# Harvard/Meta Study Quantifies AI Voice-Phishing (Vishing) Persuasiveness Gap: 16.5% Compliance, 70.3% Detection, and Profitable Economics for Attackers

> A peer-reviewed study by Harvard Kennedy School, Harvard SEAS, and Meta researchers (arXiv:2607.09970, accepted to Expert Systems with Applications) surveyed 4,100 US adults against recordings from six commercial AI voice systems (Llama Full-Duplex, OpenAI AVM, Gemini, Sesame, Play.AI, ElevenLabs) and human callers performing vishing scenarios. Participants correctly identified synthetic voices only 70.3% of the time, falsely flagged real humans as synthetic 54.2-67.4% of the time, and overall scam compliance averaged 16.5%, reaching 36.1% in an ElevenLabs-cloned 'sister in distress' scenario. Economic modeling shows several AI voice models (Gemini, Sesame, ElevenLabs) are already profitable for attackers per-call, unlike human-operated vishing.

- **Published:** 2026-07-17T00:00:00Z
- **Last reviewed:** 2026-07-17T00:00:00Z
- **Canonical:** https://intel.threadlinqs.com/threat/TL-2026-1431
- **Card image:** https://intel.threadlinqs.com/og/threat/threat-intel-harvard-meta-ai-vishing-study-tl-2026-1431.webp?v=ed090f1c (Threadlinqs threat intel: Harvard-Meta Voice Phishing Study: Analysis & Hunting (TL-2026-1431) — Medium severity, 15 IOCs, 9 detection rules.)
- **ID:** TL-2026-1431
- **Severity:** MEDIUM
- **Category:** THREAT_INTEL
- **Status:** ACTIVE
- **Detections:** 9 (1 shown free below) · **IOCs:** 15

## Description

Researchers Fred Heiding, Claudio Mayrink Verdun, Simon Lermen, Andrew Kao, Vitor Albiero, Lauren Deason, Irina-Elena Veliche, and Christine Lehane (affiliations: Harvard Kennedy School, Harvard School of Engineering and Applied Sciences, Meta Platforms Inc., and independent researchers) published 'Evaluating AI Models' Capability to Automate Voice Phishing Attacks' (arXiv:2607.09970), accepted for publication in Expert Systems with Applications and covered by Help Net Security on 2026-07-17.

The study combined a large-scale survey experiment (N=4,100 US adults) with qualitative interviews (N=12). Participants were exposed to audio recordings or transcripts of five vishing scenarios generated by six leading voice-AI systems — Llama Full-Duplex (Meta, experimental mode, April 2025), OpenAI Advanced Voice Mode ('Sol' voice, GPT-4o), Google Gemini ('Ursa' voice, mobile app), Sesame ('Maya' voice, open-source conversational model), Play.AI ('Celeste' voice), and ElevenLabs (voice cloning) — alongside human-caller control recordings, with model refusals, disclaimers, and generation beeps edited out to simulate a competent attacker.

The five scam scenarios were: (1) MasterCard credential phishing impersonating card support (~7-15% compliance), (2) Gmail credential phishing impersonating Google support (~6-13% compliance), (3) a charitable Donation solicitation from an unknown caller (20-32% compliance), (4) a 'Police-Grandma' bail/relative-in-trouble scenario from an unknown caller (24-25% compliance), and (5) a 'Sister-in-Distress' scenario using an ElevenLabs-cloned voice claiming a family emergency requiring immediate funds, which produced the highest compliance at 36.1% (combined 'yes'/'maybe' responses; 6.5% hard 'yes'). Overall compliance across all five scenarios and both AI and human callers averaged 16.5%.

Detection performance varied sharply by model: participants correctly flagged OpenAI AVM as synthetic 93.7% of the time, Llama FD 88.6%, Gemini 88.4%, Play.AI 85.6%, ElevenLabs clones 75.8%, but Sesame only 66.3% — the hardest system to distinguish from a human. Overall voice-AI detection accuracy across all six systems was 70.3%, while real human callers were incorrectly flagged as synthetic between 54.2% and 67.4% of the time, indicating participants applied a strong 'guess AI' bias rather than genuine discrimination ability. Neither self-reported AI familiarity nor voice-assistant usage improved detection (51.2% for frequent AI users vs. 54.4% for non-users, not statistically significant).

Regression analysis identified message persuasiveness as the dominant predictor of victim compliance (odds ratio 2.58, p<.001), followed by caller sentiment (OR=1.64) and a 9-item trustworthiness scale covering expertise, credibility, empathy, urgency, and confidence (OR=1.48, Cronbach's alpha=.927). A separate 6-item Partner Modelling Questionnaire human-likeness scale (alpha=.93) did not independently predict compliance once persuasiveness and trustworthiness were controlled for — meaning voice realism itself was not the primary driver of victim harm; script quality and emotional manipulation were. Participants who cited 'repetitive responses' (OR=1.38, p=.048), 'long-winded responses' (OR=3.80, p<.001), or 'unnatural phrasing' in text transcripts (OR=3.39, p<.001) were more likely to correctly detect AI, but these cues went unnoticed by the majority of participants despite being present in the recordings.

The paper's economic model computed expected hourly attacker profit per voice system using a revenue formula r_j(t,X_i) = m(X_i) x p_j(t,X_i) x q, where m(X_i) is average extractable value per victim (~$450), p_j is per-call success probability, and q is a 0.6% call-to-contact conversion rate calibrated from marketing-industry data. Human operators paid US minimum-adjacent wages ($34.55/hr) were unprofitable at -$27.10/hr expected profit despite a 23% success rate, because human labor cost dominates. Among AI systems, Gemini (12% success rate, $0.13/hr compute cost) yielded +$2.38/hr expected profit, Sesame (15.2% success, $0.33/hr cost) yielded +$1.03/hr, and ElevenLabs (36.9% success rate — driven by the cloned-voice sister scenario — $0.75/hr cost) yielded the highest expected profit at +$2.97/hr; Llama FD, OpenAI AVM, and Play.AI remained unprofitable at current success/cost ratios. The authors estimate a one-time attacker development cost of $16,120 (260 engineering hours at $62/hr), with break-even periods of continuous operation ranging from 226 days (ElevenLabs) to 655 days (Sesame) to recoup that investment — figures the authors expect to shrink further as model costs decline and success rates improve.

The authors conclude that 'the primary risk of present-day AI-enabled vishing thus lies in the economics of automation rather than novel or superhuman persuasive techniques,' and that 'even a 5% success rate across millions of automated calls represents a transformative shift in the economics of fraud.' They recommend consumer education pivot from 'detect the synthetic voice' toward recognizing manipulative conversational strategies and resisting urgency/emotional-pressure tactics, plus out-of-band verification through independently-established channels, developer-side abuse-prevention and provenance tracking for voice-AI platforms, risk-based Know-Your-Customer controls for AI voice service signup, and regulatory frameworks that account for the 'automation dividend' of near-zero marginal-cost attacks.

This finding sits within a broader 2025-2026 vishing/imposter-scam surge documented by federal agencies: the FBI has issued repeated advisories on AI voice-cloning attacks impersonating senior US officials and family members in distress, and the FTC's June 2026 fraud report found Americans reported $3.5 billion in imposter-scam losses in 2025 (up roughly threefold since 2020), with imposter scams the top-reported fraud category for a fifth consecutive year and social media the fastest-growing initial contact vector ($2.1B in 2025 losses, an eightfold increase since 2020). A separate, since-withdrawn arXiv preprint (2602.20061, submitted Feb 2026, withdrawn Mar 2026 pending revision) reported even lower human discrimination ability (37.5% accuracy, below chance) on a smaller 22-participant, 16-clip sample, consistent directionally with the Harvard/Meta findings though not independently verified due to withdrawal.

## MITRE ATT&CK

- T1589 Gather Victim Identity Information
- T1593.001 Social Media
- T1587.001 Malware
- T1585 Establish Accounts
- T1583.006 Web Services
- T1566.004 Spearphishing Voice
- T1199 Trusted Relationship
- T1204.001 Malicious Link
- T1621 Multi-Factor Authentication Request Generation
- T1539 Steal Web Session Cookie
- T1684.001 Impersonation
- T1036 Masquerading
- T1123 Audio Capture
- T1213 Data from Information Repositories
- T1657 Financial Theft

## Free detection (Splunk SPL, Microsoft KQL, Sigma)

**AI Vishing Precursor — Bulk Identity/OSINT Harvest Feeding Voice-Clone Pretext Script** (`TL-2026-1431-SPL-001`)

Detects local reconnaissance behavior consistent with an attacker assembling a victim dossier (name, relative names, employer, phone) to script an AI-voice-clone vishing pretext, per the Harvard/Meta vishing study's five scenario archetypes. Flags PowerShell reading bulk PII-shaped files and writing a staged 'dossier' artifact shortly before a persistence-adjacent registry write.

- **Languages:** Splunk SPL, Microsoft KQL, Sigma
- **ATT&CK:** T1589, T1593.001, T1585
- **Data sources:** sysmon, windows_security, endpoint_edr
- **Severity:** medium
- **Confidence:** medium

### Splunk SPL

```spl
index=endpoint sourcetype="XmlWinEventLog:Microsoft-Windows-Sysmon/Operational" EventCode=1
| eval cmdline=lower(CommandLine)
| where match(Image, "(?i)powershell\.exe$") AND (like(cmdline, "%social_export%") OR like(cmdline, "%osint%") OR like(cmdline, "%dossier%") OR like(cmdline, "%profile_export%"))
| stats earliest(_time) as first_seen latest(_time) as last_seen count by ComputerName, User, CommandLine, ParentImage
| where count > 0
| eval risk="AI-vishing pretext staging (T1589/T1593.001)"
| table first_seen, last_seen, ComputerName, User, CommandLine, ParentImage, risk
```

### Microsoft KQL

```kql
DeviceProcessEvents
| where Timestamp > ago(24h)
| where FileName =~ "powershell.exe"
| where ProcessCommandLine has_any ("social_export", "osint", "dossier", "profile_export")
| project Timestamp, DeviceName, AccountName, FileName, ProcessCommandLine, InitiatingProcessFileName
| summarize FirstSeen=min(Timestamp), LastSeen=max(Timestamp), Events=count() by DeviceName, AccountName, ProcessCommandLine
```

### Sigma

```yaml
title: Bulk PII-Shaped File Access Preceding Registry Marker Write (AI-Vishing Pretext Staging)
id: 7c1e2f2a-4b6a-4b0a-9d3e-1a2b3c4d5e6f
status: experimental
description: Detects PowerShell reading/writing bulk identity-export-shaped files consistent with staging an AI-voice-clone vishing pretext dossier (Harvard/Meta AI vishing study, arXiv:2607.09970).
references:
  - https://arxiv.org/abs/2607.09970
  - https://www.helpnetsecurity.com/2026/07/17/research-ai-voice-phishing/
author: Threadlinqs Team
date: 2026-07-17
tags:
  - attack.reconnaissance
  - attack.t1589
  - attack.t1593.001
  - attack.resource_development
  - attack.t1585
logsource:
  product: windows
  category: process_creation
detection:
  selection:
    Image|endswith: '\powershell.exe'
    CommandLine|contains|all:
      - 'mock_social_export'
      - 'staged_pretext_dossier'
  selection_generic:
    Image|endswith: '\powershell.exe'
    CommandLine|contains:
      - 'social_export'
      - 'osint'
      - 'dossier'
      - 'profile_export'
  condition: selection or selection_generic
falsepositives:
  - Legitimate HR/recruiting or CRM export automation scripts using similar filenames
  - Authorized insider-risk or OSINT-defense tooling performing sanctioned lookalike activity
level: medium
```

**Known false positives:**

- Legitimate HR/recruiting export scripts using similarly named files
- Authorized red-team or insider-risk simulation tooling

**Tuning notes:** Tune the filename keyword list to your environment's naming conventions for exports/dossiers; correlate with the registry-marker detection (SIG-002) within a short window to raise confidence and suppress isolated single-hit alerts.

### 8 more detection rules (Splunk SPL, Microsoft KQL, Sigma), unlock with TL Creds

- Voice-AI/VoIP Service Account Establishment Following Identity-Lookup Burst (locked: unlock with TL Creds)
- MFA Push/Approval Bombing Immediately Following Voice-Support Impersonation Call (locked: unlock with TL Creds)
- New Cheap Cloud/SIP-Trunk Infrastructure Acquisition Followed by Rapid Outbound Calling Activation (locked: unlock with TL Creds)
- Spearphishing-Voice Pretext Rehearsal Log Pattern (High-Risk Persuasion Cue Combination) (locked: unlock with TL Creds)
- Session Cookie / Token Theft Following Vishing-Style Support Impersonation (locked: unlock with TL Creds)
- Impersonation-Consistent Caller-ID/Trusted-Relationship Spoofing Preceding Funds-Transfer or Credential Action (locked: unlock with TL Creds)
- Inbound-Call-Triggered Account-Sensitive Action Without Out-of-Band Verification (locked: unlock with TL Creds)
- Bursty Automated Outbound Call/Connection Pattern Consistent With Low-Marginal-Cost AI-Vishing Campaign (locked: unlock with TL Creds)

Sign up free to unlock the detection rules: https://threadlinqs.com/?modal=signup

## Sources

- [Research: AI voice phishing](https://www.helpnetsecurity.com/2026/07/17/research-ai-voice-phishing/)
- [Evaluating AI Models' Capability to Automate Voice Phishing Attacks (arXiv:2607.09970)](https://arxiv.org/abs/2607.09970)
- [Evaluating AI Models' Capability to Automate Voice Phishing Attacks (HTML full text)](https://arxiv.org/html/2607.09970v1)
- [Can You Tell It's AI? Human Perception of Synthetic Voices in Vishing Scenarios (withdrawn preprint, arXiv:2602.20061)](https://arxiv.org/abs/2602.20061)
- [FTC Data Show People Reported Losing $3.5 Billion to Imposter Scams in 2025](https://www.ftc.gov/news-events/news/press-releases/2026/06/ftc-data-show-people-reported-losing-3-point-5-billion-imposter-scams-2025)
- [New trends in reports of imposter scams](https://consumer.ftc.gov/consumer-alerts/2026/05/new-trends-reports-imposter-scams)
- [FBI warns of AI voice-cloning scam that mimics loved ones in distress](https://www.click2houston.com/news/local/2026/06/02/fbi-warns-of-ai-voice-cloning-scam-that-mimics-loved-ones-in-distress/)
- [Senior U.S. Officials Continue To Be Impersonated in Malicious Messaging Campaign](https://www.fbi.gov/investigate/cyber/alerts/2025/senior-us-officials-continue-to-be-impersonated-in-malicious-messaging-campaign)
- [FTC warns of record $3.5 billion losses to imposter scams in 2025](https://www.bleepingcomputer.com/news/security/ftc-warns-of-record-35-billion-losses-to-imposter-scams-in-2025/)

## Full data

The other 8 detection rules (Splunk SPL / Microsoft KQL / Sigma) unlock with TL Creds; a free account includes daily creds: https://threadlinqs.com/?modal=signup
Indicator values unlock with TL Creds, at a lower price on Red and Purple; AI agents can query rules and IOCs through the Threadlinqs MCP server (Purple): https://intel.threadlinqs.com/mcp

Canonical: https://intel.threadlinqs.com/threat/TL-2026-1431
