Spell-check for privacy
Blog
Email privacy 10 min read

Gmail at work: you can contract away Google, not the recipient

It is easy to worry about what Google does with your Gmail. But the platform risk is largely a contract problem - and contracts exist. The risk no agreement covers is the one that does the real damage: the data you hand to another person.

Gmail
New message
Todr.smith@clinic.co.uk
SubjectClient file: Client
BeeSensible highlights sensitive details before send.

The example above is interactive. Click a highlighted value to see your action options.

Quick answer

Gmail can be run securely, and most of the worry about Google is contractible: a data processing agreement, EU data residency, no ad-scanning since 2017, and Google's stated position that it does not train Gemini on Gmail content. What no contract changes is what you put in front of a recipient. The most common email breach is not Google reading your mail - it is a person sending the wrong data to the wrong human, and that is where Gmail privacy actually leaks.

01

The Google platform risk is largely contractible: DPA, EU data residency, no ad-scanning, no Gemini training on Gmail

02

What no contract covers is the data you choose to share with a recipient

03

Misdirected emails were the single most common data security incident reported to the ICO in 2024

04

Putting addresses in To or CC instead of BCC is one of the most-reported breaches every year

05

Confidential mode does not reduce what you share - the recipient still sees and can screenshot the content

A charity sends its monthly update to the families it supports. A staff member pastes the list into the To field instead of BCC. Every recipient can now see every other recipient - and because of what this particular charity does, the email address alone reveals something deeply personal about each family on it.

This is not a hypothetical. In 2016 the UK regulator fined an NHS sexual-health clinic £180,000 after a newsletter went to 781 people with their addresses in the open instead of BCC, exposing names tied to HIV status. It was the second time the same organisation had made the same mistake.

No firewall stops this. No encryption setting prevents it. It is not Google's systems failing - it is a person, in a hurry, handing data to other people. And that, far more than anything Google does on its servers, is where Gmail privacy actually goes wrong.

The risk you can put in a contract - and the one you can't

It is natural to start with Google itself. The familiar worry runs: it reads everything, trains its AI on your mail, and monetises your inbox. That concern deserves a straight answer - and the answer is more reassuring than the worry.

Most of it is contractible. On Google Workspace you get a data processing agreement under GDPR. You can restrict data residency to the EU. Google stopped scanning consumer Gmail content for ad personalisation in 2017. And it has stated publicly that it does not use Gmail content to train its Gemini model. You do not have to take any of this on trust alone - much of it is written into terms you can hold Google to.

In other words, the platform risk has an owner, a contract, and a remedy. You can buy it down.

What you cannot buy down is the recipient. No agreement Google signs changes who an employee emails, how much they include, or whether they forward a thread with three weeks of history attached. The moment a person decides what to put in front of another person is the moment no contract reaches - and it is the moment where the overwhelming majority of email breaches begin.

How Gmail handles your data

Consumer Gmail (free)Google Workspace (business)
Account typePersonal @gmail.comManaged work domain
Data processing agreementNoYes
Content scanned for adsNo (since 2017)No
Content used to train GeminiNo (per Google)No (per Google)
Admin controlNoneFull, via Admin console
Changes who you emailNoNo

The table makes the point that matters: moving from consumer Gmail to managed Workspace fixes the contract column. It does not touch the bottom row. A business should be on Workspace - a personal Gmail account processing client data has no processor agreement behind it and is the wrong tool for the job. But even a perfectly configured Workspace, with EU residency and every AI feature reviewed, still sends exactly the email its user tells it to, to exactly the person in the To field.

The biggest privacy risks in Gmail

Every one of these is about data reaching a human - not a server.

1. Autocomplete resolves to the wrong contact

Gmail suggests recipients aggressively from your history: former colleagues, similarly named contacts, an external partner from a project two years ago. The suggestion appears before you finish the name. One tab and the email is addressed to the wrong person, with whatever it contains.

2. Reply All on a thread that grew

A message arrives via a list or a wide CC. You reply to one point and click Reply All. Everyone who was ever on the thread receives your answer - and the older messages beneath it, which may hold context those people were never meant to see.

3. To or CC instead of BCC on a group email

Send to a group in To or CC and every recipient's address is exposed to every other recipient. When those addresses belong to patients, claimants, job applicants, or members of a support group, the addresses themselves are sensitive personal data. This single mistake is among the most-reported breaches to regulators every year.

4. Forwarding a thread with its history

Forwarding carries everything below the latest message. A quick "FYI, see below" can pass salary figures, legal advice, or another client's details to someone who had no reason to receive them. Most people do not scroll down before they forward.

5. Attaching a Drive file is sharing it

In Gmail, attaching a Google Doc or Sheet often means sharing a Drive link, not sending a copy. If the file's permission is "anyone with the link," the recipient - and anyone they forward it to - can open far more than the one document you meant to send, sometimes including its edit history and comments. The convenience hides a sharing decision.

6. Putting more in the body than the recipient needs

The quiet, constant risk. A reply that needed a date of birth to confirm identity does not need the full record pasted beneath it. A status update for a colleague does not need the client's account number. Every extra identifier in the body is data handed over for no reason - and the recipient now holds it.

7. Confidential mode that feels safer than it is

Gmail's Confidential mode sets an expiry and blocks forwarding, copying, and downloading. It looks protective, but it does not reduce what you share: the intended recipient still reads the full content on screen and can screenshot or photograph it, and Google can still process it. It is mild access control, not confidentiality - and it can make people share more freely, believing it is safe.

Each of these is a decision about what one person hands to another. None is fixed by a setting on Google's side.

What is actually at stake: consequences

Under GDPR, an email that reaches the wrong person, or exposes a list of recipients' addresses, can be a personal data breach. If the data carries risk to people's rights and freedoms - health details, financial information, a list that reveals something by its mere existence - the organisation must notify its supervisory authority within 72 hours of becoming aware.

Fines reach €20 million or 4% of global annual turnover, whichever is higher, and the liability sits with the organisation as controller. But the regulatory number is rarely the worst part. A misdirected email to a client, a patient, or a vulnerable person creates harm that cannot be recalled - the recipient has already read it. In healthcare, legal, financial, and support settings, that damage is immediate and personal.

And note where the liability does not move: a flawless contract with Google offers no defence when the breach was an employee placing the wrong data in front of the wrong human.

Verified incidents

2016 - NHS sexual-health clinic, BCC failure

A London clinic sent an HIV-related newsletter to 781 recipients using To instead of BCC, revealing 730 full names alongside addresses tied to a deeply sensitive context. The ICO fined the trust £180,000, noting it had made a near-identical error years earlier. Source: ICO / Digital Health, May 2016.

2024 - Misdirected email, the most common incident

UK ICO data showed misdirected emails were the single most common data security incident reported in 2024. Failure to use BCC correctly remains one of the most-reported non-cyber breaches year after year, with the education and health sectors leading the count. Source: ICO, 2024.

2024 - ICO guidance on bulk email

Responding to the persistence of these breaches, the ICO published guidance urging organisations to switch off email autocomplete, build in send delays so a mistake can be caught, and use proper bulk-email or mail-merge tools rather than BCC for sensitive sends. Source: ICO, 2024.

The pattern across all of them is the same: not an attacker, not a platform flaw, but an ordinary person sending data to other people slightly too fast.

Settings that help

These reduce the odds. None removes the decision.

1. Slow down or disable autocomplete The ICO specifically recommends switching off autocomplete for sensitive contexts, so a half-typed name cannot resolve to the wrong contact before you look.

2. Turn on Undo Send and extend the window In Gmail Settings, set the cancellation period to its maximum. It gives you a few seconds to catch a wrong recipient or a forgotten BCC.

3. Use BCC - or a real bulk tool - for groups For any group message, use BCC, or better, a mail-merge or bulk-email service. For genuinely sensitive lists, BCC alone is not the recommended control.

4. Check Drive link permissions before attaching When attaching a Google file, confirm the sharing scope. Default to "restricted" or specific people, not "anyone with the link."

5. Treat Confidential mode as access control, not secrecy Use it to limit forwarding, but never rely on it to keep content from a recipient who can simply screenshot.

Steps verified in June 2026; Gmail's settings labels shift over time.

What no setting - or contract - solves

A data processing agreement governs Google. Autocomplete settings and send delays reduce slips. But nothing on either list decides who an employee emails, how much they paste, or whether they forward a thread with its history attached.

No contract un-sends a message. No undo window helps once the few seconds pass. No confidential-mode flag stops a recipient from reading what you chose to give them. The most consequential Gmail exposures are not Google processing your data - they are people sharing data with other people, quickly, during ordinary work.

That is the gap the platform cannot close. It sits between the moment an employee has sensitive information and the moment they hand it to someone else.

How BeeSensible helps before you send

BeeSensible checks personal data in text fields - including the Gmail compose window - as you type. Through the desktop app, detection runs entirely on the device and no text leaves the machine. For browser-only use, the extension sends the text to BeeSensible's EU detection service, where analysis runs in working memory and the text is discarded after detection. When sensitive content appears in the body or subject of a draft, BeeSensible marks it inline so the sender can see exactly what they are about to share - and delete it, replace it with a placeholder, or mask it before it reaches anyone.

Gmail
New message
Todr.smith@clinic.co.uk
SubjectClient file: Client
BeeSensible highlights sensitive details before send.
Hover or tap a highlighted value to replace, mask, or delete it - before the draft reaches anyone.

Message content is not stored. The user makes every decision.

For Gmail, this works on the one risk no contract and no setting reaches: the data a person is about to put in front of another person. It does not choose your plan or sign your DPA - those matter, and they are separate. It keeps the sender sharp in the compose window, so a client name beside an account number, or a recipient who never needed that detail, is caught before it becomes someone else's to read.

You can hold Google to a contract. You cannot recall an email. Gmail's real privacy story is human: the platform is the part you can sign away, and the recipient is the part you have to stay awake for.

FAQ

Common questions

Is Gmail safe for business data?

On managed Google Workspace, the platform side is largely covered: there is a data processing agreement, EU data residency options, no ad-scanning, and Google states it does not use Gmail content to train Gemini. What none of that addresses is the most common failure - sending the wrong data to a recipient. Gmail is safe enough as a platform; the risk lives in what people put in front of other people.

Does Google read my Gmail or use it to train its AI?

Google stopped scanning consumer Gmail content for ad personalisation in 2017, and has publicly stated it does not use Gmail content to train its Gemini model. On Workspace, content is processed under business terms, not for advertising. These are real, contractible commitments. The bigger day-to-day risk is not Google's processing - it is the personal data an employee shares with the wrong human.

What is the most common email data breach?

Misdirected emails - sending to the wrong recipient, or putting addresses in To or CC instead of BCC. Misdirected email was the single most common data security incident reported to the UK ICO in 2024, and BCC failures sit in the top breaches reported every year. These are human errors made in seconds, not platform failures.

How do I stop sending a Gmail to the wrong person?

Turn off or slow down autocomplete so a half-typed name does not resolve to the wrong contact, enable Undo Send to give yourself a correction window, and use BCC - or a proper bulk-email or mail-merge tool - whenever you email a group. For sensitive sends, pause and re-read the recipient field and the thread history before pressing send.

Is Gmail Confidential mode secure?

Confidential mode lets you set an expiry and limit forwarding, copying, and downloading, but it is not end-to-end encryption and it does not reduce what you share. The intended recipient still reads the full content on screen and can screenshot or photograph it, and Google can still process it. Treat it as a mild access control, not a confidentiality guarantee.

Is a data processing agreement enough to use Gmail for personal data?

A data processing agreement with Google is necessary for processing personal data on Workspace, but it is not sufficient on its own. It governs how Google handles the data; it does nothing about which recipient an employee sends it to, how much detail they include, or whether a thread is forwarded with history attached. The contract covers the platform, not the human decision.

See how BeeSensible works

Detect sensitive data before it leaves your team, in any app, in real time.