AegisAI Perspectives · 141
Gmail's built-in settings judge mail as it arrives, and an API layer adds checks and removal after delivery, with access your Workspace admin can revoke.
See it on your own Google Workspace or Microsoft 365 tenant. No MX change.
Or get the Q3 2026 State of the AI Attack report (free, instant download)
Not ready for a demo? Check your domain.

Gmail's built-in protection judges mail as it arrives: by default it displays warnings and moves untrustworthy email to the spam folder1. AegisAI investigates every email for what it is trying to do, and removes the ones that fail from every mailbox they reached17. AegisAI reports on its inbound email security page that 3 in 10 confirmed phishing emails cleared native Gmail and Microsoft filters in Q3 202623.
If you run IT for a company of 20 to 500 people on Google Workspace, with a small IT or security team and no SOC, this page shows what Gmail covers and what the API layer adds after delivery. It also lists the access controls that stay in your hands. Whether the built-in layer is enough on its own is answered in is Google Workspace's built-in email security enough. The setup steps are in how to add a third-party email security layer to Google Workspace.
Key takeaways
- Gmail's default is to display warnings and move untrustworthy email to the spam folder.
- Unopened phishing found after delivery is moved out of inboxes automatically; opened messages stay until someone removes them7.
- The added layer is out of band: "It reads mail through the Microsoft 365 or Google Workspace API after delivery. No MX change, no gateway."18
- "Runs in monitoring mode first. Removal needs write access your admin grants and can revoke."
- Google classifies both gmail.readonly and gmail.modify as restricted scopes11.
- The inbound email security page states: "The median verdict takes 1.9 seconds, and a convicted message is removed from every mailbox it reached within seconds."
| Question | Answer | Source |
|---|---|---|
| Is Google Workspace's built-in email security enough, or do I need a third-party tool? | Built in, the Safety settings act on attachments and links, plus spoofing2. Opened phishing found after delivery stays until someone removes it. An API layer adds checks after delivery and removes the emails that fail from every mailbox they reached. | Google Workspace Admin Help; product page |
| What does Gmail do by default? | It displays warnings and moves untrustworthy email to spam. Security Sandbox is off by default5. | Google Workspace Admin Help |
| How does the added layer connect? | Out of band, through the Google Workspace API after delivery. | Inbound email security page |
| What does it read? | "Specialist agents investigate different evidence: sender reputation, threat indicators, the link chain, the business context of the message."24 | Product blog |
| What does the admin control? | Write access for removal, which the admin grants and can revoke. In the Admin console an app is Trusted, Limited or Blocked13. | Inbound email security page; Google Workspace Admin Help |
Short on time? Watch the explainer, or book a demo.
Native Gmail security covers attachments and links before delivery, plus spoofing and authentication. Each spoofing and authentication option defaults to keeping the email in the inbox with a warning. Changing these settings requires the Gmail Settings administrator privilege3.
Two options add checks before delivery. Enhanced pre-delivery message scanning slightly delays a suspicious message so Gmail can run extra security checks4. Security Sandbox is off by default, and only some editions support it6.
It falls short after delivery. Opened phishing that Gmail finds later stays in the inbox until someone removes it. Keep every Safety setting on, and write down the action each one uses today.
It reads the evidence around each message after delivery, not only the attachment or link. Per can I trust AI agents inside my inbox: "Specialist agents investigate different evidence: sender reputation, threat indicators, the link chain, the business context of the message."
Business context matters for BEC. The FBI's 2025 IC3 report records 24,768 business email compromise complaints and $3,046,598,558 in reported losses16.
Gmail's own post-delivery tools depend on privilege and edition. The alert center lists "Phishing message detected post-delivery" as an alert type, and needs the Alert center administrator privilege8. The investigation tool can delete an email from users' inboxes, and needs the Security center administrator privilege10. Gmail log events come only with Frontline Plus, Enterprise Plus or Education Plus9.
The added layer works in that window. The inbound email security page states: "The median verdict takes 1.9 seconds, and a convicted message is removed from every mailbox it reached within seconds." Check which of the Google tools your edition and admin roles give you.
No MX change and no inline gateway. "It reads mail through the Microsoft 365 or Google Workspace API after delivery. No MX change, no gateway."
Setup is an access grant, not a routing change. "Runs in monitoring mode first. Removal needs write access your admin grants and can revoke." The connection steps are on the third-party layer setup page. Grant read access first and hold write access until the trial ends.
Benign mail is not kept for training. "Emails classified as benign stay in the inference pipeline and never enter training. They are processed in memory and discarded."25
Each verdict comes with its reason in plain language21 The access itself sits on Google's restricted scopes. Take these three facts to your own compliance review; this page makes no claim about any law or framework.
Check access in the Admin console, not in a vendor's slides. The full review of vendor API risk is in is it safe to give an email security vendor API access to our mailboxes. These steps cover the controls Google gives you.
Open Security > Access and data control > API controls. Find the app and note whether it is Trusted, Limited or Blocked.
Sign in as a super administrator and open Manage Domain Wide Delegation. List each client and the scopes it holds15.
Read each Gmail scope against Google's list. gmail.readonly means view your email messages and settings.
Treat gmail.modify as write access. It lets an app read, compose and send email, but it cannot permanently delete a message at once, bypassing the trash.
Start the layer without write access. "Runs in monitoring mode first."
To cut access, delete the client in domain-wide delegation. Dependent applications stop immediately. Or set the app to Blocked, and it can't access any Google service.
Allow 48 hours for the accessed apps list to show a grant or a revoke14.
Run steps 1 to 4 again after any change of scope.
Each control sits with a named admin role or a setting you can open.
| Control | Layer | Where it is set | Default or rule | Who holds it |
|---|---|---|---|---|
| Safety settings: attachments; links; spoofing | Gmail, before delivery | Gmail Safety settings | Warnings shown; untrustworthy mail to spam | Gmail Settings administrator |
| Pre-delivery message scanning | Gmail, before delivery | Gmail settings | Suspicious mail slightly delayed for extra checks | Gmail Settings administrator |
| Security Sandbox | Gmail, before delivery | Gmail settings | Off; Frontline Plus; Business Standard; Business Plus; Enterprise Standard; Enterprise Plus | Gmail Settings administrator |
| Post-delivery phishing removal | Gmail, after delivery | Automatic | Unopened moved out; opened stays | Automatic |
| Post-delivery alerts | Gmail, after delivery | Alert center | Alert type listed | Alert center administrator |
| Delete from inboxes | Gmail, after delivery | Investigation tool | Log events on three editions | Security center administrator |
| App access level | Admin console | API controls | Trusted; Limited; Blocked | Admin |
| Domain-wide delegation | Admin console | Manage Domain Wide Delegation | Deleting a client stops dependent apps | Super administrator |
| Mail scopes | Google OAuth | The app's grant | gmail.readonly and gmail.modify restricted | Admin who approves |
| Added layer, connection | API, after delivery | Google Workspace API | No MX change, no gateway | Admin who connects it |
| Added layer, removal | API, after delivery | Write access grant | Monitoring mode first; revocable | Admin |
Print this table and fill in who holds each role in your tenant.
It reads mail through the Microsoft 365 or Google Workspace API after delivery. Every verdict comes with its reason in plain language, so your team can check it and share it. Write access, verdict time and data handling are covered in the sections above.
Track three numbers, each from a place you can open.
To get the third number on your own tenant, ask for a monitoring-mode test of your Google Workspace mail. Put the three figures side by side on day 30 before you grant write access.
Built-in settings judge attachments and links, plus spoofing, and opened phishing found after delivery stays in the inbox. If that gap matters to you, add a layer after delivery. AegisAI investigates every email for what it is trying to do, and removes the ones that fail from every mailbox they reached.
No. The layer connects through the API after delivery, so mail routing stays as it is.
Look for gmail.readonly, which means view your email messages and settings. Then look for gmail.modify, which lets an app read, compose or send email. Google classifies both as restricted.
Delete its client under Manage Domain Wide Delegation, and dependent applications stop immediately. Or set the app to Blocked under API controls. The accessed apps list updates 48 hours later.
No. "Emails classified as benign stay in the inference pipeline and never enter training. They are processed in memory and discarded."