The FBI counted 22,364 internet crime complaints with an artificial intelligence component in 2025, carrying $893 million in reported losses, and adults aged 60 and over absorbed about $352 million of that total (FBI IC3, 2025 Internet Crime Report). That is the first year the bureau published a dedicated AI section, and it names voice cloning in ‘family in distress’ calls as one of the patterns behind the number. Consumer surveys point in the same direction: 1 in 4 adults in seven countries had experienced an AI voice scam or knew someone who had, and 77% of victims lost money (McAfee, Beware the Artificial Impostor 2023), while 28% of UK adults say a voice cloning scam targeted them in the past year (Starling Bank, 2024). This roundup narrows the lens of our broader voice AI scam statistics to one scheme: the fake call from a child, grandchild or partner in trouble. We aggregated data from the FBI Internet Crime Complaint Center, the Federal Trade Commission, the FCC, McAfee, Starling Bank, Hiya, Pindrop, Consumer Reports, University College London, AARP, and the Canadian and Australian anti-scam agencies.
TL;DR
- 22,364 AI-linked complaints and $893 million in losses reported to the FBI in 2025 (FBI IC3, 2025 Internet Crime Report).
- Distress or grandparent scams using voice cloning cost victims more than $5 million in 2025 (FBI IC3, 2025 Internet Crime Report).
- Adults 60+ filed 14.05% of AI-related complaints but carried 39.46% of AI-related losses (FBI IC3, 2025).
- 1 in 4 adults experienced an AI voice scam or knew someone who did; 77% of victims lost money (McAfee, 2023).
- 28% of UK adults were targeted by a voice cloning scam in the past year, and 46% had never heard of the scam (Starling Bank, 2024).
- 31% of US consumers encountered a deepfake voice fraud call in the past year (Hiya, Q4 2024 Global Call Threat Report).
- Three seconds of audio produced an 85% voice match (McAfee, 2023).
- Listeners spotted deepfake speech only 73% of the time (UCL, PLOS ONE 2023).
- 4 of 6 voice cloning products tested lacked meaningful consent safeguards (Consumer Reports, March 2025).
- US consumers reported $3.5 billion lost to imposter scams in 2025, nearly three times the 2020 level (FTC, June 2026).
- Scam reports identified as AI-enabled rose twentyfold between 2023 and 2025 (AARP with Microsoft AI for Good Lab).
- Only 13% of parents have created a family codeword (McAfee, State of the Scamiverse 2026).
1. How Many People Are Targeted
Survey data converges on a striking range: somewhere between a quarter and a third of adults in rich countries now report brushing against a cloned-voice scam. 1 in 4 adults across seven countries had experienced an AI voice scam or knew someone who had, with 1 in 10 targeted personally (McAfee survey of 7,054 adults, April 2023). The more recent UK and multi-country studies show the scheme has moved from novelty to background noise.
The problem is awareness lagging exposure. In the UK, 28% of adults say they were targeted by a voice cloning scam in the past year, yet 46% had never heard of this type of scam (Starling Bank and Mortar Research, 3,010 UK adults, August 2024). A scam people have never heard of is a scam they have not rehearsed a response to.
| Metric | Value | Source |
|---|---|---|
| Adults who experienced an AI voice scam or know someone who did | 1 in 4 | McAfee, Beware the Artificial Impostor 2023 |
| Adults targeted personally | 1 in 10 | McAfee, 2023 |
| Adults who say it happened to someone they know | 15% | McAfee, 2023 |
| UK adults targeted by a voice cloning scam in the past year | 28% | Starling Bank / Mortar Research, 2024 |
| UK adults who had never heard of voice cloning scams | 46% | Starling Bank / Mortar Research, 2024 |
| US / Canada / UK consumers who met a deepfake voice fraud call in the past year | 31% / 27% / 26% | Hiya, Q4 2024 Global Call Threat Report (Jan 2025 survey) |
| Respondents across six countries who met a deepfake voice fraud call | More than one-third | Hiya, Q4 2024 Global Call Threat Report |
| Americans’ average deepfakes seen per day | 3 | McAfee, State of the Scamiverse 2026 |
Context note: Hiya’s figures come from a survey of 12,000 consumers in the US, Canada, UK, Germany, France and Spain fielded in January 2025 (Hiya press release). Self-reported ‘deepfake call’ exposure depends on what respondents believe they heard, so treat these as perception data rather than measured incidence.
2. Money Lost: What Federal Data Shows
The AI line in the FBI report is small relative to total cybercrime, but it is the first time the bureau has measured it at all. $893,346,472 in losses across 22,364 AI-linked complaints sits inside a record $20.877 billion in total reported losses for 2025 (FBI IC3, 2025 Internet Crime Report). The FBI is explicit that the AI figure is a floor: AI tagging depends on victims recognizing the technology, and a convincing cloned voice is, by design, hard to recognize.
Family emergency scams are filed by the FTC under imposter scams, the most reported fraud category for the fifth year running. Imposter scams cost US consumers $3.5 billion in 2025, nearly three times the 2020 figure, and nearly one in three fraud reports involved an imposter (FTC, June 2026). Total reported fraud losses reached about $16 billion, the highest on record.
| Metric | Value | Source |
|---|---|---|
| AI-linked internet crime complaints, 2025 | 22,364 | FBI IC3, 2025 Internet Crime Report |
| AI-linked reported losses, 2025 | $893 million | FBI IC3, 2025 Internet Crime Report |
| Distress / grandparent scam losses using voice cloning, 2025 | More than $5 million | FBI IC3, 2025 Internet Crime Report |
| All internet crime complaints / losses, 2025 | 1,008,597 / $20.877 billion (+26%) | FBI IC3, 2025 Internet Crime Report |
| US imposter scam losses, 2025 | $3.5 billion | FTC Consumer Sentinel, June 2026 |
| Imposter scam reports, 2025 | More than 1 million | FTC testimony to the Joint Economic Committee, March 2026 |
| Total US fraud losses reported to the FTC, 2025 | $15.9 billion (about $16 billion) | FTC, March and June 2026 |
| Imposter scam losses, 2024 | $2.95 billion | FTC Consumer Sentinel, March 2025 |
The distress-scam line (more than $5 million) looks small next to investment fraud, but it counts only cases a victim reported to the IC3 and linked to that scheme. Older FTC data shows how fast the family category can move: losses to family and friend imposters jumped from $26 million to $41 million in a single year in 2018 (FTC Data Spotlight, December 2018; most recent available spotlight dedicated to this scam).
3. Older Adults and the Grandparent Scam
Age is where the loss data turns sharp. Adults 60 and over filed 14.05% of AI-related complaints to the FBI in 2025 but accounted for 39.46% of AI-related losses, roughly $352 million (FBI IC3, 2025 Internet Crime Report). Their total internet crime losses reached $7.75 billion, up 59% in one year, with an average loss of $38,501 per complainant. Our elder fraud statistics cover the wider picture; the voice-clone slice matters because it attacks the one thing older adults are told to trust, a familiar voice.
The FTC’s own trend lines show the same escalation in large losses. Reports from older adults losing $10,000 or more to impersonation scams rose from 1,790 in 2020 to 8,269 in 2024, more than fourfold, and 41% of those big-loss cases started with a phone call (FTC Data Spotlight, August 2025). Total fraud losses reported by people 60+ grew from about $600 million in 2020 to $2.4 billion in 2024 (FTC, December 2025).
| Metric | Value | Source |
|---|---|---|
| AI-related losses, adults 60+ | About $352 million | FBI IC3, 2025 Internet Crime Report |
| Share of AI complaints / AI losses from adults 60+ | 14.05% / 39.46% | FBI IC3, 2025 Internet Crime Report |
| Internet crime complaints / losses, adults 60+ | 201,266 / $7.75 billion (+59%) | FBI IC3, 2025 Internet Crime Report |
| Average loss per complainant 60+ | $38,501 | FBI IC3, 2025 Internet Crime Report |
| Complainants 60+ losing more than $100,000 | 12,444 | FBI IC3, 2025 Internet Crime Report |
| Older adult reports of $10,000+ impersonation losses | 1,790 (2020) to 8,269 (2024) | FTC Data Spotlight, Aug 2025 |
| Fraud losses reported by adults 60+ | About $600 million (2020) to $2.4 billion (2024) | FTC report to Congress, Dec 2025 |
| Estimated total losses of older adults including unreported cases, 2024 | As high as $81.5 billion | FTC report to Congress, Dec 2025 (via Fox Business) |
Organized call centers sit behind many of these calls. In March 2025, 25 Canadian nationals were charged in Vermont over a grandparent scheme that allegedly took more than $21 million from elderly victims in over 40 US states between 2021 and June 2024, with callers posing as grandchildren arrested after car crashes and needing bail (ICE / HSI press release). The prosecutors did not allege voice cloning; the case shows the call-center infrastructure that cloned voices now plug into.
| Metric | Value | Source |
|---|---|---|
| Defendants charged in the Vermont grandparent scam case | 25 (23 arrested on March 4, 2025) | ICE / HSI, March 2025 |
| Alleged losses in the Vermont case | More than $21 million, 40+ states | ICE / HSI, March 2025 |
| Canadian losses to emergency (grandparent) scams since 2021 | More than $23 million | Canadian Anti-Fraud Centre, April 2024 |
| Money mules arrested in Canada since 2021 | More than 70 | Canadian Anti-Fraud Centre, April 2024 |
| Australian ‘Hi Mum’ family impersonation victims, Jan-Jul 2022 | 1,150+ victims, $2.6 million lost | ACCC Scamwatch, 2022 |
| Share of ‘Hi Mum’ losses from people over 55 | 95% | ACCC Scamwatch, 2022 |
Context note: the Canadian figure comes from the Canadian Anti-Fraud Centre and the Australian figures from Scamwatch; the Australian data dates from 2022 and is the most recent available breakdown for that text-based variant.
4. Anatomy of a Fake Emergency Call
The script has barely changed in a decade; only the voice has. Messages claiming the sender had been in a car incident (48%) or been robbed (47%) were the most likely to get a response, followed by a lost phone or wallet (43%) and needing help while traveling abroad (41%) (McAfee, 2023). In older FTC data, the caller claimed to be in jail or other legal trouble in about half of cash-payment reports, and posed as a grandchild, usually a grandson, about 70% of the time (FTC Data Spotlight, 2018).
Who the voice belongs to shapes who answers. 45% of adults said they would reply to a voicemail or voice note from a friend or loved one in need of money, most often if it seemed to come from a spouse (40%), a parent (31%) or a child (20%), and parents aged 50 and over were the group most likely to respond to a child, at 41% (McAfee Beware the Artificial Impostor report). Scammers clone the voice that disarms the listener most.
| Metric | Value | Source |
|---|---|---|
| Adults who would reply to a distress voice message from a loved one | 45% | McAfee, 2023 |
| Most persuasive scenarios: car incident / robbery / lost phone or wallet / abroad | 48% / 47% / 43% / 41% | McAfee, 2023 |
| Response likelihood by sender: spouse / parent / child | 40% / 31% / 20% | McAfee, 2023 |
| Parents 50+ most likely to respond to a child | 41% | McAfee, 2023 |
| Grandparent-scam reports where the caller posed as a grandchild | About 70% | FTC Data Spotlight, 2018 |
| Cash-payment reports involving a jail or legal-trouble story | About half | FTC Data Spotlight, 2018 |
| UK adults who would send money even if the call seemed suspicious | 8% | Starling Bank / Mortar Research, 2024 |
Payment rails tell the rest of the story. Among older adults who lost $10,000 or more to an impersonation scam in 2024, 33% paid in cryptocurrency, 20% by bank transfer and 16% in cash (FTC Data Spotlight, August 2025). Back in 2018, 25% of people 70 and over who reported a family imposter payment sent cash, and their median loss was $9,000 versus about $2,000 across all ages (FTC, 2018). Cash and crypto share one feature: once handed over, neither can be clawed back.
| Metric | Value | Source |
|---|---|---|
| Victims who lost money after an AI voice scam | 77% | McAfee, 2023 |
| Victims who lost more than $1,000 | More than one-third | McAfee, 2023 |
| Victims who lost $5,000 to $15,000 | 7% | McAfee, 2023 |
| Payment method, older adults losing $10,000+: crypto / bank transfer / cash | 33% / 20% / 16% | FTC Data Spotlight, Aug 2025 |
| Median loss for people 70+ who mailed cash to a family imposter | $9,000 | FTC Data Spotlight, 2018 |
| US average loss per phone fraud victim | $539 | Hiya, Q4 2024 Global Call Threat Report |
5. Why Ears Fail: Cloning Speed and Detection Limits
The economics of the scam collapsed when cloning got short and cheap. Three seconds of audio was enough to produce an 85% voice match, and further training pushed accuracy to 95%, using one of more than a dozen freely available tools McAfee researchers found online (McAfee, 2023). Meanwhile 53% of adults shared their voice online at least once a week, so source audio is rarely the bottleneck.
Human hearing does not close the gap. In a controlled study of 529 listeners, people identified deepfake speech only 73% of the time, and training raised accuracy by just 3.84 percentage points on average (Mai et al., PLOS ONE 2023). Accuracy rose to 85.59% only when listeners could compare against a reference clip, a luxury no one has during a panicked phone call. For lab and machine benchmarks, see our audio deepfake detection accuracy statistics.
| Metric | Value | Source |
|---|---|---|
| Audio needed for an 85% voice match | 3 seconds | McAfee, 2023 |
| Voice match accuracy after further training | 95% | McAfee, 2023 |
| Adults who share their voice online at least weekly | 53% | McAfee, 2023 |
| Adults not confident they could identify a cloned voice | 70% | McAfee, 2023 |
| Human accuracy detecting deepfake speech (no reference) | 73% | UCL, PLOS ONE 2023 |
| Human accuracy with a genuine reference clip | 85.59% | UCL, PLOS ONE 2023 |
| UK adults confident they could spot the warning signs | 30% | Starling Bank / Mortar Research, 2024 |
| Americans not confident they could identify a deepfake scam | 35% | McAfee, State of the Scamiverse 2026 |
Call-network data confirms synthetic voices are already in the phone system at scale. In Hiya’s Q1 2025 honeypot of 8,455 analyzable calls, 25% contained AI-generated audio and 55% of those AI calls were identified as fraud (Hiya, 2025). Pindrop, analyzing 1.2 billion customer calls, measured deepfake attempts rising more than 1,300% in 2024, from about one a month to seven a day (Pindrop 2025 Voice Intelligence and Security Report). An analysis of 531,000 fraud reports by Microsoft’s AI for Good Lab with AARP found scams identified as AI-enabled, with cloned voices or synthetic video, increased twentyfold between 2023 and 2025 (AARP).
| Metric | Value | Source |
|---|---|---|
| Honeypot calls containing AI-generated audio | 25% | Hiya, Q1 2025 |
| AI-generated calls identified as fraud | 55% | Hiya, Q1 2025 |
| Deepfake fraud attempts growth, 2024 | +1,300% (1 per month to 7 per day) | Pindrop, 2025 Voice Intelligence and Security Report |
| Projected growth in deepfaked calls, 2025 | +155% | Pindrop, 2025 |
| Growth in AI-enabled scam reports, 2023 to 2025 | Twentyfold (531,000 reports analyzed) | AARP and Microsoft AI for Good Lab |
| Voice cloning products lacking meaningful consent safeguards | 4 of 6 tested | Consumer Reports, March 2025 |
Context note: Pindrop’s data covers enterprise contact centers (banks, insurers, retailers), not family calls. It is included as a measure of how quickly synthetic voice attacks scale once tooling is cheap.
6. Defenses, Rules and the Safe-Word Gap
Regulators have moved, but slowly relative to the tooling. On February 8, 2024 the FCC unanimously ruled that AI-generated voices count as ‘artificial’ under the Telephone Consumer Protection Act, so robocalls using them need prior express consent, after 26 state attorneys general urged the change (FCC news release). Consumer Reports found in March 2025 that 4 of 6 voice cloning products it assessed had no meaningful mechanism to confirm the speaker’s consent, and 4 of 6 required only a name or email to sign up (Consumer Reports). The consent debate behind legitimate voice cloning software is covered in our AI voice cloning ethics statistics.
The cheapest defense remains low-tech. The FBI advises families to create a secret word or phrase to verify identity, limit public audio and video of themselves, and listen for unusual tone and word choice (FBI IC3 PSA, December 2024). Adoption trails concern by a wide margin: 67% of people worry about voice cloning, but only 19% have disabled voicemail recordings, and 81% of parents fear their child’s likeness will be stolen while only 13% have created a family codeword (McAfee, State of the Scamiverse 2026).
| Metric | Value | Source |
|---|---|---|
| FCC vote classifying AI voices as ‘artificial’ under the TCPA | Unanimous, Feb 8, 2024 | FCC, 2024 |
| State AGs urging the FCC ruling / AGs in FCC robocall MOU | 26 / 48 | FCC, 2024 |
| FTC Voice Cloning Challenge prize pool (3 winning teams) | $35,000 | FTC, April 2024 |
| People worried about voice cloning | 67% | McAfee, State of the Scamiverse 2026 |
| People who have disabled voicemail recordings | 19% | McAfee, State of the Scamiverse 2026 |
| Parents who fear a child’s likeness will be stolen / have a family codeword | 81% / 13% | McAfee, State of the Scamiverse 2026 |
| UK adults worried about becoming a voice cloning victim | 79% | Starling Bank / Mortar Research, 2024 |
The FTC’s Voice Cloning Challenge funded detection, watermarking and voice-authentication prototypes in April 2024, and in September 2026 the agency opened public comment on whether its impersonation rule should address platforms’ role in promoting impersonation scams. Neither replaces the simple habit of hanging up and calling the person back on a number you already know.
Summary: Voice Cloning Family Emergency Scams by the Numbers
| Metric | Value | Source |
|---|---|---|
| AI-linked complaints / losses reported to the FBI, 2025 | 22,364 / $893 million | FBI IC3, 2025 Internet Crime Report |
| Distress / grandparent scam losses using voice cloning, 2025 | More than $5 million | FBI IC3, 2025 Internet Crime Report |
| AI-related losses from adults 60+ | About $352 million (39.46% of AI losses) | FBI IC3, 2025 Internet Crime Report |
| Internet crime losses, adults 60+ | $7.75 billion (+59%) | FBI IC3, 2025 Internet Crime Report |
| US imposter scam losses, 2025 | $3.5 billion | FTC, June 2026 |
| Total US fraud losses reported to the FTC, 2025 | $15.9 billion | FTC, March 2026 |
| Older adult reports of $10,000+ impersonation losses | 1,790 (2020) to 8,269 (2024) | FTC Data Spotlight, Aug 2025 |
| Big-loss impersonation cases starting with a phone call | 41% | FTC Data Spotlight, Aug 2025 |
| Adults who experienced an AI voice scam or know someone who did | 1 in 4 | McAfee, 2023 |
| Voice scam victims who lost money | 77% | McAfee, 2023 |
| UK adults targeted by voice cloning in the past year | 28% | Starling Bank, 2024 |
| US consumers who met a deepfake voice fraud call | 31% | Hiya, Q4 2024 |
| Audio needed for an 85% voice match | 3 seconds | McAfee, 2023 |
| Human accuracy detecting deepfake speech | 73% | UCL, PLOS ONE 2023 |
| Honeypot calls containing AI-generated audio | 25% | Hiya, Q1 2025 |
| Growth in AI-enabled scam reports, 2023-2025 | Twentyfold | AARP and Microsoft AI for Good Lab |
| Voice cloning products lacking consent safeguards | 4 of 6 | Consumer Reports, 2025 |
| Parents with a family codeword | 13% | McAfee, 2026 |
| Alleged losses in the Vermont grandparent scam case | More than $21 million | ICE / HSI, 2025 |
Methodology and Sources
Every figure above was traced to the organization that produced it. Where a primary document could not be read directly (the FBI IC3 PDF and some McAfee pages did not render or timed out during research), we used figures that matched across multiple reports quoting that document and named the original source. Widely shared numbers we could not trace to a primary document, including several ‘AI fraud grew X%’ claims attributed to the FBI, were excluded.
- FBI Internet Crime Complaint Center: 2025 Internet Crime Report and PSA on criminal use of generative AI (December 2024)
- Federal Trade Commission: imposter scam data for 2025 (June 2026), 2024 fraud data (March 2025), testimony to the Joint Economic Committee (March 2026), Data Spotlight on older adults (August 2025), annual report to Congress on older adults (December 2025), grandparent scam Data Spotlight (2018) and Voice Cloning Challenge winners (April 2024)
- Federal Communications Commission: declaratory ruling on AI-generated voices in robocalls (February 2024)
- McAfee: Beware the Artificial Impostor report and press release (May 2023), State of the Scamiverse 2026
- Starling Bank and Mortar Research: Safe Phrases campaign survey (September 2024)
- Hiya: Q4 2024 Global Call Threat Report (February 2025) and AI-generated audio honeypot analysis (2025)
- Pindrop: 2025 Voice Intelligence and Security Report
- University College London: Warning: Humans cannot reliably detect speech deepfakes, PLOS ONE (August 2023)
- Consumer Reports: assessment of AI voice cloning products (March 2025)
- AARP: AI makes it next to impossible to detect scams, reporting the Microsoft AI for Good Lab analysis
- U.S. Immigration and Customs Enforcement: 25 Canadian nationals charged in grandparent scam (March 2025)
- Canadian Anti-Fraud Centre: emergency scam investigation update (April 2024)
- ACCC Scamwatch: ‘Hi Mum’ scam warning (2022)
- Data watch: No agency publishes a clean count of voice-cloned family emergency calls. The FBI’s AI figures rely on victims recognizing AI involvement and are described by the bureau as a floor; the more than $5 million distress-scam figure covers only reports tied to that scheme. The FTC’s 2025 total appears as $15.9 billion in March 2026 testimony and ‘about $16 billion’ in the June 2026 release; we cite both as the same figure. The $81.5 billion older-adult estimate is a modeled figure that adjusts for underreporting, read through Fox Business coverage of the FTC report. McAfee’s 2023 survey, the UCL study (2023) and the FTC grandparent-scam spotlight (2018) and ACCC data (2022) are older than three years or close to it and are shown as most recent available data. McAfee State of the Scamiverse 2026 figures were read from McAfee’s published summary because the full report did not load. Hiya and Starling exposure figures are self-reported. Pindrop measures enterprise contact centers, not households. The FTC’s 2025 Consumer Sentinel Data Book, with a family and friend impersonation breakdown, is the next primary dataset to watch.
Last updated: October 3, 2026. We update this roundup quarterly, and the next refresh is expected when the FTC publishes its next Consumer Sentinel Network Data Book and the FBI releases its 2026 Elder Fraud and Internet Crime reports.