Pre-delivery vs post-delivery email security

Pre-delivery email security screens a message before it reaches an inbox, so a bad message is refused or held unread. Post-delivery security scans mail that has already landed and pulls bad messages back out. Pre-delivery prevents the first read. Post-delivery limits the damage from what got through.

Updated

What is the difference?

The difference between pre-delivery and post-delivery email security is the order of two events: the check and the delivery. Pre-delivery puts the check first. Post-delivery puts the delivery first.

QuestionPre-delivery screeningPost-delivery remediation
When it decidesBefore the message is readableAfter the message is in the mailbox
Where it sitsIn the mail path: the receiving server, or a gateway in front of itInside the mailbox platform, usually through its API
What it does with a bad messageRefuses it, or holds it outside the inboxMoves it to junk or quarantine, or deletes it
Exposure windowNone for what it catchesFrom delivery until the cleanup runs
What it knowsThe message itself, at one momentThe message, plus what was learned later
What it missesThreats that show themselves laterAnything read or acted on before it runs

What is pre-delivery screening?

Pre-delivery screening is checking a message before it reaches an inbox, while the sending server is still connected or before the message is filed. A bad message is refused or held, so nobody reads it.

Gmail and Outlook do this with their spam and malware filters. Secure email gateways do it in front of a mail server. The full mechanics are in pre-delivery email screening.

What is post-delivery remediation?

Post-delivery remediation is finding and removing a harmful message after it has been delivered. A tool connected to the mailbox platform rescans delivered mail as new information arrives, then moves or deletes what it now knows is bad.

Microsoft 365 has this built in. Its zero-hour auto purge retroactively detects and neutralizes phishing, spam and malware messages that were already delivered to cloud mailboxes. API-based email security products work in a similar way across Google Workspace and Microsoft 365.

Post-delivery remediation exists for good reasons. Microsoft's documentation names two: malware that was undetectable during mail flow, and content weaponized after delivery.

What can each approach do, and not do?

Pre-delivery screening

  • It can stop a message before any reader, human or agent, sees it.
  • It can tell the sending server at once that a message was refused.
  • It can keep a doubtful message outside the inbox for a human to review.
  • It cannot change its mind. The verdict is made once, so a link that turns harmful later is missed.
  • It cannot use context that arrives later, such as other recipients reporting the same message.

Post-delivery remediation

  • It can act on new information: a link that changed, or a campaign reported elsewhere.
  • It can remove one message from many mailboxes at once.
  • It can cover internal mail that never crosses a gateway.
  • It cannot undo a read, a click, a reply, a payment or a forwarded file.
  • It cannot take a message back out of an AI agent that already read it.

Neither approach catches everything. Each one covers a gap in the other.

Why does order matter when an agent reads the mail?

Order matters because post-delivery remediation depends on a gap between delivery and reading, and an AI agent leaves no gap. An agent with a webhook or a waiting API call reads a new message within seconds of delivery.

A human gives a cleanup scan a head start. An unread message can be pulled back before anyone opens it. For an agent, the message is in the model's context before a later scan has run.

The same message, in two orders
Delivery first (post-delivery)
  1. Message accepted and delivered to the agent's inbox
  2. Seconds later: the webhook fires and the agent reads it
  3. The agent acts on a hidden instruction
  4. Later: a scan flags the message and removes it

Check first (pre-delivery)
  1. Message arrives, sending server still connected
  2. Screened: held outside the inbox
  3. The agent sees that a message is held, and nothing its sender wrote
  4. Later: a human releases it or discards it

The agent also needs no click. EchoLeak (CVE-2025-32711) showed the pattern: researchers showed that one crafted email could make Microsoft 365 Copilot disclose data with no click from anyone. Microsoft rated it 9.3 (critical) and published its advisory in June 2025, saying the flaw was already mitigated.

For an agent, reading is the moment of attack, so the check has to come before the read. More on this in zero-click AI attacks and indirect prompt injection.

Do you need both?

Most organizations benefit from both, in that order. Pre-delivery screening removes the first read for what it catches. Post-delivery remediation handles what only time reveals.

The balance shifts with the reader. For a human mailbox on Gmail or Outlook, post-delivery cleanup is a useful second layer, because a message can sit unread for a while.

For an agent's inbox, the second layer arrives too late to help much. The first layer has to be strict, and uncertain mail has to wait for a human instead of being delivered on a guess.

How does Trotta Mail order it?

Trotta Mail checks first, for agents and humans. Every message is screened while the sending server is still connected. Attacks, known and new, are refused or held outside the inbox for a human. Everything else is delivered with a verdict.

A held message cannot be read by an agent's API key on any route. The key sees that the message exists and why, and nothing its sender wrote. See held mail.

Trotta Mail does not count on a later sweep. If a delivered message turns out to be bad, a human or an agent can report it as a threat, which moves it to held for a human.

No filter catches everything. A cleverly reworded scam with no technical tell can still be delivered, which is why uncertain mail is held for a human. The details are in concepts.

Questions

What is post-delivery remediation?

Post-delivery remediation is removing a harmful email from a mailbox after it was delivered. A security tool rescans delivered mail when new information arrives and moves or deletes what it now knows is bad.

Is pre-delivery security better than post-delivery?

Neither replaces the other. Pre-delivery prevents the first read, and post-delivery handles threats that appear later. When an AI agent reads the mail, pre-delivery matters more, because the agent reads within seconds.

Do Gmail and Outlook use pre-delivery or post-delivery security?

Both. They filter spam and malware before delivery. After delivery, Microsoft 365 can move or quarantine a message later found to be harmful, and Google Workspace alerts admins on some editions, who can then delete it.

What is the exposure window?

The exposure window is the time between a message being delivered and a later scan removing it. During that time the message can be opened, clicked or read by an AI agent.

Can an email be pulled back after an AI agent has read it?

The message can be removed from the mailbox, but the agent has already taken it as input and may have acted on it. Removal does not undo a reply, a forwarded file or a tool call.

Sources

Give your agent an inbox of its own. Free for two inboxes. No card.