with your Gmail account, the process is available in Google's settings. According to the steps outlined in Neuburger's piece (sourced from the Twitter thread he cites):

• Open Gmail in a browser.
• Go to Settings (the gear icon) and then See all settings.
• Navigate to the General tab and look for AI-related features, including Smart Compose, Smart Reply, and Nudges.
• Toggle each off individually.
• Also check Settings > Data & Privacy in your broader Google Account for AI and activity controls.

Note that disabling these features in Gmail does not necessarily stop Google from processing your email data for other purposes under its Terms of Service. Full data isolation from Google's infrastructure is not achievable while using Gmail.

What Remains Unresolved

The source base here is thin: one article, one political lean, no response from Google on the record. This article cannot independently verify every specific claim Neuburger makes about government AI use, and some of his framing—particularly around Social Security AI replacement—cites secondary reporting that isn't fully sourced in the available material.

What is verifiable: Gmail's AI features are on by default, opt-out settings exist, and Google's Terms of Service are written broadly enough that the scope of data use for AI training is genuinely ambiguous to ordinary users.

The unresolved policy question is whether default-on AI scanning of private communications should require affirmative opt-in under U.S. law. The EU's GDPR imposes stricter consent requirements for exactly this kind of data processing. The U.S. has no equivalent federal standard as of June 18, 2026, and no legislation directly addressing this specific practice is currently enacted at the federal level.