AI-driven email fraud is exposing gaps in legacy SEG controls

At a glance

What this is: This is an analysis of why secure email gateways miss modern payload-less social engineering and what behaviour-based email defence changes.

Why it matters: It matters because identity teams now have to govern user trust, vendor impersonation, and payment-change fraud alongside conventional email filtering and access controls.

By the numbers:

• Replacing third-party SEGs saves organisations an average of 42% in licensing costs and cuts SOC time by 95%.
• 76% of Abnormal's 550-plus enterprise customers now operate without a third-party SEG after migration.

👉 Read Abnormal AI's analysis of why legacy SEGs miss payload-less email attacks

Context

Secure email gateways were built for spam, malicious links, and attachment-based malware. That model breaks down when the attacker sends a convincing, text-only message that contains no obvious payload, no malware, and no known-bad indicators for the filter to catch.

The identity governance problem is that modern email abuse now targets trust relationships, not just inbox content. When a spoofed vendor or coworker can push a payment change or account-takeover prompt through a trusted channel, email security becomes part of NHI, IAM, and fraud control rather than a stand-alone mail hygiene issue.

Key questions

Q: How should security teams stop text-only email fraud when there is no malware to block?

A: Security teams need controls that evaluate sender behaviour, conversation history, and business context, not just message payloads. Static gateways can still filter spam and known malicious artifacts, but they will miss believable social engineering that exploits trust. The practical answer is to add behavioural detection and verification around sensitive workflows such as payments and account changes.

Q: Why do secure email gateways fail against modern phishing and invoice fraud?

A: SEGs fail because they were designed around malicious links, attachments, and known-bad indicators. Modern phishing often arrives as plain text from a trusted-looking sender and uses the business process itself as the attack surface. When there is no payload to sandbox, the gateway has little to stop, so human trust becomes the main control gap.

Q: How do teams know whether a second email security layer is actually adding value?

A: A second layer adds value only if it detects a different class of risk from the native mailbox security stack. If it merely repeats spam and malware filtering, it adds cost, tuning effort, and false positives without improving resilience. Teams should test for incremental coverage of behavioural anomalies, vendor impersonation, and workflow abuse.

Q: Who should verify payment changes when a trusted email request looks legitimate?

A: The request should be verified by a separate control owner or process, not by replying within the same email thread. Finance, procurement, and executive teams should use out-of-band verification for bank-detail changes, urgent transfers, and supplier updates. That breaks the attacker’s ability to convert trust into immediate action.

Technical breakdown

Why secure email gateways miss payload-less social engineering

Secure email gateways inspect messages before they reach the cloud mailbox and score them using rules, reputation, and signatures. That works well when the message contains malware, a bad link, or a known spoofing pattern. It fails when the attack is purely linguistic and behavioural. A text-only email from a compromised or spoofed trusted contact can look normal to a gateway because the message itself carries no malicious artifact. The control is tuned to detect object-level threats, not relationship abuse or business-process manipulation.

Practical implication: treat SEG coverage as necessary but insufficient for invoice fraud, vendor impersonation, and account takeover prompts.

How behavioural AI changes email defence

Behavioural email security flips the detection model from known-bad to known-good. Instead of asking whether a message matches a malicious pattern, it builds a baseline from sender relationships, message tone, timing, device context, and normal payment or collaboration behaviour. That lets the system spot deviations such as an unexpected bank-detail change or a message that is technically plausible but contextually wrong. The key architectural difference is that identity and communication context become detection inputs, so the system can intervene before a user acts on the message.

Practical implication: baseline the communication graph around finance, executive, and vendor workflows so deviations can be isolated before business action occurs.

Why duplicate email controls create cost without coverage

Many enterprises run a SEG alongside native Microsoft 365 or Google Workspace protections, assuming that two layers automatically mean better defence. In practice, both layers often overlap on spam, commodity phishing, and malware triage, so the organisation pays twice for similar coverage. The real gap remains the same: contextual detection of socially engineered abuse. If the second control does not add behavioural visibility or workflow context, it mainly increases policy maintenance, false positives, and analyst effort.

Practical implication: measure whether each email control adds unique detection value before keeping both in the stack.

Threat narrative

Attacker objective: The attacker wants to exploit trusted communication channels to trigger fraudulent payment or account actions before the deception is recognised.

1. Entry occurs through a text-only email that appears to come from a trusted contact or vendor and contains no malicious payload for the gateway to block.
2. Escalation happens when the recipient trusts the message enough to change payment details, disclose information, or approve a transfer outside normal verification flow.
3. Impact follows as the organisation suffers direct financial loss, account takeover exposure, or business email compromise with limited opportunity for post-delivery filtering to help.

Breaches seen in the wild

• Shai Hulud npm malware campaign — Shai Hulud campaign: npm malware exposed secrets on GitHub.
• DeepSeek breach — DeepSeek breach exposed 1M+ log lines and sensitive secret keys.

Read our 52 NHI Breaches Analysis report for a comprehensive view of breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Payload-less email fraud is a trust-control failure, not a spam problem: The attack works because the control stack still assumes malicious email must contain detectable payloads or obvious indicators. That assumption fails when the actor uses a believable conversation thread and the business process itself becomes the target. Practitioners should treat this as a shift from mail hygiene to identity-aware trust governance.

Behavioural email security creates a distinct identity signal that legacy gateways cannot see: Static filters inspect message content, while behavioural systems inspect whether the message fits the normal relationship between people, vendors, and payment workflows. That is a different control class, not a nicer version of the same gate. The implication is that email security now belongs in the same conversation as IAM, fraud prevention, and vendor trust.

Dual-layer email stacks often create coverage duplication instead of resilience: When the cloud email platform already blocks commodity spam and malware, a second SEG frequently repeats the same controls without adding context. That leaves the organisation with higher administrative load, more false positives, and no better view of social engineering. Practitioners should judge stack value by unique detection scope, not by the number of tools in the path.

Identity-blind email filtering will underperform in finance-heavy workflows: Payment instructions, supplier changes, and executive requests depend on recognising who normally communicates with whom and when. If the control cannot model that baseline, it cannot reliably separate ordinary business from manipulated business. The practical conclusion is that finance, procurement, and executive mail flows need identity-linked monitoring, not just perimeter filtering.

From our research:

• Systems with least-privileged AI access had a 17% incident rate vs 76% for over-privileged systems, according to The 2026 Infrastructure Identity Survey.
• Only 44% of organisations have implemented any policies to manage their AI agents, despite 92% agreeing that governing AI agents is critical to enterprise security.
• For adjacent guidance, review The 52 NHI breaches Report for patterns of identity abuse that turn trusted access into business loss.

What this signals

Behavioural email defence is becoming part of the wider identity boundary, because trust abuse now sits between IAM, fraud, and mailbox security. As organisations move more business processes into cloud collaboration, the mail channel becomes a control point for payment validation, supplier changes, and account recovery rather than just message delivery.

Identity-weighted email risk: the next maturity step is to treat sender relationships, vendor history, and workflow context as security signals, not just user metadata. That shift matters because static filters cannot distinguish a routine request from a manipulated one once the attacker uses a legitimate-looking thread.

The operational signal for practitioners is straightforward: if a control stack cannot detect a believable request before a human responds, the programme still relies on manual skepticism as a security layer. That is not a stable control model in environments where finance and collaboration happen at machine speed.

For practitioners

• Map high-trust mail flows first Identify the conversations that can trigger money movement, credential reset, or privileged approval, then assign tighter verification rules to those paths. Start with finance, procurement, executive assistants, and supplier onboarding channels.
• Test controls against payload-less attacks Run monitor-only simulations using text-only vendor impersonation, invoice-change fraud, and reply-thread abuse so you can see what passes static filters. Use the results to compare behavioural detection against SEG-only coverage.
• Measure unique detection value per email layer Inventory which threats each layer actually stops, then remove or reconfigure controls that only duplicate spam and malware handling. Retain only the layers that add context, identity correlation, or workflow awareness.
• Tie payment verification to out-of-band identity checks Require a separate verification path for bank-detail changes, urgent transfer requests, and vendor account updates, even when the email appears to come from a trusted sender. Build the check around the transaction, not the inbox.

Key takeaways

• Legacy email gateways are not designed to stop payload-less social engineering, which leaves trusted communication as the primary attack surface.
• A real-world text-only fraud case produced a $753,000 loss, showing that email abuse now creates direct financial impact without malware.
• Practitioners should measure whether each email layer adds unique behavioural visibility, then anchor payment and vendor-change verification outside the inbox.

Key terms

• Secure Email Gateway: A secure email gateway is a filtering layer placed before a mailbox service to block spam, phishing, and malware. It relies heavily on signatures, reputation, and policy rules, which makes it effective against commodity attacks but weaker against text-only social engineering and business-process fraud.
• Behavioural Email Security: Behavioural email security uses communication patterns, identity context, and historical relationships to judge whether a message fits normal activity. Instead of asking whether an email is obviously malicious, it asks whether the request is consistent with how people, vendors, and workflows normally behave.
• Payload-less Attack: A payload-less attack is a malicious message or interaction that contains no attachment, link, or malware for traditional tools to detect. The attacker relies on trust, urgency, and business context to get a human to act, which shifts the control problem from content inspection to behavioural verification.

Deepen your knowledge

NHI governance, agentic AI identity, and machine identity security are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are responsible for identity security strategy or programme maturity, it is worth exploring.

This post draws on content published by Abnormal AI: Key Insights on why SEGs miss payload-less, socially engineered attacks. Read the original.