They should require independent verification for payments, account changes, and sensitive approvals, and they should document who can approve what under what conditions. The goal is to make a convincing voice, face, or message insufficient on its own to trigger irreversible action.
Why governance has to be stricter when synthetic media is credible
Synthetic voice, video, and text reduce the reliability of the most natural human shortcut: believing the channel that seems familiar. That shifts governance from “does this request sound right?” to “has it been independently verified against an approved decision path?” High-risk actions need controls that survive impersonation, urgency, and executive-looking messages.
The practical implication is that organisations should treat synthetic media as an amplification layer for social engineering, not as the core control problem. The control objective is to prevent a convincing message from becoming a standalone authorization event for money movement, privileged access, or sensitive approvals.
For teams building awareness around this threat, a focused treatment such as Deepfakes, Social Engineering and AI Impersonation Guide is useful because it ties the impersonation problem to payment controls and out-of-band verification rather than to media authenticity alone.
Which requests should be treated as high-risk by default?
Not every request deserves the same friction. The requests that most justify hard governance are the ones that can create irreversible or hard-to-reverse impact: payments, bank detail changes, vendor or payroll changes, account recovery, credential resets, wire instructions, and approvals that grant access, privilege, or contractual commitment. These are the requests where impersonation can quickly become loss.
Governance should define which action types are always in scope, which are conditional, and what evidence is required before approval. The important distinction is between a routine operational ask and a request that changes authority, money flow, or access posture. Once a request can alter those states, the approval path must be explicit and repeatable.
Organisations that want a broader framework for control design can map this kind of decision discipline to NIST SP 800-53 Rev 5 Security and Privacy Controls, especially around access control, authentication, and auditability. For operational governance and monitoring, NIST Cybersecurity Framework 2.0 is a useful umbrella for defining who approves, who verifies, and how exceptions are tracked.
What should the approval model look like in practice?
The approval model should separate request intake, independent verification, and final execution. A message may trigger review, but it should never be enough by itself to execute an irreversible action. The strongest pattern is to use pre-defined approvers, known-good callback or out-of-band verification, and explicit conditions for when a request can be actioned.
That also means governance needs to define escalation paths for exceptions. If a request is urgent, unusual, or comes through an unexpected channel, the response should be to slow the action, not to improvise a new approval rule in the moment. Good governance makes the safe path easy and the exception path deliberate.
When the organisation is also setting policy for AI-enabled content and provenance, NIST AI 600-1 GenAI Profile and the EU AI Act regulatory framework help frame the broader governance expectation: treat content authenticity as a risk signal, not a trust guarantee.
Risk and Threat Considerations
Synthetic media increases the chance of trust abuse, especially when staff are trained to react quickly to urgent requests from senior-looking voices or messages. The main exposure is not the media itself, but the decision process that lets a believable impersonation bypass verification and trigger a payment, access change, or sensitive approval.
Failure mechanism: The attacker or impersonator exploits urgency, authority bias, or channel confusion so that staff accept the request as authenticated without checking it against an independent control path.
Impact: Organisations can suffer fraud, unauthorized access, account takeover, contract changes, or irreversible financial loss before the deception is detected.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
OWASP Agentic AI Top 10 addresses the attack and risk surface, while NIST SP 800-53 Rev 5 and NIST CSF 2.0 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| NIST SP 800-53 Rev 5 | AC-3 — Access Enforcement | High-risk approvals often grant or change access, so approval rules must be enforced before action. |
| IA-2 — Identification and Authentication (Organizational Users) | Synthetic impersonation makes independent authentication central to request handling. | |
| AU-2 — Event Logging | High-risk approvals need traceable records of who approved what and when. | |
| Recommendation — Enforce approval conditions before privileged or sensitive changes are executed. Require strong authentication before accepting high-risk approval requests. Log approval, verification, and exception events for later review. | ||
| NIST CSF 2.0 | GV.RM-01 — Risk Management Strategy | Governance must define how the organisation handles impersonation-driven approval risk. |
| Recommendation — Define risk thresholds and mandatory verification for irreversible requests. | ||
| OWASP Agentic AI Top 10 | ASI09 — Human-Agent Trust Exploitation | Synthetic media exploits trust in familiar voices, faces, and messages. |
| Recommendation — Add independent verification before acting on persuasive synthetic content. | ||
Practitioner Guidance
What to prioritise: Start with the few request types that can cause the greatest irreversible harm, then define mandatory verification steps for each. Payments, beneficiary changes, account recovery, and privilege changes should never rely on a single message, call, or video clip.
What to verify: Confirm that approval authority is documented by role, dollar threshold, business context, and channel, and that the verifier is different from the requester where the risk is material. If the process cannot be independently replayed from records, it is too informal for high-risk requests.
Common mistake: Teams often add awareness training but leave the approval path unchanged. Training helps, but the control must be procedural and observable, otherwise a convincing synthetic voice can still bypass the workflow.
Practitioner takeaway: Treat synthetic media as a reason to harden approval governance, not to debate authenticity in isolation, because the control objective is to make irreversible action impossible without independent verification.
Related resources from NHI Mgmt Group
- When should organisations treat an NHI as a high-priority risk?
- How should security teams govern access requests for high-risk cloud resources?
- Why do business email compromise and synthetic identity attacks create such high risk for organisations?
- How should organisations govern collaboration and social media tools so they reduce insider risk without losing business value?