Synthetic Identity Fraud Statistics 2026
By Axis Intelligence Research
Co-authors: Sarah Davis & Marcus Chen | Last updated: August 6, 2026 | License: CC BY 4.0
Federal institutions reported $182 million in synthetic identity fraud through suspicious activity filings for 2021 — the only measured national figure that exists. Industry estimates for the same crime run past $35 billion. Axis Intelligence Research finds the gap between what is estimated and what is measured now exceeds 190 to 1.
Quick Answer
Synthetic identity fraud is the fastest-growing fraud type globally, and the least reliably counted. The only federally measured figure — $182 million in suspicious activity reported by financial institutions for 2021, across roughly 3,000 filings — sits alongside industry loss estimates that crossed $35 billion in 2023. According to Axis Intelligence Research, the Synthetic Identity Verification Gap (SIVG™) reads 55.3 at its August 2026 baseline, quantifying how far the federal detection and verification apparatus sits below the fraud’s estimated scale.
Key Findings
- According to Axis Intelligence Research, the ratio between industry synthetic identity loss estimates and federally measured suspicious activity value is 192 to 1, derived from FiVerity’s $35 billion 2023 estimate cited by the Federal Reserve Bank of Boston and FinCEN’s $182 million in 2021 synthetic identity suspicious activity reporting.
- According to Axis Intelligence Research, 57.5% of all US identity theft reports filed in 2024 involved account origination rather than misuse of an existing account — 652,842 of 1,135,291 reports, computed from Federal Trade Commission Consumer Sentinel subtype data.
- The Social Security Administration’s electronic SSN verification service now projects 68 million verifications per year against an original design projection of 1.1 billion — Axis Intelligence Research calculates the federal verification system operates at 6.2% of its planned volume.
- According to the Federal Reserve’s payments security division, synthetic identity fraud was the fastest-growing fraud type globally in 2025, rising eight-fold from 2024, driven by generative AI automation of identity creation and document forgery.
- According to Axis Intelligence Research, a small lender paying for federal SSN verification faces an effective cost of $0.51 per check while the largest users pay $0.029 — a 17.4× price advantage computed from the SSA tier schedule effective April 7, 2025.
What Is Synthetic Identity Fraud, and How Is It Counted?
The Federal Reserve convened a focus group of twelve fraud experts to settle a definitional problem that was wrecking industry reporting. Their recommended definition is deliberately narrow: synthetic identity fraud is the use of a combination of personally identifiable information to fabricate a person or entity in order to commit a dishonest act for personal or financial gain.
The distinction from ordinary identity theft matters for every number in this dataset. Traditional identity theft misuses a real person’s identity, and a real person notices. Synthetic identity fraud manufactures a person who does not exist, so nobody files a complaint. The Government Accountability Office states the consequence plainly in its 2024 review of federal verification infrastructure: synthetic identity losses are frequently booked as credit losses rather than fraud, which means the fraud statistics never see them.
That single accounting habit is why this page reads differently from most fraud datasets. There is no authoritative national loss total for synthetic identity fraud. There is a measured floor, a set of estimates, and a very large space in between.
The counting problem in three federal numbers
| Measure | Value | Period | Source |
|---|---|---|---|
| FinCEN suspicious activity value, synthetic identity | $182 million | 2021 | FinCEN, Identity-Related Suspicious Activity: 2021 Threats and Trends (Jan 2024), via GAO-24-106770 |
| Suspicious activity reports filed, synthetic identity | ~3,000 | 2021 | FinCEN, via GAO-24-106770 |
| Boston Fed measured loss estimate | $6 billion | 2016 | Federal Reserve Bank of Boston, Synthetic Identity Fraud in the U.S. Payment System (July 2019) |
| Study cited by Boston Fed | $20 billion | 2020 | Federal Reserve Bank of Boston, A Victimless Crime? Hardly (Aug 2022), via GAO-24-106770 |
| Anti-fraud platform estimate | $35 billion+ | 2023 | FiVerity, cited by Federal Reserve Bank of Boston (Apr 17, 2025) |
Axis Intelligence Research does not average these figures and will not publish a blended national loss total. The $182 million is a count of filings. The $35 billion is a modeled estimate from a commercial anti-fraud platform. Averaging a filing count with a vendor model produces a number that describes nothing. What the two figures jointly describe — and what we do publish — is the distance between them.
Sarah Davis, Digital Finance Lead: Follow the accounting and the whole thing makes sense. A charge-off is a line item a lender already knows how to book; a fraud loss triggers investigation, SAR filing, and questions from the examiner about origination controls. When the cheapest path for a portfolio manager is to call it credit deterioration, the number that reaches Washington is the number of institutions that chose the expensive path. Three thousand filings for a crime the same regulators call the fastest-growing in the country is not a measurement. It is a sample of who bothered.
How Much Does Synthetic Identity Fraud Cost in 2026?
There is no verified 2026 total, and any page publishing one is extrapolating. What can be sourced is the direction and the acceleration.
The Federal Reserve’s payments security team reported in June 2026 that synthetic identity fraud was the fastest-growing fraud type globally in 2025, with an eight-fold increase over 2024. The mechanism is documented rather than speculative: generative AI now automates identity assembly, produces supporting documents that reproduce government markings and security features, satisfies liveness and facial recognition checks during onboarding, and conducts patterned transactions across multiple institutions to mature accounts and grow credit lines.
Writing for the Federal Reserve Bank of Boston in April 2025, payments fraud specialist Mike Timoney described the same shallowness that makes synthetic identities detectable and the same generative tooling that is filling it in — fabricated parent records, aged email histories, deepfake audio and video attached to the fake profile.
For context on the enclosing category, the FBI’s Internet Crime Complaint Center logged 1,008,597 complaints and $20.877 billion in reported losses for 2025, a 26% year-over-year increase in losses and the first year past one million complaints. Synthetic identity fraud is not broken out as a category in that report, which is itself part of the measurement problem this page documents. Our online fraud statistics page breaks the IC3 and FTC loss data down by scheme type.
Synthetic Identity Fraud Statistics by Account Type
Because synthetic identities have no victim to complain, the closest sourced proxy is the share of identity theft reports that involve opening something rather than misusing something. Origination is where a fabricated identity has to pass; existing-account misuse requires a real account belonging to a real person.
New-account origination share of US identity theft reports, 2024
| Identity theft type | Reports | Origination-linked | Source |
|---|---|---|---|
| Credit card — new accounts | 406,110 | Yes | FTC Consumer Sentinel Network Data Book 2024 |
| Loan or lease (all subtypes) | 176,400 | Yes | FTC Consumer Sentinel Network Data Book 2024 |
| Government documents or benefits | 70,332 | Yes | FTC Consumer Sentinel Network Data Book 2024 |
| Credit card — existing accounts | 52,428 | No | FTC Consumer Sentinel Network Data Book 2024 |
| Bank account (all subtypes) | 114,608 | Mixed | FTC Consumer Sentinel Network Data Book 2024 |
| Employment or tax-related | 87,470 | Mixed | FTC Consumer Sentinel Network Data Book 2024 |
| Phone or utilities | 82,626 | Mixed | FTC Consumer Sentinel Network Data Book 2024 |
| Other identity theft | 358,993 | Mixed | FTC Consumer Sentinel Network Data Book 2024 |
| All identity theft reports | 1,135,291 | — | FTC Consumer Sentinel Network Data Book 2024 |
Axis calculation: (406,110 + 176,400 + 70,332) ÷ 1,135,291 = 57.5%. Consumers may report more than one type, and 14% of 2024 identity theft reports included multiple types, so this measures share of reports rather than share of unique victims.
The ratio inside the credit card category is the sharpest single number in the FTC’s 2024 data book: 406,110 reports of fraudulent new-account applications against 52,428 reports of misuse on cards that already existed. Roughly 7.7 fraudulent applications are reported for every fraudulent use of an existing card.
Loan or lease identity theft — the category where bust-out schemes surface — rose 17.8% year over year, from 149,803 reports in 2023 to 176,400 in 2024. Credit card identity theft rose 7.8% over the same period. The origination categories are growing faster than the misuse categories.
Sarah Davis: Underwriting teams read this table backwards. They see 449,032 credit card identity theft reports and staff the disputes queue. The number that should move headcount is 406,110 — applications, not transactions. A disputed charge is a chargeback; a funded application to a person who does not exist is an account that will behave perfectly for eighteen months and then vanish with the full line drawn. The first costs basis points. The second costs the principal.
Which States Report the Most Identity Theft?
State-level synthetic identity data does not exist. State-level identity theft reporting does, and it maps onto the same origination corridors — high-velocity credit markets, dense metro auto lending, and the states where new-account applications cluster.
| Rank | State | Reports per 100K | Total reports | Credit card share of ID theft types | Source |
|---|---|---|---|---|---|
| 1 | Florida | 528 | 115,840 | 50% | FTC Consumer Sentinel Network Data Book 2024 |
| 2 | Georgia | 517 | 55,955 | 47% | FTC Consumer Sentinel Network Data Book 2024 |
| 3 | Nevada | 466 | 14,631 | 46% | FTC Consumer Sentinel Network Data Book 2024 |
| 4 | Texas | 393 | 116,484 | 38% | FTC Consumer Sentinel Network Data Book 2024 |
| 5 | Delaware | 392 | 3,942 | 45% | FTC Consumer Sentinel Network Data Book 2024 |
| 6 | Massachusetts | 388 | 27,141 | 42% | FTC Consumer Sentinel Network Data Book 2024 |
| 7 | California | 356 | 139,665 | 49% | FTC Consumer Sentinel Network Data Book 2024 |
| 8 | Louisiana | 346 | 15,991 | 44% | FTC Consumer Sentinel Network Data Book 2024 |
| 9 | Illinois | 339 | 43,028 | 41% | FTC Consumer Sentinel Network Data Book 2024 |
| 10 | Maryland | 324 | 19,990 | 45% | FTC Consumer Sentinel Network Data Book 2024 |
| — | South Dakota (lowest) | 94 | 845 | 35% | FTC Consumer Sentinel Network Data Book 2024 |
Axis calculation: the median state reporting rate across all fifty states is 208.5 per 100,000. Florida’s 528 runs 2.53× the national state median. At the metro level the concentration is sharper still: Miami–Fort Lauderdale–West Palm Beach reported 903 identity theft reports per 100,000 residents, more than four times the median state rate.
Puerto Rico is the outlier worth naming. Employment or tax-related theft accounts for 32% of its identity theft types — against a national pattern where credit card dominates — which points to a materially different fraud economy than the mainland corridors.
The Federal Verification System: How eCBSV Actually Performs
This is the part of the synthetic identity story that almost nobody publishes, and it is the most consequential.
Congress anticipated synthetic identity fraud in 2018. Section 215 of the Economic Growth, Regulatory Relief, and Consumer Protection Act — the Banking Bill — directed the Social Security Administration to build a database letting financial institutions electronically verify that a name, date of birth, and SSN combination matches SSA records. That system, eCBSV, launched in June 2020. It is the single federal control designed specifically to break synthetic identities at origination, because a fabricated identity built on a real child’s SSN attached to a fabricated name will fail the match.
Six years in, the design assumptions and the operating reality have diverged by more than an order of magnitude.
eCBSV: projected versus actual
| Measure | Original projection | Current position | Source |
|---|---|---|---|
| Annual verifications | 1.1 billion (FY2021, projected Nov 2020) | 68 million/year projected | SSA Federal Register notices, via GAO-24-106770 and 90 FR 15030 |
| Direct users (permitted entities) | 123 | 21 (FY2025) | SSA Federal Register notices, via GAO-24-106770 and 90 FR 15030 |
| Actual verifications performed | — | 76.8 million (FY2023) | GAO-24-106770 |
| Total program cost | ~$27 million (initial estimate) | $66.3 million through FY2024 | GAO-24-106770; 90 FR 15030 |
| Costs unrecovered | — | $25.5 million | SSA, 90 FR 15030 (Apr 7, 2025) |
| Break-even target | End of FY2023 | End of FY2027 | GAO-24-106770; 90 FR 15030 |
Three findings from the GAO audit deserve to travel further than they have. Seventy-seven percent of the entities that applied for the eCBSV pilot had still not enrolled by FY2023. Of the 22 direct users renewing or starting subscriptions that year, 19 were service providers or credit reporting companies rather than banks — meaning 87% of the 76.8 million FY2023 transactions were submitted on behalf of roughly 946 indirect institutions. And SSA officials told GAO they did not plan to take significant steps to increase use of the service, because the agency’s stated goal is cost recovery rather than adoption.
What a federal SSN check costs
Under the tier schedule effective April 7, 2025, permitted entities buy an annual subscription band, not per-check pricing. Axis Intelligence Research computed the effective cost per verification at each band ceiling — the best case a subscriber in that tier can achieve.
| Tier | Annual volume | Annual fee | Effective cost per check at ceiling | Source |
|---|---|---|---|---|
| 1 | 1–10,000 | $5,100 | $0.510 | SSA, 90 FR 15030 |
| 2 | 10,001–75,000 | $37,125 | $0.495 | SSA, 90 FR 15030 |
| 3 | 75,001–200,000 | $98,000 | $0.490 | SSA, 90 FR 15030 |
| 4 | 200,001–500,000 | $240,000 | $0.480 | SSA, 90 FR 15030 |
| 5 | 500,001–1,000,000 | $470,000 | $0.470 | SSA, 90 FR 15030 |
| 6 | 1,000,001–2.5M | $907,500 | $0.363 | SSA, 90 FR 15030 |
| 7 | 2,500,001–5M | $1,765,500 | $0.353 | SSA, 90 FR 15030 |
| 8 | 5,000,001–10M | $3,206,250 | $0.321 | SSA, 90 FR 15030 |
| 9 | 10,000,001–15M | $3,562,500 | $0.238 | SSA, 90 FR 15030 |
| 10 | 15,000,001–20M | $4,453,125 | $0.223 | SSA, 90 FR 15030 |
| 11 | 20,000,001–25M | $5,165,625 | $0.207 | SSA, 90 FR 15030 |
| 12 | 25,000,001–200M | $5,878,125 | $0.029 | SSA, 90 FR 15030 |
Axis calculation: $0.510 ÷ $0.0294 = 17.4×. A community lender running 10,000 checks pays 17.4 times per verification what a national credit bureau pays at the top of Tier 12.
The band structure punishes anyone who lands just above a threshold. A subscriber at 10,001 transactions pays Tier 2’s $37,125, or $3.71 per check — more than seven times the Tier 1 ceiling rate. At 75,001 transactions the Tier 3 entry rate is $1.31; at 200,001, Tier 4 costs $1.20.
Sarah Davis: This is a pricing sheet doing policy work nobody voted for. Ten of the sixteen institutions GAO interviewed said the tier structure actively discourages volume, and several said they had narrowed eCBSV to high-risk applicants only in order to control spend. That is the exact opposite of how a fraud control should behave — you want the authoritative check on the whole funnel, because the synthetic that gets through is the one your risk model did not flag. When the authoritative source is priced as a scarce good, institutions ration it, and the rationing decision is made by whoever owns the vendor budget rather than whoever owns the loss.
To be fair to SSA: fees came down in April 2025, twice inside a single fiscal year, and the agency has now recovered all costs incurred through FY2020. The Tier 12 rate of under three cents per check is genuinely cheap. The problem is not the price at the top. It is that only 21 entities are in the building.
The Axis Synthetic Identity Verification Gap (SIVG™)
Every figure on this page describes the same structural condition: synthetic identity fraud is the most-estimated and least-measured fraud category in the US financial system. The Synthetic Identity Verification Gap (SIVG™) puts a single reproducible number on that condition.
What SIVG measures: the distance between synthetic identity fraud’s estimated scale and the measured capacity of the federal apparatus built to detect and verify against it. Higher readings mean a wider gap.
Formula:
SIVG = 0.50 × N(G1) + 0.30 × N(G2) + 0.20 × N(G3)
where N(G) = (log₁₀(G) ÷ 3) × 100 [log normalisation against a 1,000× ceiling]
G1 = Measurement Gap = industry loss estimate ÷ federally reported SAR value
G2 = Capacity Gap = original projected annual verifications ÷ current projected verifications
G3 = Participation Gap = originally projected direct users ÷ current participating entities
Inputs and baseline reading, as of August 6, 2026:
| Component | Inputs | Ratio | Normalised | Weight | Source |
|---|---|---|---|---|---|
| G1 Measurement Gap | $35,000M ÷ $182M | 192.3× | 76.1 | 50% | FiVerity via Boston Fed (2023 est.); FinCEN via GAO-24-106770 (2021) |
| G2 Capacity Gap | 1,100M ÷ 68M verifications | 16.2× | 40.3 | 30% | SSA projections via GAO-24-106770; 90 FR 15030 |
| G3 Participation Gap | 123 ÷ 21 entities | 5.9× | 25.6 | 20% | SSA projections via GAO-24-106770; 90 FR 15030 |
| SIVG baseline | — | — | 55.3 | 100% | Axis Intelligence Research |
This is the inaugural reading and therefore a baseline. It carries no history and makes no claim relative to prior periods.
Why the weights sit where they do. G1 carries half the index because measurement failure is the upstream cause: a fraud that is booked as credit loss cannot be resourced against. G2 carries 30% because verification capacity is the operational bottleneck that a policy change could actually move. G3 carries 20% because participation is downstream of both — institutions join a system they can afford and trust.
Why G1 combines an estimate with a measurement. Section 1.4 methodology normally prohibits merging incompatible sources. G1 is the deliberate exception, because the incompatibility is the measured quantity. A vendor model and a regulatory filing count answer the same question by different means; the ratio between them is the epistemic gap itself.
Scope and caveats. The G1 inputs carry different reference years — a 2023 estimate against 2021 filings — because those are the most recent figures each source publishes. FiVerity’s estimate is a commercial model, not a federal measurement, and is presented as such. G2 compares SSA’s own projections across two notices rather than actual-to-actual, since FY2025 actuals are not yet public; the FY2023 actual of 76.8 million verifications sits above the current 68 million projection, which is itself worth noting. SIVG measures institutional response capacity, not fraud volume.
Recomputation: every input above is public. A competent analyst can reproduce 55.3 from the six numbers in the table. SIVG is licensed CC BY 4.0 for citation and reuse with attribution to Axis Intelligence Research.
Why AI Changed Synthetic Identity Economics
Two things about generative AI matter for this dataset, and only two.
The first is unit cost. The Federal Reserve’s June 2026 analysis documents that generative tools capable of producing convincing identification documents, utility bills, and liveness-passing selfies are readily accessible and often free or low-cost. When the marginal cost of a plausible fake person approaches zero, the constraint on synthetic identity fraud shifts entirely to the institution’s ability to detect it — which is a defensive spend, not a criminal one.
The second is behavioural mimicry after onboarding. The Fed describes AI systems programmed to time transactions, select expected merchant categories, and hold amounts below alerting thresholds, then adjust as they learn which behaviours trigger review. Rule-based post-onboarding monitoring assumes a fraudster who does not iterate. That assumption no longer holds — a shift our cybersecurity statistics dataset tracks across attack categories.
This connects directly to the breach economy. Boston Fed notes the raw material is stolen PII from breaches — thousands of breaches compromising hundreds of millions of records annually. Our data breach statistics page tracks the supply side of that pipeline, and our phishing statistics page tracks the credential harvesting that feeds it.
Sarah Davis: The fee is always somewhere. In synthetic identity fraud the fee is a credit line that behaves normally for four to eighteen months while the account seasons — small balances, on-time payments, gradual limit increases granted by the lender’s own growth model. The lender pays the fraudster to build the exposure, then funds it. That is not a detection failure at onboarding. That is a detection failure that onboarding ratified, and every month of good behaviour after it made the eventual loss larger.
Who Gets Hurt When the Identity Is Fake?
The “victimless crime” framing collapses under three documented mechanisms, all named by the Federal Reserve Bank of Boston.
Children are the preferred SSN source precisely because nobody checks a minor’s credit file for fifteen years. The damage surfaces when the real person applies for a first card, a student loan, or an apartment lease. Seniors are targeted for the opposite reason — long, clean credit histories that give a synthetic profile immediate scoring lift. And businesses absorb the charge-offs, then price them into consumer terms.
The FTC’s 2024 identity theft data records age for 86% of the year’s 1,135,291 identity theft reports, and the two cohorts most useful to a synthetic identity builder — minors and the oldest consumers — sit at the bottom of that distribution. Low reporting from those groups is a floor, not a count, for the same reason the national loss total is unknowable: the population least likely to discover the fraud is the population most likely to be used for it.
Consumers who want to monitor exposure across all three bureaus can review our comparison of identity theft protection services; our identity theft statistics page covers per-victim loss data and the ACRI metric in more depth.
Methodology
Collection. All figures were retrieved from primary sources during production between August 5 and August 6, 2026. Federal sources: the Federal Trade Commission Consumer Sentinel Network Data Book 2024 (published March 2025, the most recent fully published annual edition); the Government Accountability Office report GAO-24-106770 (September 2024); the Social Security Administration Federal Register notice 90 FR 15030 (April 7, 2025); the SSA eCBSV programme documentation; Federal Reserve payments security publications (June 2026 and April 2025); and the FBI Internet Crime Complaint Center 2025 Annual Report (April 2026). FinCEN figures are cited as reported in GAO-24-106770.
What we measure. Reported figures only. FTC Consumer Sentinel data is based on unverified consumer reports, not a survey, and consumers may report multiple identity theft types — 14% of 2024 identity theft reports did. Because synthetic identities have no complaining victim, no dataset on this page is a direct count of synthetic identity fraud; the FTC origination share is presented explicitly as a proxy.
Axis calculations. Four figures on this page are computed by Axis Intelligence Research rather than reported by a source: the 57.5% origination share of identity theft reports; the 2.53× ratio of Florida’s reporting rate to the fifty-state median of 208.5 per 100,000; the effective cost-per-verification curve derived from the SSA tier schedule; and the SIVG index. Each formula and its inputs appear inline above, and each has a corresponding row in the downloadable dataset.
Non-primary sources. FiVerity’s $35 billion estimate is a commercial model reported by the Federal Reserve Bank of Boston; it is named in plain text and used only as the numerator of an explicitly labelled gap ratio. The LexisNexis-derived eight-fold growth figure reaches this page through the Federal Reserve’s own citation of it.
Scope and caveats. No national synthetic identity fraud loss total exists, and this page does not construct one. State-level and metro-level figures describe identity theft generally, not synthetic identity fraud specifically. The eCBSV FY2025 transaction actuals were not public at retrieval; comparisons use SSA’s published projections, dated.
Review trigger. This dataset will be revised when any of the following occurs: the FTC publishes the Consumer Sentinel Network Data Book 2025; SSA issues a new eCBSV tier fee notice in the Federal Register; FinCEN publishes updated identity-related suspicious activity analysis; or GAO issues a follow-up on eCBSV cost recovery. Scheduled review otherwise: Q4 2026.
About This Dataset
Title: Synthetic Identity Fraud Statistics 2026 — Federal Measurement, Verification Capacity, and Loss Estimates
Publisher: Axis Intelligence Research
Licence: CC BY 4.0
Format: CSV, UTF-8, one row per observation, full provenance columns
Temporal coverage: 2016–2026
Spatial coverage: United States
Citation
APA: Axis Intelligence Research. (2026). Synthetic identity fraud statistics 2026: Federal measurement, verification capacity, and loss estimates. https://axis-intelligence.com/synthetic-identity-fraud-statistics/
MLA: Axis Intelligence Research. “Synthetic Identity Fraud Statistics 2026: Federal Measurement, Verification Capacity, and Loss Estimates.” Axis Intelligence, 6 Aug. 2026, axis-intelligence.com/synthetic-identity-fraud-statistics/.
Chicago: Axis Intelligence Research. “Synthetic Identity Fraud Statistics 2026: Federal Measurement, Verification Capacity, and Loss Estimates.” Axis Intelligence. August 6, 2026. https://axis-intelligence.com/synthetic-identity-fraud-statistics/.
Frequently Asked Questions
Why do synthetic identity fraud loss estimates vary by more than 190 times?
Because one number counts filings and the other models exposure. FinCEN’s $182 million reflects what institutions actually reported as suspicious activity tied to synthetic identities in 2021. Commercial estimates in the $20–35 billion range model total system exposure including losses booked as credit charge-offs. GAO states directly that suspicious activity filings understate the true cost because many synthetic losses are never classified as fraud.
At what point in the account lifecycle does a synthetic identity actually cost money?
Not at origination. The Federal Reserve’s documented pattern is an account that opens cleanly, makes on-time payments, and accepts credit line increases for months before the bust-out. The loss is realised at the end and sized by everything the lender extended in between, which is why origination-stage controls are worth disproportionately more than post-booking monitoring.
Can eCBSV alone stop synthetic identity fraud?
No, and SSA says so explicitly. eCBSV returns a binary yes or no on whether a name, date of birth, and SSN combination matches SSA records; it does not verify that the applicant is the person behind that combination. Ten of the sixteen institutions GAO interviewed cited the binary result as a practical limit, since a “no” may indicate fraud or may indicate a typo, an alias, or a legal name change.
Why do so few banks use the federal SSN verification service directly?
Cost structure and access model. Nineteen of the 22 direct users that started or renewed in FY2023 were service providers or credit reporting companies, and 87% of transactions were submitted on behalf of roughly 946 indirect institutions. Most banks reach eCBSV through a vendor rather than subscribing, which means the tier fee is embedded in a vendor contract rather than a direct SSA subscription.
Is first-party fraud the same thing as synthetic identity fraud?
They overlap and are routinely confused at the case-classification stage. First-party fraud involves a real person misrepresenting intent or information; synthetic identity fraud involves a fabricated person. Hybrid schemes — a real applicant using a partially fabricated identity — get classified as first-party, closed, and never counted as synthetic, which is one documented route by which the measured total stays low.
Does a “no match” from eCBSV mean the applicant is a synthetic identity?
Not reliably. SSA’s response covers all three data elements jointly, and a mismatch on any one produces the same “no.” Applicants using a legal name that differs from SSA records, data-entry errors, and unreported name changes all produce false negatives. GAO recorded at least one institution limiting its use of the service specifically because false negatives risked declining legitimate customers.
Are children’s Social Security numbers really the preferred input?
They are a documented preferred input, per the Federal Reserve Bank of Boston, because a minor’s SSN carries no credit file and no monitoring, so a synthetic profile built on it can season for years undetected. Seniors’ SSNs are targeted for the inverse property — an established credit history that lends immediate scoring credibility to a fabricated profile.
What would move the SIVG reading down?
Three things, in order of impact: FinCEN or a bank regulator publishing a synthetic-identity loss series that reflects charge-off reclassification; SSA closing the gap between eCBSV’s design capacity and its actual volume; and permitted-entity participation rising above the 21 institutions enrolled for FY2025. A change in fraud volume alone would not move SIVG, because SIVG measures institutional response, not the crime.
Which US states report the highest identity theft rates?
Florida leads at 528 reports per 100,000 residents, followed by Georgia (517), Nevada (466), Texas (393), and Delaware (392), per the FTC’s 2024 data. Axis Intelligence Research calculates Florida’s rate at 2.53 times the fifty-state median of 208.5. These are identity theft rates generally; no state-level synthetic identity fraud data is published by any federal source.
How does AI change detection, not just attack?
Symmetrically, in principle. The Federal Reserve documents institutions deploying AI to detect pixel-level artefacts in submitted images, metadata signatures indicating machine-generated documents, and audio-video inconsistencies characteristic of deepfakes. The Boston Fed’s framing is that synthetic identities are structurally shallow — no long-lived email, no years of utility history — and that depth-checking is exactly the pattern-matching task models do well.
Charts
- The measurement gap — bar comparison, log scale: FinCEN measured SAR value ($182M, 2021) vs Boston Fed 2016 estimate ($6B), 2020 cited study ($20B), FiVerity 2023 estimate ($35B). Source: FinCEN via GAO-24-106770; Federal Reserve Bank of Boston.
- Origination vs misuse — stacked bar: credit card new accounts (406,110) vs existing accounts (52,428); loan or lease (176,400); government documents or benefits (70,332). Source: FTC Consumer Sentinel Network Data Book 2024.
- eCBSV projected vs actual — grouped bars: 1.1B projected verifications vs 68M current projection vs 76.8M FY2023 actual; 123 projected users vs 21 participating. Source: GAO-24-106770; 90 FR 15030.
- The federal verification price curve — line chart, effective cost per check at each of the twelve tier ceilings, $0.510 to $0.029. Source: SSA, 90 FR 15030; Axis Intelligence Research calculation.
- Identity theft reports per 100,000 by state — choropleth, all fifty states plus DC and Puerto Rico. Source: FTC Consumer Sentinel Network Data Book 2024.
- SIVG component contributions — horizontal bar: G1 (76.1 × 0.50), G2 (40.3 × 0.30), G3 (25.6 × 0.20) summing to 55.3. Source: Axis Intelligence Research.