AI-powered business email compromise is outpacing legacy email controls

At a glance

What this is: This webinar argues that AI-powered business email compromise is defeating legacy email controls because the attacks now exploit identity, context, and conversation patterns rather than malware.

Why it matters: IAM, email security, and identity governance teams need to treat BEC as an identity problem across human and non-human trust boundaries, not just a mail-filtering problem.

By the numbers:

• In 2025 alone, organizations reported over $3 billion in losses due to business email compromise (BEC) to the FBI.
• Business email compromise caused over $3 billion in reported losses in 2025, according to the FBI and Abnormal AI.

👉 Watch Abnormal AI's webinar on why legacy email controls miss AI-powered BEC

Context

Business email compromise is a form of identity abuse in email flows, where attackers impersonate trusted people or organisations to trigger payment, credential, or data-handling actions. The primary weakness is not message volume alone, but the absence of reliable identity and behaviour signals inside many mail security stacks.

That matters to IAM and security teams because email remains a high-trust business control plane. When detection tools only inspect domains, links, and attachments, they miss the social and behavioural cues that separate a routine exchange from a convincing impersonation attempt.

Key questions

Q: How should organisations defend against business email compromise when attackers use real conversations?

A: Organisations should combine technical detection with process controls that break the attacker’s ability to turn trust into action. That means relationship-aware monitoring, stronger verification for payment and banking changes, and clear escalation paths for suspicious requests. A familiar thread should never be treated as proof of legitimacy.

Q: Why do secure email gateways miss many BEC attacks?

A: Secure email gateways are strongest against known bad content, not socially engineered requests that use clean infrastructure and trusted language. BEC often succeeds because the email looks legitimate while the intent is fraudulent, so domain checks and attachment scanning do not provide enough context to stop it.

Q: What signals help detect email impersonation before money moves?

A: The most useful signals are deviations in sender-recipient relationships, request timing, tone, and thread history. A request that fits the mailbox but not the relationship is a stronger warning than a suspicious link. Teams should escalate any payment or vendor-change request that breaks the normal communication pattern.

Q: Who should own business email compromise defence in the enterprise?

A: BEC defence should be shared across email security, IAM, fraud, and finance operations. Email controls catch part of the problem, but identity and approval governance determine whether a fraudulent request becomes a real transaction. High-risk workflows need coordinated ownership, not isolated tooling.

Background and context

Why secure email gateways miss identity-driven impersonation

Secure email gateways were designed to spot malware, malicious links, and suspicious file attachments. Business email compromise bypasses that model by using legitimate-looking infrastructure, real sender relationships, and context-aware language to induce action. The detection problem shifts from content inspection to trust inference, because the message itself may be clean while the intent is fraudulent. In practice, that means domain reputation alone is too narrow to catch executive spoofing, vendor impersonation, and thread hijacking.

Practical implication: teams need detection logic that evaluates identity and conversation context, not just message payloads.

How behavioral baselines change BEC detection

Behavioral AI in this context means building a normal pattern for who communicates with whom, when, from where, and in what tone or sequence. Once a baseline exists, the system can flag deviations such as an unusual request path, a copied but out-of-pattern vendor exchange, or a sudden shift in payment language. This is not about replacing mail security with generic AI; it is about adding identity and relationship intelligence that legacy filters never had.

Practical implication: establish baselines for high-trust accounts and route anomalies into manual verification before action is taken.

Conversation hijacking as an identity continuity problem

Conversation hijacking works because an attacker inserts themselves into an existing trusted thread, then reuses the established relationship to push a fraudulent request. The security failure is continuity loss: systems often authenticate the message transport but not the continuity of the relationship behind it. That makes human trust decisions fragile, especially where finance, procurement, or executive assistants rely on email history as proof of legitimacy.

Practical implication: protect high-value email threads with stronger verification for payment, banking, and vendor-change requests.

NHI Mgmt Group analysis

Business email compromise is now an identity abuse problem, not an email hygiene problem. The article is correct that legacy secure email gateways were built for payload-based threats, while modern BEC succeeds by exploiting trust relationships and communication context. That shift matters because the control failure is not just detection coverage, but the assumption that legitimacy can be inferred from message characteristics alone. Practitioners should treat trusted email exchange as an identity surface, not a transport channel.

Conversation hijacking exposes the gap between authentication and behavioural assurance. A mailbox can be authenticated while the conversation behind it is no longer trustworthy. That is the governance problem: controls that verify access do not necessarily verify intent, especially in high-trust finance and executive workflows. The implication is that email security and IAM teams need shared ownership of high-risk communication paths.

Behavioral baselines create a useful signal only when they are tied to business relationships. Generic anomaly detection is not enough if it cannot distinguish a genuine vendor thread from a fraudulent one. The strongest use case is relationship-aware monitoring for payment changes, banking details, and urgent exception handling, where normal patterns are predictable and deviations are costly. Security teams should focus on the business moments where trust is most actionable.

High-trust inboxes need stronger governance than the rest of the mail estate. Executive assistants, finance approvers, procurement contacts, and account managers sit at the intersection of identity and money movement. Those roles deserve stronger verification thresholds because BEC does not need broad compromise, only one credible decision point. Practitioners should classify these workflows as high-impact trust paths and govern them accordingly.

Identity-context detection is the right named concept for this problem. It describes the need to evaluate who is speaking, how they normally behave, and whether the request fits the relationship, not just whether the email is technically clean. That concept captures the failure mode of legacy email security in a way practitioners can operationalise across human identity programmes. Teams should adopt identity-context detection as a governance lens for BEC.

From our research:

• Organisations maintain an average of 6 distinct secrets manager instances, creating fragmentation that undermines centralised control, according to The State of Secrets in AppSec.
• 43% of security professionals are concerned about AI systems learning and reproducing sensitive information patterns from codebases.
• For a wider control perspective, review Ultimate Guide to NHIs , Key Challenges and Risks to see how identity sprawl and trust gaps compound across programmes.

What this signals

Identity-context detection: BEC resilience will increasingly depend on whether teams can fuse behavioural signals, relationship history, and approval logic into one decision path. With organisations maintaining an average of 6 distinct secrets manager instances, fragmentation is already a governance problem in adjacent identity domains, and the same pattern shows up in email trust controls.

Security leaders should expect finance-facing and executive-facing inboxes to become higher-value targets because they compress trust, urgency, and transaction authority into a single channel. That means anomaly review needs to move closer to the business process, not stay trapped inside the mail gateway.

When business users can approve payments or vendor changes from the same channel that carries routine correspondence, the security model has to assume conversation reuse as an attack path. Programmes that separate identity verification from request execution will absorb less fraud than those that rely on email history as evidence.

For practitioners

• Map high-trust communication paths Identify the inboxes and workflows where a single convincing email can trigger payment, credential reset, or vendor-bank-change action. Prioritise executive, finance, procurement, and supplier-facing accounts, then define the verification step required before any sensitive request is honoured.
• Add relationship-aware detection to mail controls Use behavioural signals such as thread history, sender-recipient patterns, request timing, and language shifts to supplement secure email gateway checks. Route anomalies into a secondary review path instead of relying on message reputation alone.
• Require out-of-band verification for payment changes Mandate a separate approval channel for banking detail changes, urgent transfers, and new beneficiary requests. Make the verification step independent of the email thread so a hijacked conversation cannot complete the fraud chain.
• Train approvers on conversation hijack cues Teach business users to look for subtle continuity breaks such as changed reply patterns, unusual urgency, or requests that do not match prior relationship behaviour. Reinforce that a familiar thread is not proof of legitimacy.

Key takeaways

• Business email compromise now exploits identity and trust, which makes content-only email filtering insufficient.
• The reported $3 billion loss figure shows that BEC is a material enterprise risk, not a niche phishing variant.
• Practitioners need relationship-aware detection, out-of-band verification, and tighter governance for high-trust approval paths.

Key terms

• Business Email Compromise: Business email compromise is fraud that uses trusted email relationships to trick people into sending money, changing payment details, or sharing sensitive information. The attack succeeds by abusing identity trust rather than exploiting software flaws, which makes it especially hard for content-only filters to catch.
• Conversation Hijacking: Conversation hijacking is the insertion of a fraudulent actor into an existing email thread so the request appears to continue a legitimate discussion. The attacker relies on prior message history, familiar tone, and trusted recipients to bypass human suspicion and complete the fraud path.
• Behavioral Baseline: A behavioral baseline is a record of normal communication patterns for people, teams, or workflows. In email security, it includes sender relationships, timing, tone, and request sequences, allowing teams to spot requests that fit the mailbox but not the established business relationship.
• Identity-Context Detection: Identity-context detection is the practice of judging an email or request using who is speaking, how they normally behave, and whether the action fits the relationship. It goes beyond simple content scanning and helps separate routine business communication from impersonation and abuse.

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 responsible for identity security strategy or NHI governance in your organisation, it is worth exploring.

This post draws on content published by Abnormal AI: why legacy secure email gateways fail against AI-powered business email compromise. Read the original.