Because the attack is no longer a single event. Deepfakes, synthetic identities, and injection attacks can pass the first gate and then behave normally long enough to stay trusted. Once the organisation treats that first pass as durable proof, the attacker only needs to preserve the illusion of legitimacy.
Why This Matters for Security Teams
Point-in-time checks are built to answer a narrow question: did this request look legitimate at the moment it crossed the gate? AI-driven fraud breaks that model because the attacker only needs the system to be convinced once, then can keep adapting. A deepfake voice, synthetic identity, or prompt injection can pass an initial check and then continue operating under the cover of normal behaviour.
That is why static trust decisions become risky. Once an identity, device, or transaction is marked trusted, many controls stop looking for change. Current guidance suggests organisations should shift from single-verification thinking toward continuous, context-aware evaluation, especially where fraud chains can unfold over multiple steps. The control problem is not just authentication, but persistence of legitimacy across the whole interaction.
NHIMG research on the DeepSeek breach shows how quickly attackers can exploit exposed or compromised credentials, and NIST’s NIST SP 800-53 Rev 5 Security and Privacy Controls reinforces the need for monitoring beyond initial access approval. In practice, many security teams encounter fraud only after the first trust decision has already been treated as durable proof.
How It Works in Practice
Defending against AI-driven fraud requires treating identity as something that must be re-evaluated, not merely established. A point-in-time check may validate a document, a voice sample, a device fingerprint, or a login challenge, but fraudsters can layer techniques so the first pass is only the beginning. The better pattern is to combine initial verification with runtime signals such as behaviour, transaction context, device consistency, session integrity, and anomalies in tool use or navigation.
That is why point-in-time checks should be paired with continuous controls. In practice, teams often use:
- step-up verification when risk changes mid-session;
- behavioural analytics to detect drift after initial approval;
- short-lived credentials or tokens so a trusted session cannot persist indefinitely;
- policy decisions that are re-evaluated for each high-risk action;
- correlation across channels, since fraud often starts in one channel and completes in another.
This aligns with current NIST guidance on layered control design, including transaction monitoring, access enforcement, and auditability. It also matches NHIMG guidance in the DeepSeek breach analysis, where initial compromise becomes dangerous only when the attacker can keep operating without fresh scrutiny. For identity and secrets governance, the State of Secrets in AppSec report highlights how exposed secrets and slow remediation create a long tail of risk that point checks cannot absorb.
These controls tend to break down in high-volume customer journeys where teams optimise for low friction and disable re-checks to reduce abandonment.
Common Variations and Edge Cases
Tighter verification often increases friction, operational overhead, and false positives, so organisations must balance fraud reduction against customer impact. That tradeoff becomes sharper in edge cases where legitimate users also behave unpredictably, such as travel, shared devices, assistive technologies, or urgent account recovery.
Best practice is evolving, and there is no universal standard for exactly how often to re-check identity in fraud workflows. Some environments can rely on transaction-level risk scoring, while others need stronger step-up controls for payouts, account changes, or high-value transfers. The key is to avoid treating one successful proof event as a permanent trust decision.
For deeper control design, NIST’s Security and Privacy Controls can be used to map monitoring, authentication, and anomaly detection requirements, while NHIMG’s DeepSeek breach coverage is a useful reminder that fraud and compromise often persist after the first successful bypass. The practical failure mode is simple: once a team lowers scrutiny after a single clean check, the attacker only needs to maintain the performance of legitimacy.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
OWASP Non-Human Identity Top 10 and OWASP Agentic AI Top 10 address the attack and risk surface, while NIST CSF 2.0, NIST SP 800-63 and NIST AI RMF set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| NIST CSF 2.0 | DE.CM-1 | Continuous monitoring is needed because fraud evolves after the first check. |
| NIST SP 800-63 | AAL2 | Assurance must be maintained beyond one successful identity proofing event. |
| NIST AI RMF | AI-driven fraud requires ongoing risk assessment, not only initial verification. | |
| OWASP Non-Human Identity Top 10 | NHI-04 | Static trust in identities and secrets enables abuse after initial access. |
| OWASP Agentic AI Top 10 | A1 | Autonomous systems can preserve legitimacy across multiple steps after first access. |
Operate continuous AI risk monitoring and revisit trust decisions as context changes.
Related resources from NHI Mgmt Group
Deepen Your Knowledge
Reviewed and updated by the NHIMG editorial team on August 11, 2026.
NHI Mgmt Group — the #1 independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org
