Deepfake fraud attempts against contact centers surged more than 1,300% in 2024, climbing from roughly one per month to seven per day (Pindrop, 2025 Voice Intelligence and Security Report). That same year, U.S. contact centers were hit by a fraud attempt every 46 seconds, and the report projects USD 44.5 billion in fraud exposure for 2025. Meanwhile the voice biometrics market that authentication vendors are racing to sell reached USD 2.63 billion in 2025 and is forecast at USD 3.06 billion in 2026 (Mordor Intelligence, 2026). The technology built to verify a human voice and the technology built to fake one are now scaling against each other in real time. This analysis consolidates data from Pindrop, Deloitte, Gartner, the FBI Internet Crime Complaint Center, Mordor Intelligence, and 16 other primary sources published in 2024-2026.
TL;DR
- Deepfake fraud attempts against contact centers rose +1,300% in 2024, from about one a month to seven a day (Pindrop, 2025 Voice Intelligence and Security Report).
- The voice biometrics market reached USD 2.63B in 2025 and is projected at USD 3.06B in 2026, growing 16.41% a year to USD 6.54B by 2031 (Mordor Intelligence, 2026).
- Deloitte projects U.S. generative-AI-enabled fraud losses hitting USD 40B by 2027, up from USD 12.3B in 2023 (Deloitte Center for Financial Services, 2024).
- 62% of organizations experienced a deepfake attack in the past 12 months; 43% involved a deepfaked audio call (Gartner, 2025).
- Deepfakes now account for 1 in 5 biometric fraud attempts (Entrust, 2026 Identity Fraud Report).
- Native virtual-camera injection attacks used to bypass biometric checks rose +2,665% in 2025 (iProov, 2025 Threat Intelligence Report).
- Americans lost USD 3.5B to imposter scams in 2025, with complaints up about 19% to roughly 1 million (FTC, 2025).
- Top audio-deepfake detectors hit 98.1% accuracy in a 2026 benchmark, but audio models lost ~50% of AUC on in-the-wild 2024 attacks (Resemble AI, 2026; arXiv, 2025).
- HSBC Voice ID blocked GBP 249M of attempted phone fraud in a year and cut telephone-banking fraud 50% (HSBC, 2021).
- A single deepfake video call cost engineering firm Arup USD 25M across 15 transfers (CNN, 2024).
- Voice phishing surged +442% between the first and second halves of 2024 (CrowdStrike, 2025 Global Threat Report).
- 69% of consumers now hold at least one passkey, part of the shift away from knowledge-based logins (FIDO Alliance, 2025).
1. Market Size and Growth
Five research firms agree on direction and disagree on magnitude. Their 2026 valuations cluster near USD 3-5 billion, but their mid-2030s forecasts span USD 6.5 billion to USD 22.8 billion. That roughly 3.5x spread tells buyers the category is real while the sizing methodology is still immature, so any single CAGR should be read as a range. The consistent signal underneath the noise: banking pays for most of it, and the contact-center IVR is where it runs. BFSI accounted for 31.40% of voice biometrics revenue in 2025, and IVR and contact centers made up 45.30% of deployments - voice biometrics is, first and foremost, a bank phone-line product. Cross-referenced figures from Mordor Intelligence and Fortune Business Insights are below.
| Metric | Value | Source |
|---|---|---|
| Voice biometrics market, 2026 to 2031 | USD 3.06B to USD 6.54B (16.41% CAGR, from USD 2.63B in 2025) | Mordor Intelligence, 2026 |
| Voice biometrics market, 2034 | USD 22.76B (25.88% CAGR from USD 3.61B in 2026) | Fortune Business Insights, 2026 |
| Voice biometrics market, 2033 | USD 18.93B (21.5% CAGR from USD 3.99B in 2025) | SkyQuest, 2026 |
| Voice biometrics market, 2032 | USD 15.27B (20.71% CAGR from USD 4.08B in 2025) | Research and Markets, 2026 |
| Voice biometrics market, 2034 | USD 11.78B (17.12% CAGR from USD 2.84B in 2025) | Renub Research, 2026 |
| BFSI share / IVR deployment share, 2025 | 31.40% / 45.30% | Mordor Intelligence, 2026 |
Outlier: forecasts diverge nearly 3.5x by the mid-2030s (USD 6.5B vs USD 22.8B), a reminder that voice-biometrics TAM estimates are directional, not precise.
2. The Deepfake Threat to Voice Authentication
Voice authentication rests on one assumption: the voice on the line belongs to a live person. Synthetic speech breaks that assumption at industrial scale, and the growth curve is the story. 42.5% of analyzed fraud attempts are now AI-driven, and nearly one in three of those succeed - a hit rate that turns voice-only verification from a control into a liability. The attack is also cheap to launch: cloning a usable voice can take as little as three seconds of recorded audio (FTC, 2024), which is why voice phishing volume is compounding across every measurement. Figures below come from Pindrop’s 2025 Voice Intelligence and Security Report. For the broader synthetic-media picture, see our deepfake statistics 2026 roundup.
| Metric | Value | Source |
|---|---|---|
| Deepfake fraud attempt surge, 2024 | +1,300% (1/month to 7/day) | Pindrop, 2025 VISR |
| Synthetic voice attacks, insurance, 2024 | +475% | Pindrop, 2025 VISR |
| Fraud attempts that are AI-driven | 42.5% | Pindrop, 2025 VISR |
| Voice phishing surge, H1 to H2 2024 | +442% | CrowdStrike, 2025 Global Threat Report |
| Voice phishing as initial infection vector, 2025 | #2, 11% of investigations | Mandiant, M-Trends 2026 |
Outlier: CrowdStrike found H1 2025 voice-phishing volume had already exceeded the full-year 2024 total (CrowdStrike, 2025).
3. Fraud Losses and Economic Impact
The dollar figures separate hype from balance-sheet risk. Deloitte projects U.S. generative-AI-enabled fraud losses reaching USD 40 billion by 2027, up from USD 12.3 billion in 2023 - a 32% compound annual growth rate. These losses are not theoretical: reported imposter-scam losses have tripled since 2020, and a single deepfake video call moved USD 25 million out of engineering firm Arup. Regulators are now measuring the AI layer directly, with authorities breaking out AI-related complaints for the first time in the 2025 report. Primary data below is drawn from Deloitte and the FBI Internet Crime Complaint Center; see also our voice cloning fraud statistics 2027.
| Metric | Value | Source |
|---|---|---|
| Gen-AI fraud losses, U.S., projected 2027 | USD 40B (from USD 12.3B in 2023, 32% CAGR) | Deloitte, 2024 |
| Imposter-scam losses, U.S., 2025 | USD 3.5B | FTC, 2025 |
| Imposter-scam complaints, 2025 | ~1M (+19%) | FTC, 2025 |
| Total internet-crime losses, 2025 | USD 20.877B (+26%) | FBI IC3, 2025 |
| AI-related complaints / losses, 2025 | 22,364 / USD 893M | FBI IC3, 2025 |
| Elder-fraud losses, 2025 | USD 7.748B (+59% vs 2024) | FBI IC3, 2025 |
| Arup deepfake CFO video-call loss, 2024 | USD 25M (15 transfers, 5 accounts) | CNN, 2024 |
Outlier: Deloitte’s conservative scenario still puts 2027 gen-AI fraud losses near USD 22 billion (Deloitte, 2024).
4. Enterprise Exposure and the Biometric Attack Surface
Deepfakes moved out of consumer scams and into the enterprise authentication stack. 62% of organizations experienced a deepfake attack in the prior 12 months, and 43% were hit by a deepfaked audio call specifically (Gartner, 2025). The uncomfortable corollary for security teams: when one in five biometric fraud attempts already carries a deepfake, “add biometrics” stops being a complete answer, and injection attacks that feed synthetic media straight into a verification pipeline are climbing faster than any single modality can defend. Data below comes from Gartner and iProov; broader context is in our cybersecurity statistics 2026.
| Metric | Value | Source |
|---|---|---|
| Orgs hit by a deepfake attack, past 12 months | 62% | Gartner, 2025 |
| Orgs hit by a deepfaked audio call | 43% | Gartner, 2025 |
| Orgs hit by a deepfaked video call | 37% | Gartner, 2025 |
| Biometric fraud attempts using deepfakes | 1 in 5 (20%) | Entrust, 2026 Identity Fraud Report |
| Native virtual-camera injection attacks, 2025 | +2,665% | iProov, 2025 Threat Intelligence Report |
| Face-swap attempts, year on year, 2025 | +300% | iProov, 2025 |
| Consumers who spotted every fake and real clip | 0.1% | iProov, 2025 |
Outlier: Gartner’s figures come from 302 cybersecurity leaders surveyed across North America, EMEA, and Asia/Pacific between March and May 2025 (Gartner, 2025).
5. Detection Accuracy and the Generalization Gap
Vendors quote near-perfect lab accuracy; the field number is the one that matters. The top audio-deepfake detector scored 98.1% accuracy in a 2026 benchmark, yet the same class of models lost roughly half their AUC when tested against in-the-wild 2024 attacks. The reason is structural: a detector trained on 2019 attack data does not generalize to 2026 voice synthesis, and models tuned against one text-to-speech architecture can fail against another. Reading detection scores means asking which dataset produced them. Benchmark data below is from Resemble AI and a peer-reviewed study in Springer’s Circuits, Systems, and Signal Processing.
| Metric | Value | Source |
|---|---|---|
| Top audio-deepfake detector accuracy, 2026 benchmark | 98.1% (F1 0.981) | Resemble AI, 2026 |
| Second-ranked detector accuracy | 96.8% | Resemble AI, 2026 |
| Third-ranked detector accuracy | 83.5% | Resemble AI, 2026 |
| In-the-wild deep-learning detector accuracy | 99.84% | Springer (Circuits, Systems, and Signal Processing), 2025 |
| Cross-dataset accuracy (ASVspoof 5, WaveFake) | >98% (EER 0.0013) | Springer, 2025 |
| Audio-model AUC drop on Deepfake-Eval-2024 | ~50% | arXiv (Deepfake-Eval-2024), 2025 |
| Fortune 500 insurer deepfake detection rate | 97% | Pindrop case study, 2025 |
Outlier: in the same 2026 benchmark, four open-source detectors trained on outdated 2019 data scored between 48% and 71.3% accuracy - a gulf that shows why lab numbers rarely survive contact with new attacks (Resemble AI, 2026).
6. Adoption and Deployment in the Field
Voice authentication is not a pilot; it is embedded in banks and processing billions of logins. HSBC’s Voice ID prevented GBP 249 million of attempted telephone fraud in a single year and cut telephone-banking fraud 50% year on year (HSBC, 2021). The wider authentication market is pivoting toward device-bound credentials at the same time, with 69% of consumers now holding at least one passkey. That shift matters for privacy: on-device, consent-based voice processing - where a person’s own voice model never leaves their machine - is the counterweight to centralized voiceprint databases that become fraud targets the moment they are breached. Deployment data below is from HSBC and the FIDO Alliance. For the consent-based, self-cloning angle, VoxBooster’s voice cloning software processes a user’s own voice locally on Windows.
| Metric | Value | Source |
|---|---|---|
| HSBC Voice ID fraud prevented, year to May 2021 | GBP 249M | HSBC, 2021 |
| HSBC telephone-banking fraud change, year on year | -50% | HSBC, 2021 |
| HSBC customers enrolled in Voice ID per week | ~14,000 | HSBC, 2021 |
| Nuance voice authentications performed annually | 5B+ (300M consumers) | Nuance, 2018 |
| Voiceprints enrolled via Nuance | 600M+ (500+ organizations) | Nuance, 2021 |
| Consumers holding at least one passkey | 69% | FIDO Alliance, 2025 |
| Top-100 websites supporting passkeys | 48% | FIDO Alliance, 2025 |
Most recent available: HSBC and Nuance operational totals date to 2018-2021; both stopped publishing updated enrollment figures afterward, so treat them as floors, not current counts.
Summary: Voice Biometrics And Voice Authentication by the Numbers
| Metric | Value | Source |
|---|---|---|
| Voice biometrics market, 2026 to 2031 | USD 3.06B to USD 6.54B (16.41% CAGR) | Mordor Intelligence, 2026 |
| Voice biometrics market, 2034 (high estimate) | USD 22.76B (25.88% CAGR) | Fortune Business Insights, 2026 |
| Deepfake fraud attempt surge, 2024 | +1,300% (1/month to 7/day) | Pindrop, 2025 VISR |
| Voice phishing surge, H1 to H2 2024 | +442% | CrowdStrike, 2025 |
| Voice phishing as initial infection vector, 2025 | #2 (11% of cases) | Mandiant, M-Trends 2026 |
| Gen-AI fraud losses, U.S., projected 2027 | USD 40B (from USD 12.3B in 2023) | Deloitte, 2024 |
| Imposter-scam losses, U.S., 2025 | USD 3.5B | FTC, 2025 |
| Total internet-crime losses, 2025 | USD 20.877B (+26%) | FBI IC3, 2025 |
| AI-related complaint losses, 2025 | USD 893M (22,364 complaints) | FBI IC3, 2025 |
| Elder-fraud losses, 2025 | USD 7.748B (+59%) | FBI IC3, 2025 |
| Arup deepfake video-call loss, 2024 | USD 25M | CNN, 2024 |
| Orgs hit by a deepfake attack, past 12 months | 62% | Gartner, 2025 |
| Biometric fraud attempts using deepfakes | 1 in 5 | Entrust, 2026 |
| Virtual-camera injection attacks, 2025 | +2,665% | iProov, 2025 |
| Top audio-deepfake detector accuracy, 2026 | 98.1% | Resemble AI, 2026 |
| Audio-model AUC drop on in-the-wild 2024 attacks | ~50% | arXiv, 2025 |
| HSBC Voice ID fraud prevented, one year | GBP 249M | HSBC, 2021 |
| Nuance voice authentications performed annually | 5B+ | Nuance, 2018 |
| Consumers holding at least one passkey | 69% | FIDO Alliance, 2025 |
Methodology and Sources
Data was gathered from primary reports, vendor threat-intelligence publications, government crime statistics, peer-reviewed research, and named market-research firms; every market-size and growth figure was cross-referenced across two or more firms, and figures older than 2024 are flagged as most recent available. Every statistic in this article was verified against a source viewed during research. Sources cited:
- Pindrop - 2025 Voice Intelligence and Security Report; 2023 Deepfake and Voice Clone Consumer Report; Fortune 500 insurer case study (report)
- Mordor Intelligence - Voice Biometrics Market, 2026 (report)
- Fortune Business Insights - Voice Biometrics Market, 2026 (report)
- SkyQuest - Voice Biometrics Market, 2026
- Research and Markets - Voice Biometrics Market Global Forecast, 2026
- Renub Research - Voice Biometrics Market, 2026
- Deloitte Center for Financial Services - Deepfake Banking Fraud Risk on the Rise, 2024 (report)
- U.S. Federal Trade Commission - Imposter scam / consumer fraud data, 2024-2025
- FBI Internet Crime Complaint Center - 2025 Internet Crime Report (report)
- Gartner - GenAI/deepfake attack survey, September 2025 (press release)
- Entrust - 2026 Identity Fraud Report (report)
- iProov - 2025 Threat Intelligence Report (press release)
- Resemble AI - Audio Deepfake Detection Benchmark, 2026 (benchmark)
- Springer, Circuits, Systems, and Signal Processing - Hybrid Deep Learning Framework for Real and Deepfake Voice Detection, 2025 (paper)
- arXiv - On Deepfake Voice Detection / Deepfake-Eval-2024, 2025 (paper)
- HSBC UK - Voice ID fraud-prevention data, 2021 (release)
- Nuance Communications - voice biometrics milestones, 2018-2021 (milestone)
- FIDO Alliance - World Passkey Day / passkey adoption, 2025 (release)
- CrowdStrike - 2025 Global Threat Report
- Mandiant - M-Trends 2026
- CNN - Arup deepfake fraud reporting, 2024 (article)
Data watch: Pindrop’s Voice Intelligence and Security Report, CrowdStrike’s Global Threat Report, Mandiant’s M-Trends, and the FBI Internet Crime Complaint Center’s annual report are all recurring publications typically released in Q1; the next editions, covering 2026 data, are expected in early-to-mid 2027 and will be folded in here. Gartner’s deepfake-attack survey and the iProov and Entrust threat/fraud reports also run on annual cycles and are due for refreshed editions around the same window.
Last updated: July 5, 2026. We review and update this page quarterly as new data is published.