TL;DR: AI-enabled phishing is rising sharply, including a 41% increase in teams-based attacks, a 449% surge in vishing, and a 92% jump in polymorphic attacks, according to KnowBe4 research. Human risk management, not awareness alone, is now the control gap that matters most.
At a glance
What this is: KnowBe4’s analysis says AI-enabled phishing is increasing across email and collaboration channels, with human risk management emerging as the more relevant control gap than awareness training alone.
Why it matters: It matters because IAM, email security, and SOC teams now have to govern human decision points and identity trust paths that attackers can exploit at machine speed.
Context
AI-enabled phishing is no longer limited to obvious email scams. It now blends email, collaboration tools, voice, and impersonation to exploit trust at the point where users decide whether to click, approve, or respond. In identity terms, the weak point is not just message filtering, but the human and account trust chain that sits behind it.
For IAM and security teams, this is a governance problem as much as a detection problem. Human risk management, email identity controls, and account recovery discipline all intersect here, because attackers increasingly target the relationships between people, identities, and channels rather than a single inbox.
The article’s examples fit a now-typical pattern: phishing is becoming more adaptive, more multi-channel, and more dependent on social engineering than on malware delivery alone.
Key questions
Q: What breaks when AI phishing is treated as only an email problem?
A: Teams miss the identity layer of the attack. AI phishing often aims to capture credentials, sessions, or trusted actions, so filtering messages without correlating authentication and account behaviour leaves the highest-risk part of the compromise unaddressed. Effective defence has to connect inbox signals to identity and response workflows.
Q: Why do AI phishing attacks create more risk than traditional phishing?
A: AI lowers the cost, time, and skill needed to produce personalised lures, so attackers can run more campaigns and iterate faster. That increases both exposure and realism. The result is a higher probability that a target will trust a message long enough to hand over credentials or payment information.
Q: What are the signs that phishing-resistant controls are not being applied effectively?
A: A common warning sign is that users still authenticate with passwords, SMS codes, or push approvals for sensitive access while rare device registrations and unusual login events are not being reviewed. Another indicator is that suspicious requests are handled in the same communication channel as the original message. If controls do not force independent verification and create clear alerting on high-risk events, they are not working as intended.
Q: How should teams govern email and collaboration-channel trust together?
A: Teams should govern them as one trust path because attackers move between email, chat, and voice to find the weakest verification point. That means aligning email security, SSO safeguards, help-desk checks, and approval workflows around the same risk signals. The goal is to stop identity abuse wherever the request first appears.
Technical breakdown
Why AI-powered phishing bypasses traditional email controls
AI-powered phishing reduces the telltale signals that older email controls were built to catch. Polymorphic wording, better impersonation, and rapid channel switching make message signatures less reliable, while collaboration platforms and voice introduce trust cues outside the inbox. That shifts the security challenge from message inspection to contextual verification, identity assurance, and behavioural detection. In practice, the attacker is not trying to defeat one filter, but to move the victim into a channel where the organisation has weaker verification and slower response.
Practical implication: expand detection and verification beyond email-only controls, especially where users can approve requests in chat, voice, or collaboration tools.
Human risk management as an identity control problem
Human risk management is best understood as a control layer around decision-making, not a replacement for awareness training. Traditional awareness assumes people can be taught to recognise every lure in advance. That breaks down when attacks are personalised, time-compressed, and delivered through trusted channels. The stronger model is to instrument risky behaviours, measure susceptibility, and apply step-up controls where a user’s action could expose credentials, payments, or privileged access. This makes human identity governance part of the broader trust architecture.
Practical implication: treat risky user actions as governed events and connect them to identity, access, and response workflows.
Machine-speed social engineering changes the defence model
When phishing adapts faster than users can interpret it, defenders have to assume that prevention will be incomplete. The relevant mechanism is speed: attackers can iterate lures, shift channels, and personalise content faster than manual review or ad hoc training can respond. That creates a gap between exposure and intervention. Security programmes therefore need telemetry on user interaction patterns, centralised reporting, and automated containment paths that reduce the time between suspicious interaction and control action.
Practical implication: shorten the path from user report to containment by automating triage, suppression, and account review workflows.
Threat narrative
Attacker objective: The objective is to convert human trust into access, financial theft, or downstream account abuse without triggering the controls built for conventional phishing.
- Entry begins with AI-generated phishing delivered through email, collaboration tools, or vishing, using tailored language and trusted-looking context to lower suspicion.
- Credential harvest or trust exploitation follows when the victim clicks, consents, shares information, or authorises a request that appears legitimate within the chosen channel.
- Impact occurs when the attacker turns that trust into account compromise, payment fraud, or broader access to the organisation’s communication and identity environment.
Breaches seen in the wild
- CoPhish OAuth phishing via Copilot Studio: Datadog showed Copilot Studio agents on a Microsoft domain can front OAuth consent phishing and forward stolen tokens; no victims reported.
Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.
NHI Mgmt Group analysis
Human risk management is now an identity governance issue, not just an awareness issue. AI phishing succeeds when organisations treat user judgment as the last line of defence rather than a governable control surface. The practical shift is toward measuring risky behaviour, enforcing step-up verification, and connecting human decisions to identity controls.
AI phishing compresses the time available for detection and response. That changes the control objective from perfect prevention to faster interruption. Email security still matters, but it is no longer sufficient on its own when threats move across inbox, chat, and voice with little friction. Practitioners should expect more cross-channel trust abuse, not more obvious spam.
Email identity and workforce identity now overlap operationally. Attackers increasingly exploit trusted senders, delegated approvals, and help-desk-style confidence cues, which means email security, SSO discipline, and account recovery controls belong in the same governance conversation. The field is moving toward trust-path management, where practitioners must know which identities can vouch for which actions.
Human risk telemetry will become a core signal for security programmes. Organisations that can measure behavioural exposure, repeat susceptibility, and response latency will have a better basis for targeted controls than those relying on annual training alone. The lesson is to govern human risk like any other attack surface: observed, measured, and bounded.
What this signals
Human risk management is becoming a control architecture, not a communications programme. Security teams that still treat phishing as a training problem will miss the operational issue, which is governing risky decisions across channels. The practical shift is toward measuring exposure, linking it to identity events, and using that data to trigger intervention.
Email security and identity security are converging at the workflow level. The next failure mode is not a single malicious message but a chain of trusted actions across inbox, chat, voice, and account recovery. Practitioners should map those trust paths now, before attackers normalise them as routine business interaction.
For practitioners
- Strengthen human risk telemetry Track repeated risky behaviours such as link clicks, credential submission, and approval of unexpected requests, then use those signals to target interventions and step-up controls.
- Expand verification beyond email Require secondary verification for money movement, password resets, and sensitive approvals when requests arrive through chat, voice, or collaboration tools.
- Instrument collaboration-channel trust paths Review how approval flows, file sharing, and delegated messaging work across Microsoft 365, Teams, and similar tools so phishing cannot hide inside legitimate workflows.
- Automate phishing report triage Feed user-reported messages into a response workflow that can quarantine similar messages, review related accounts, and alert the SOC without manual delay.
Key takeaways
- AI phishing is now a multi-channel trust problem, not just an inbox filtering problem.
- The article’s core signal is that human risk management has to be governed as part of identity and access control.
- Security teams should connect behavioural telemetry, verification steps, and automated containment to reduce exposure faster.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
NIST CSF 2.0, NIST SP 800-53 Rev 5, CIS Controls v8 and OWASP ASVS set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | Phishing often succeeds by tricking users into authorising access or actions they would not otherwise permit. |
| DE.CM-03 — Detect malicious code | Behavioural phishing detection depends on monitoring suspicious user interactions and reporting signals. | |
| Recommendation — Review approval and authentication flows so risky user actions trigger step-up verification under PR.AA-05. Build monitoring and reporting pipelines that surface suspicious user interactions early under DE.CM-03. | ||
| NIST SP 800-53 Rev 5 | IA-5 — Authenticator Management | Phishing often targets credentials, recovery paths, and authentication bypass points. |
| Recommendation — Harden authenticator lifecycle controls so stolen or misused credentials cannot be reused easily under IA-5. | ||
| CIS Controls v8 | CIS-5 — Account Management | Account misuse and rapid containment rely on strong account governance and response. |
| Recommendation — Apply account governance controls to disable, review, and recover suspect accounts quickly under CIS-5. | ||
| OWASP ASVS | V10 — OAuth and OIDC | OAuth consent phishing is part of the email-security threat pattern discussed in the source article. |
| Recommendation — Review consent and federation flows so OAuth abuse is blocked with stronger approval checks under V10. | ||
Key terms
- Human Risk Management: The practice of managing how people interact with security controls, especially under pressure, distraction, or deception. It combines training, policy, and friction management so identity systems are still usable enough that users do not bypass them in day-to-day work.
- AI-Personalised Phishing: Phishing that uses context, tone, and timing tailored to the target rather than generic mass messaging. In practice, it reduces obvious red flags and forces defenders to rely on identity, behaviour, and channel correlation instead of message signatures alone.
- Trust Path: A trust path is the sequence of systems a request passes through before identity controls or service access complete. It includes resolution, routing, verification, and policy enforcement, so weak controls anywhere in the path can undermine the assurance of the final access decision.
- Step-Up Verification: Step-up verification is a stronger identity check applied when risk increases, such as during password reset, device change, or privileged access request. It uses higher-assurance signals than a static question, such as device possession, authenticated context, or approved administrative review.
Deepen your knowledge
NHI Mgmt Group covers identity security, NHI governance, and agentic AI through independent research, practitioner guides, and the NHI Foundation Level course, the industry's only accredited NHI security programme. It is designed for practitioners who need to govern identities, privileges, and trust paths across modern security programmes.
Published by the NHIMG editorial team on August 2, 2026.
Updated on October 12, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org