Business email compromise is outpacing legacy email defenses

At a glance

What this is: NHI Mgmt Group's analysis, drawing on Abnormal AI's blog post: Business email compromise is becoming harder to stop because attackers use plain text, legitimate domains, and AI-generated messaging that bypasses legacy email filters.

Why it matters: It matters because identity and access teams need controls that evaluate context, relationships, and behaviour, not just message content, when trust itself is the attack surface.

By the numbers:

• Business email compromise caused $2.8 billion in losses in 2024 alone, according to the FBI IC3 Internet Crime Report cited by Abnormal AI.
• Business email compromise has generated $17.1 billion in reported losses since 2015, according to the FBI IC3 Internet Crime Report cited by Abnormal AI.
• Abnormal AI says business email compromise attack volume rose 54% between 2023 and 2024.

Context

Business email compromise is a trust attack, not a malware delivery problem. It uses convincing human language, spoofed identities, and urgent requests to get employees to move money or change account details.

That is why legacy email security controls struggle here. Gateways are tuned for malicious links, attachments, and known-bad infrastructure, while BEC often arrives as plain text from legitimate-looking domains and pushes the decision burden to the recipient.

For IAM, PAM, and fraud teams, the real issue is not email content alone. It is whether an organisation can validate the relationship, authority, and business context behind a request before money or access changes hands.

Key questions

Q: What breaks when email security only scans attachments?

A: Attachment-only DLP misses the most common modern leak paths: data pasted into message bodies, screenshots, and text copied into AI tools. That leaves organisations blind to context, intent, and downstream reuse. A mailbox can appear compliant while the sensitive content is already being repackaged or exfiltrated elsewhere.

Q: Why does generative AI make business email compromise harder to detect?

A: Generative AI removes the language mistakes and awkward phrasing that once helped people spot fraudulent messages. That raises the quality of impersonation and reduces the value of instinct-based review. Defenders now need context and behaviour signals because the text itself no longer separates real from fake reliably.

Q: How should organisations verify high-risk requests that arrive by email?

A: Use an independent confirmation step outside the email thread for payments, vendor changes, payroll updates, and other high-impact actions. Verification should check the requester through a separate trusted channel and confirm the business event before any action is taken.

Q: When should organisations prioritise automated BEC detection over employee awareness training?

A: Once messages are AI-polished and embedded in normal business workflows, automated detection should take priority because training cannot reliably distinguish legitimate-looking fraud at scale. Awareness still matters, but it should support machine-enforced controls rather than carry the primary burden alone.

Technical breakdown

Why plain-text BEC bypasses legacy email filtering

Business email compromise often avoids the indicators that secure email gateways were built to detect. There may be no malicious attachment, no payload to detonate, and no obvious phishing link. Instead, the message is socially engineered to look routine, using authentic tone, domain similarity, and business context that make static content inspection weak. This is why plain-text fraud can survive perimeter filtering and land in the inbox as if it were normal correspondence. The control failure is architectural: the detection model is focused on message artifacts rather than intent, sender behaviour, and relationship context.

Practical implication: email security teams need detection that evaluates identity and communication patterns, not only content signatures.

How generative AI changes BEC tradecraft

Generative AI lowers the effort required to produce polished, personalised, and context-aware fraudulent messages. That matters because older BEC campaigns often betrayed themselves through awkward language, formatting errors, or generic wording. AI reduces those tells and makes messages harder for employees to judge on instinct alone. The result is not just more volume, but better calibration to the victim’s role, process, and expected communication style. In identity terms, the attacker is borrowing credibility from legitimate organisational relationships and using language quality to simulate authority.

Practical implication: awareness training alone cannot keep pace with AI-assisted impersonation without automated behavioural detection.

Why relationship mapping matters more than inbox inspection

Modern BEC defence increasingly depends on modelling who normally talks to whom, when, and about what. Relationship mapping creates a baseline of organisational communication patterns so anomalous requests can be tested against expected behaviour. That includes signals such as risky logins, mailbox rule changes, sudden sender drift, and unusual authority claims inside a thread. This is a different control philosophy from classic email hygiene. It treats the message as one signal inside a broader identity and context graph, which is where BEC often becomes visible.

Practical implication: integrate mailbox, identity, and behavioural signals into a single detection path for high-risk requests.

Threat narrative

Attacker objective: The attacker wants employees to authorise payments or account changes that move money to criminal-controlled destinations.

1. Entry occurs through a convincing email that uses a trusted-looking identity and plain text instead of malware or links.
2. Credential harvesting is often unnecessary because the attacker relies on authority impersonation and process manipulation to trigger action.
3. Escalation happens when the victim follows the request and routes payment or sensitive change approval outside normal verification.
4. Impact is financial loss, operational disruption, and reputational damage once the fraudulent transaction is discovered.

Breaches seen in the wild

• TruffleNet stolen AWS keys campaign 2025: TruffleNet used 800+ attacker hosts to test stolen AWS keys and check SES capacity; one compromised account sent a fake $50,000 invoice.

Read and download The State of NHI & AI Agent Breach Report 2026, covering 150+ breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Business email compromise is now a governance problem, not just a mail-security problem. The article shows that BEC succeeds because organisations still treat trust as an informal human judgement rather than a verifiable control point. When a request can be legitimate-looking, plain text, and domain-correct, the control surface shifts from the message to the identity and relationship behind it. Practitioners should treat high-risk email requests as identity governance events, not just inbox events.

Legacy email defences are optimised for the wrong failure mode. Secure email gateways were built for malicious content, but BEC exploits authorised-looking communication paths and process familiarity. That means the more mature the email hygiene programme becomes, the less relevant it is to this specific class of fraud unless it is paired with context-aware detection. The practical conclusion is simple: content filtering is necessary, but it is no longer the deciding layer for BEC.

Identity context is the new inspection layer for business communications. The article’s strongest signal is that legitimacy can no longer be inferred from sender appearance alone. Relationship history, mailbox behaviour, and request timing are now more decisive than URL reputation in many BEC cases. That pushes identity teams toward controls that validate business context before high-risk actions are completed.

Human trust remains the attack surface, but the control point has to move upstream. Training still has value, but the article is right that training cannot be the sole defence once AI makes fraudulent messages indistinguishable from routine communication. The real question is whether the organisation can introduce machine-enforced friction before money, payroll, or vendor details are changed. That is where BEC becomes a lifecycle and workflow governance issue, not just an awareness issue. The implication is tighter verification of high-impact requests before execution.

What this signals

Identity-aware email defence is becoming the decisive control layer for BEC. Organisations that still rely on spam filtering and user judgement are defending the wrong boundary. The better model is to evaluate request legitimacy against identity history, communication norms, and workflow context before a sensitive action is allowed to complete.

Business email compromise is now a fraud and identity governance problem. The organisation that can validate who is asking, what they are asking for, and whether the request fits the known relationship has a materially better chance of stopping payment diversion and executive impersonation. That is where email defence, IAM, and finance controls converge.

Request verification should be built into the workflow, not appended to it. If approval, payment, or payroll changes can be triggered directly from an inbox, the organisation has already accepted too much trust at the point of execution. The control needs to sit in the business process itself, before the action becomes irreversible.

For practitioners

• Harden approval workflows for payment and vendor changes Require out-of-band confirmation for wire transfers, payroll edits, bank detail changes, and other high-impact financial requests before they are executed.
• Add behavioural signals to email defence Correlate mailbox rule changes, risky sign-ins, unusual thread patterns, and sender-history anomalies to flag requests that look legitimate but do not fit normal communication behaviour.
• Segment privileged communication paths Separate executive, finance, and HR request handling from ordinary inbox workflows so a single convincing email cannot trigger an irreversible action.
• Treat employee training as a control layer, not the control Keep awareness training in place, but back it with policy, verification steps, and automated monitoring because human review alone will miss AI-polished fraud.

Key takeaways

• Business email compromise succeeds because it exploits trust and business process legitimacy, not malware delivery.
• Generative AI raises the quality of impersonation and makes human-only review less reliable.
• The effective defence is context-aware detection plus workflow verification before money or account details change.

Key terms

• Business email compromise: A form of social engineering where an attacker impersonates a trusted person or domain to manipulate payment, change banking details, or extract sensitive information. It often succeeds without malware because the attacker targets process trust and human judgement instead of technical controls.
• Relationship Mapping: Relationship mapping is the process of identifying and linking connected individuals or entities around a primary subject. In compliance workflows, it helps reveal indirect risk through family ties, close associates, or ownership structures that a single-record match would miss.
• Context-Aware Detection: Context-aware detection is a method that evaluates surrounding meaning before deciding whether data is sensitive. Instead of relying only on patterns, it uses signals from nearby text, file content, or communication context to reduce misclassification. This approach is especially useful in DLP, where isolated terms can look risky but are actually benign.
• Authority Impersonation: Authority impersonation is the use of a trusted role, such as an executive, finance lead, or vendor contact, to pressure a recipient into action. It is effective because the attacker borrows organisational legitimacy instead of trying to break technical controls directly.

Deepen your knowledge

NHI governance, agentic AI identity, and machine identity lifecycle are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are building or maturing an IAM programme, it is worth exploring.