TL;DR: AI has driven the cost of phishing, voice cloning and deepfake document fraud against knowledge-based authentication toward near zero, according to IDlayr, while mobile possession factors offer structural resistance because they depend on something physically held and verified at the device or network layer. Passwords, PINs, SMS OTP and even biometrics no longer hold the trust boundary together when attackers can replicate the human side of authentication at scale.
At a glance
What this is: This is an analysis of how AI-enabled fraud is eroding knowledge-based consumer authentication and why possession-based controls are the remaining structurally resilient option.
Why it matters: It matters because IAM teams, fraud leads and security architects now have to design authentication that survives phishing, deepfakes and relay attacks without relying on user judgment.
Context
AI-enabled fraud is no longer limited by attacker skill or manual effort. The article argues that passwords, PINs, SMS OTP and knowledge questions are now too easy to steal, replay or synthesize when voice cloning, deepfake documents and phishing relay kits are widely available.
The identity problem is not just stronger fraud tooling. It is that consumer authentication models still assume the user can notice deception and protect a secret, even when the attack path now targets mobile devices, account recovery and agent-driven commerce at the same time.
Key questions
Q: What should teams do first when replacing SMS OTP?
A: Start with the single user journey that has the highest fraud loss or abandonment rate, then run the replacement in shadow mode before changing the live experience. That sequence gives you real performance data, lets you size the business case, and reduces the risk of scaling a control that has not been proven in production.
Q: Why do knowledge-based factors fail so quickly under AI-enabled fraud?
A: They fail because the attacker no longer needs to break the secret directly. AI can imitate voices, generate convincing documents and automate phishing at scale, which turns the human verifier into the weak point. Once the control depends on recognition or recall, the attacker only needs to produce a believable prompt or request.
Q: What are the signs that consumer authentication is too easy to socially engineer?
A: Look for repeated OTP interception, support-team bypasses, recovery abuse and fraud attempts that succeed without malware. If attackers can win by sounding legitimate, producing fake documents or steering users into entering a code, the current control set is still anchored in human judgment rather than resilient possession.
Q: What is the difference between possession-based authentication and passkeys in practice?
A: Possession-based authentication proves control of a specific device or SIM at the moment of verification. Mainstream passkeys often sync through the cloud, so they reduce password risk but do not always create the same hard device binding. For fraud-prone consumer journeys, the difference matters because only the stronger binding resists replay and social engineering.
Technical breakdown
Why knowledge-based authentication fails under AI fraud
Knowledge-based authentication depends on information the user knows, such as a password, PIN or one-time code. AI changes the economics of that model because the attacker no longer needs to guess a secret manually. Voice cloning, multilingual phishing and deepfake documents let the attacker reproduce the signals that humans and legacy fraud checks treat as proof. Once the human is the verification step, the attacker only has to persuade or mirror the user. That makes the trust decision dependent on recognition, speed and confusion, all of which AI can manipulate at scale.
Practical implication: stop treating user-entered secrets as the anchor for consumer identity assurance.
Why possession-based authentication is structurally different
Possession-based authentication shifts the trust decision from what a user knows to what they physically hold, such as a specific device or SIM card. The key distinction is that possession cannot be copied by language, image generation or social engineering. A carrier-verified or device-bound check creates a deterministic yes or no answer without asking the user to reveal a reusable secret. That removes the interception step that phishing and relay attacks depend on. In identity terms, the control moves from user behaviour to physical binding.
Practical implication: require proof of possession at the network or device layer for high-risk consumer flows.
How mobile identity becomes the trust anchor for agentic commerce
The article extends the problem into agentic commerce, where software agents act on behalf of users in transactions. That introduces an identity chain that now includes the person, the mobile device, the SIM and the application or agent. If that chain is weak, attackers can hijack sessions or inject false identity into the transaction path. The governance challenge is not just account authentication but transaction attribution. Every high-risk action needs to remain cryptographically bound to a real user and a real device, not a recovered session or a shared secret.
Practical implication: bind sensitive agent-led transactions to a verified device identity before approval or execution.
Threat narrative
Attacker objective: The attacker’s objective is to impersonate the user convincingly enough to steal access, authorize transactions or bypass consumer authentication controls.
- Entry occurs through AI-assisted phishing, voice cloning or deepfake documents that replace a legitimate authentication request with a convincing imitation.
- Credential capture or replay follows when the victim enters a password, PIN or SMS OTP into the attacker-controlled flow or reveals enough context for account takeover.
- Impact is account compromise, payment fraud or fraudulent session initiation, with agent-led commerce adding a new path for identity injection and transaction abuse.
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.
- Mailchimp breach 2022: Attackers socially engineered Mailchimp staff, used a support tool to export 102 customer lists and exposed customer API keys for phishing.
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
AI has collapsed the economics of identity fraud by turning the human trust step into the primary attack surface. When authentication still depends on the user spotting deception, attackers only need to produce something convincing enough to trigger a response. That shift invalidates consumer IAM designs that assume people can reliably distinguish real prompts from synthetic ones. The implication is that identity programmes must stop pricing in human vigilance as a control.
Knowledge-based identity was designed for a world where secrets stayed secret long enough to matter. That assumption fails when secrets can be stolen, inferred or replayed at machine scale in seconds. Passwords, PINs and SMS OTP are not weak because users are careless; they are weak because the underlying trust model has become economically obsolete. The implication is that consumer authentication architecture has to move away from secret transmission as the core security boundary.
Mobile possession factors create the first practical trust model that is resistant to synthetic media attacks. A device or SIM cannot be phished, narrated or deepfaked into existence. That makes possession the only control in this article with structural resilience rather than just better detection. Practitioners should treat possession-based verification as the new baseline for high-value consumer journeys.
Agentic commerce turns authentication into an attribution problem as much as a fraud problem. Once software agents can initiate or complete transactions, the real question becomes which identity is allowed to act and on what device. This broadens consumer identity governance from login assurance to transaction binding. The implication is that identity teams need controls that hold up when a human, device and agent all participate in the same workflow.
Ephemeral assurance is replacing reusable secrets as the correct mental model for modern consumer auth. Authentication should prove a live relationship between user, device and network rather than preserve a credential that can be copied later. That changes design priorities across IAM, fraud and mobile channels. Practitioners should focus on verifiable possession and session binding, not stronger versions of the same shared-secret pattern.
What this signals
Mobile possession becomes the new trust anchor when AI can replicate every human-facing factor. Consumer identity programmes that still centre passwords, OTPs or recovery questions are now defending the wrong boundary. The control question is shifting from whether a user can recognise fraud to whether the verifier can independently prove possession at the device layer.
Agentic commerce will force identity teams to govern transaction attribution, not just login success. When software can act for a person, the security model needs to show which device, which user and which session authorised the action. That pushes IAM and fraud teams toward stronger binding between identity, device and transaction context.
For practitioners
- Replace SMS OTP on high-risk flows Move the most fraud-sensitive consumer journeys away from SMS codes and other user-entered secrets. Prioritise account recovery, payments and onboarding paths where phishing and relay attacks are most likely to succeed.
- Bind authentication to device possession Use device- or SIM-verified checks for step-up authentication so the trust decision is tied to something the attacker cannot copy remotely. Make sure the control is validated at the network or device layer rather than through user recall.
- Rework fraud controls for AI-generated deception Assume callers, documents and support interactions can be synthetically generated at production quality. Update verification paths so they do not depend on a person detecting whether the interaction feels authentic.
- Add identity binding to agent-led transactions Require a verified device identity before software agents can approve or execute high-value actions. Track the link between the user, the device and the agent session so attribution survives automation.
- Retire human-in-the-loop trust prompts Remove workflows that ask users to validate authenticity through memory, pattern recognition or manual code entry. Those checks place the burden on the easiest target in the chain and are increasingly unreliable under AI-driven fraud.
Key takeaways
- AI-enabled fraud is breaking consumer authentication models that rely on secrets, human recognition or user-entered codes.
- The article argues that possession-based verification is the only control family with structural resilience against synthetic media and relay attacks.
- For practitioners, the shift is from strengthening shared secrets to binding high-risk actions to a verified device or SIM.
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 MITRE ATT&CK address the attack and risk surface, while NIST SP 800-63 and NIST CSF 2.0 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Non-Human Identity Top 10 | NHI-04 — Insecure Authentication | The article centres on authentication methods that AI can now defeat through phishing and synthesis. |
| NHI-10 — Human Use of NHI | The article highlights human-mediated verification as the weak point in modern fraud chains. | |
| Recommendation — Replace knowledge-based consumer auth with stronger possession-bound verification where fraud risk is highest. Remove user-dependent verification steps from flows that attackers can socially engineer. | ||
| NIST SP 800-63 | SP 800-63B — Authentication | The topic is consumer authentication and the strength of authenticators under phishing pressure. |
| Recommendation — Map high-risk journeys to phishing-resistant authenticators and reduce reliance on shared secrets. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | The article focuses on how access decisions depend on reliable identity proofing and authorization. |
| Recommendation — Align authorization decisions with stronger identity proofing so transactions are not granted on weak signals. | ||
| MITRE ATT&CK | TA0006;TA0008 — Credential Access; Lateral Movement | The article describes phishing relay, credential capture and downstream account abuse patterns. |
| Recommendation — Map fraud relay and OTP theft to credential access paths and monitor for rapid session abuse. | ||
Key terms
- Knowledge-based authentication: A verification method that asks a person to prove identity by answering facts such as past addresses, account details, or other shared information. It is operationally convenient but weak against social engineering, breach data, and impersonation, which is why it performs poorly in high-risk support flows.
- Possession-Based Authentication: An authentication method that relies on something the user has, such as a security key or registered device. It improves assurance by linking access to a physical or managed object, but the security outcome depends on enrolment, revocation, loss handling, and whether the possession factor can be cloned or bypassed.
- Silent Network Authentication: A phone-number verification method that confirms device and SIM association through the carrier network instead of sending a one-time code. It reduces exposure to phishing and SMS pumping because the user does not need to read or return a secret.
- Agentic Commerce: Agentic commerce is a buying and transaction model where software agents act on behalf of a person. The identity challenge is not just proving who owns the account, but constraining what the agent may do, for how long, and under what revocation and audit rules.
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.
Published by the NHIMG editorial team on July 11, 2026.
Updated on October 11, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org