TL;DR: A major U.S. healthcare provider found bot attacks account for more than half of all fraud in its systems, according to Pindrop, with more than 15,000 unique bot fraud calls since summer 2025 and $40M in account exposure tied to fraudulent AI bot calls. Static verification is failing because scripted, high-speed interactions can harvest and reuse personal data faster than legacy controls can respond.
At a glance
What this is: This is an analysis of AI-enabled bot fraud in healthcare contact centres, showing that scripted calls are now driving a majority of fraud in one provider’s systems and are being used to probe, social-engineer, and take over accounts.
Why it matters: It matters because healthcare identity assurance now has to distinguish legitimate AI-assisted callers from fraud bots while protecting contact-centre workflows, account access, and high-value member accounts.
By the numbers:
- Pindrop says a U.S. healthcare provider faced over $40M in account exposure related to fraudulent AI bot calls in 2025.
- Nearly 60% of organisations now report fraudsters using compromised personally identifiable information to quickly bypass knowledge-based authentication, according to Pindrop.
Context
Healthcare contact centres are now a fraud interface as much as a service channel. When attackers can use scripted calls to probe IVR systems, harvest personal data, and hand off to live-agent social engineering, the traditional boundary between authentication and conversation breaks down.
The problem is not only automation, but scale and adaptation. Bots that avoid obvious text-to-speech artefacts can still behave in a scripted, machine-paced way, which means defenders need controls that evaluate interaction patterns, data access, and identity signals together, not in isolation.
Key questions
A: Treat the contact centre as part of the identity perimeter. Prioritise behavioural detection on IVR and live-agent flows, raise verification requirements for high-value transactions, and remove static knowledge checks as a primary trust decision where compromised PII is likely. The goal is to make scripted abuse harder to scale without creating unnecessary friction for legitimate callers.
Q: Why do compromised personal data and KBA make healthcare fraud easier to scale?
A: Because attackers no longer need to discover identity facts from scratch. They can recycle stolen PII, answer predictable challenge questions, and use that confidence to move into live-agent manipulation. Once the same data is reused across many calls, KBA becomes a replayable control rather than a proof of identity.
Q: What are the signs that a contact-centre fraud scheme is automated rather than human?
A: Repeated IVR probing, unusual speed between prompts, scripted language patterns, consistent command-like phrasing, and call characteristics that resemble a coordinated operation rather than a spontaneous conversation are strong indicators. The most useful signal is not a single audio artifact, but a cluster of behavioural traits that stay consistent across many calls.
Q: What should teams do when AI is used for legitimate assistance and for fraud in the same channel?
A: Define the trust boundary by use case. Legitimate translation or support tooling should be allowed only when the caller’s risk level is acceptable, while the same interaction should trigger stronger verification when the session touches balances, account changes, or sensitive personal data. Governance must separate acceptable AI assistance from identity abuse.
Technical breakdown
How IVR reconnaissance becomes a fraud entry point
Interactive voice response systems often reveal enough structure for an attacker to map account pathways, confirm knowledge-based facts, and test which prompts expose useful signals. Once a bot can traverse that layer, the contact centre itself becomes a discovery surface, not just a service channel. In healthcare, that matters because balances, account status, and personal data can all be exposed through ordinary support flows. The attack is successful before any live agent is involved because the bot has already learned which branches, verification steps, and account types are worth targeting. Read Top 10 Online Fraud Attacks.
Practical implication: Instrument IVR paths for abuse patterns and treat repeated menu traversal as a security signal, not only a customer-service event.
Why scripted voice interactions beat legacy verification
Legacy knowledge-based authentication assumes personal data is both hard to obtain and hard to reuse. That assumption fails when stolen PII is already circulating and bots can replay it at machine speed while sounding operationally normal. The article’s point about programming-style commands is important: the fraud signal is behavioural, not just audio-based. If defenders rely on voice artifacts alone, they miss the more durable indicator, which is an interaction sequence that looks human enough to pass timing and flow checks while still being fully scripted. See the Identity Fraud Prevention Guide.
Practical implication: Shift verification from static challenge questions toward behavioural and risk-based authentication signals that evaluate the session, not the answer alone.
How account takeover connects to downstream identity abuse
The path does not stop at contact-centre probing. Once attackers have enough context, they can pivot into live-agent manipulation, account takeover, and access to employer-funded savings accounts such as HSAs and FSAs. That makes the fraud problem an identity governance problem as well, because the contact centre can become the point where stolen identity evidence is converted into authorisation. In practical terms, the most sensitive accounts require tighter step-up controls, stronger agent verification workflows, and clearer separation between communication convenience and transaction approval.
Practical implication: Map high-value account journeys to stronger step-up verification and transaction controls before they can be abused through the service channel.
NHI Mgmt Group analysis
Bot fraud is now a contact-centre identity problem, not just a fraud problem. When scripted callers can extract enough data to manipulate live agents, the service channel becomes part of the identity attack surface. That means healthcare organisations need to govern conversation flows, step-up thresholds, and high-value account access as one control plane, not separate teams with separate signals. Practitioners should treat the voice channel as an identity-dependent workflow.
Legacy knowledge-based authentication has lost most of its defensive value in healthcare. The article shows why compromised PII makes KBA an unsafe trust anchor for high-friction calls. Once fraudsters can reuse personal data at scale, the control is no longer proving identity, only proving access to leaked data. That is why identity fraud prevention must move toward behavioural, contextual, and transaction-bound signals. Practitioners should retire KBA as a primary trust decision for sensitive accounts.
AI-assisted callers blur the line between legitimate assistance and malicious automation. The article correctly notes that AI can support translation or administrative help, but the governance challenge is that the same channel can carry both service assistance and fraud. Organisations need policy boundaries that distinguish permitted AI mediation from scripted abuse without blocking legitimate users. That is an identity assurance design issue, not a customer-service nuance. Practitioners should define when AI assistance is allowed and when it must trigger stronger verification.
Healthcare fraud is becoming a repeatable industrial process, which changes the control objective. More than 15,000 unique bot fraud calls in a short period signals repeatability, not opportunism. Repeated, machine-paced attacks mean defenders must optimise for detection speed, signal fusion, and response consistency rather than isolated case handling. The operational goal is to make every high-risk call expensive for the attacker and low-friction for the legitimate member. Practitioners should design for scale, not anecdote.
From our research library:
- 55% of healthcare data breaches now originate from a third-party vendor, according to Ponemon Institute’s 2023 Third-Party Risk in Healthcare report.
What this signals
Healthcare fraud is shifting from isolated incidents to repeatable machine-enabled campaigns. That changes the operational target from individual case closure to pattern suppression across IVR, live-agent, and account-access layers. For practitioners, the practical test is whether one bot can be blocked without the next bot immediately succeeding through the same path.
Voice-channel controls now need to behave like identity controls. The strongest defence is no longer a single authentication question, but a layered view of interaction behaviour, customer context, and transaction sensitivity. In programmes that still treat contact-centre calls as a helpdesk issue, fraud will keep outrunning the control design.
For practitioners
- Tighten IVR abuse detection Monitor menu traversal, retry patterns, and prompt sequencing for bot-like reconnaissance across healthcare contact flows.
- Replace KBA as the primary trust check Use risk-based and behavioural verification for sensitive calls, especially when personal data may already be compromised.
- Separate service convenience from transaction approval Require stronger step-up controls before account changes, balance access, or savings-account actions in the contact centre.
- Train agents on scripted social engineering Give frontline staff a short decision path for when caller behaviour, background noise, or scripted language indicates automation.
Key takeaways
- The core risk is not only fraud volume, but the way bot-driven calls turn healthcare contact centres into a scalable identity attack surface.
- The article cites more than 15,000 unique bot fraud calls and over $40M in account exposure, which points to repeatable attack economics rather than one-off abuse.
- Practitioners need stronger behavioural detection, tighter step-up verification, and a move away from static knowledge-based checks for sensitive account journeys.
Related NHI Mgmt Group resources: Identity Fraud Prevention Guide · Top 10 Online Fraud Attacks · AI Agent Authorisation Guide
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-10 — Human Use of NHI | The article centres on automated bot activity used to manipulate human-assisted identity and fraud workflows. |
| NHI-04 — Insecure Authentication | The article shows static knowledge checks failing against compromised PII and scripted callers. | |
| Recommendation — Separate legitimate human-assisted workflows from bot-driven interaction paths and require stronger checks where automation is suspected. Replace static knowledge-based checks with stronger, context-aware authentication for sensitive contact-centre actions. | ||
| NIST SP 800-63 | SP 800-63B — Authentication | The fraud scheme exploits weak authentication decisions in call-centre identity workflows. |
| Recommendation — Apply assurance-based authentication controls to sensitive service-channel journeys and step up when risk increases. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | Account takeover and savings-account access hinge on authorisation decisions after identity proofing. |
| Recommendation — Tighten authorisation for high-value account actions and separate verification from transaction approval. | ||
| MITRE ATT&CK | TA0006;TA0008 — Credential Access; Lateral Movement | Bots harvest personal data and pivot from IVR reconnaissance into account takeover through human-assisted escalation. |
| Recommendation — Map scripted contact-centre abuse to credential access and lateral movement techniques in your detection pipeline. | ||
Key terms
- Interactive Voice Response: Interactive Voice Response is an automated phone system that lets callers interact with menus and prompts using speech or keypad input. In security terms, it is a transaction layer that can expose identity checks, account access, and backend integrations, which makes weak authentication, poor validation, and excessive feedback especially risky.
- 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.
- Behavioural fraud signal: A behavioural fraud signal is a pattern in how an account acts that suggests misuse, even when authentication appears valid. Examples include device changes, IP shifts, unusual transaction values, or profile edits that do not fit the historical customer pattern.
- Contact-Centre Identity Attack Surface: The set of voice workflows, agent interactions, and verification steps through which attackers can influence identity decisions. In healthcare, this surface matters because the contact centre can convert harvested personal data into account access, service manipulation, or financial loss.
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. Explore nhimg.org for resources that connect identity governance to the broader security disciplines your programme depends on.
Published by the NHIMG editorial team on October 22, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org