How should organisations reduce business email compromise risk when attackers use generative AI?

Organisations should stop relying on message quality as the main trust signal. Use out-of-band verification for payments, credential changes, and vendor updates, then correlate mailbox behaviour with identity and workflow telemetry so suspicious requests can be blocked even when the email looks authentic.

Why This Matters for Security Teams

Generative AI changes business email compromise because attackers no longer need perfect grammar, familiar tone, or even a long-lived foothold to look convincing. A well-written phishing message can now be generated at scale, personalised from public data, and tuned to match the target’s workflow. That makes message quality a weak trust signal. The real control point is whether the request is authorised, expected, and verified through a channel the attacker cannot fully control.

This is why guidance from NIST Cybersecurity Framework 2.0 and current email threat research both point toward identity-aware detection rather than content-only filtering. NHIMG’s analysis of identity compromise shows how quickly attackers act once they can reuse legitimate access, and the same logic applies when they use AI to impersonate a sender before the mailbox or workflow is actually breached. See also The 52 NHI breaches Report for patterns of credential abuse that begin with trust, not malware. In practice, many security teams encounter payment fraud only after a finance approval chain has already been abused, rather than through intentional verification failure.

How It Works in Practice

Reducing BEC risk starts by assuming the email itself may be authentic-looking but still untrusted. Organisations should move high-risk actions such as invoice changes, payroll redirects, bank detail updates, and new vendor onboarding to out-of-band verification. That means a separately known phone number, a signed approval workflow, or a portal-based confirmation that does not rely on replying to the same inbox.

Detection also needs to correlate mailbox behaviour with identity and business process telemetry. If a request arrives from a familiar address but the sending session, device posture, geo location, or OAuth grant pattern is unusual, the case should be treated as suspicious. Guidance from CISA cyber threat advisories and NIST AI 600-1 GenAI Profile supports layered controls that combine technical signals with process verification. For teams mapping real-world risk, Ultimate Guide to NHIs — Key Challenges and Risks is useful because the same credential abuse patterns that affect machine identities also show up in mailbox compromise and workflow hijacking.

• Require step-up verification for payment and account-change requests.
• Use conditional access and mailbox anomaly detection to flag unusual sender context.
• Separate approval authority from the communication channel used to request the action.
• Log and review delegated mailbox access, token grants, and forwarding-rule changes.

These controls tend to break down when finance, procurement, and IT still accept email reply chains as sufficient approval evidence because the attacker can stay inside the same conversation thread.

Common Variations and Edge Cases

Tighter verification often increases friction, so organisations need to balance fraud prevention against operational speed. That tradeoff is real for urgent payments, executive travel changes, and supplier exceptions, where rigid processes can create workarounds. Best practice is evolving, but there is no universal standard for this yet: the key is to tier controls by transaction risk rather than apply the same friction everywhere.

High-risk organisations should treat executive impersonation, vendor onboarding, and payroll diversion as separate BEC scenarios because the control weaknesses differ. Executive fraud often depends on urgency and authority pressure, while vendor fraud often depends on account-data manipulation and weak callback procedures. Current guidance suggests using policy-based approval routing, immutable audit logs, and separation of duties so one compromised inbox cannot complete the full workflow. The OWASP NHI Top 10 also reinforces the need to treat identity and request context as security inputs, not just the message body. Organisations that rely on a single verification method usually discover the gap only after a finance exception, vendor master change, or mailbox takeover has already been exploited.

Related resources from NHI Mgmt Group

• How can organisations reduce the identity impact of email compromise?
• How should teams reduce the risk of exposed AI credentials being abused?
• How do attackers turn a supply-chain incident into wider NHI compromise?
• How do attackers turn stolen npm secrets into broader compromise?