AI Scam Trends 2027
By Axis Intelligence Research
Co-author: Marcus Chen | Last updated: October 31, 2026 | License: CC BY 4.0
The most quoted AI fraud forecast in circulation — Deloitte’s $40 billion in US fraud losses by 2027 — was built on the FBI’s own complaint data. So it can be graded against that data. Two years in, the FBI’s crime-type losses sit 8.5% below the path Deloitte’s endpoints imply, and hitting $40 billion now requires 40.4% growth a year. Last year’s growth was 26.2%.
Quick Answer
Axis Intelligence Research projects about $27.8 billion in scam and fraud losses reported to the FBI’s Internet Crime Complaint Center (IC3) in 2027 in its base case (range $21.6–35.2 billion), up from $20.3 billion in 2025. That sits 31% below Deloitte’s $40 billion forecast. The driver to watch: how many new investment-scam victims file complaints.
Key Findings
- Axis Intelligence Research projects $27.8 billion in IC3 crime-type losses for 2027 in its base case, up 37% from $20.3 billion in 2025.
- Deloitte’s 2023 starting point of $12.3 billion matches the sum of the FBI’s 2023 crime-type loss table ($12.35 billion), an Axis Intelligence Research finding that makes Deloitte’s $40 billion forecast directly testable against IC3 data.
- Complaints flagged as AI-related carried $893.3 million in 2025 losses, only 5.6% of the $15.9 billion lost in the six crime types the FBI says AI enhances, Axis Intelligence Research calculates.
- Americans aged 60 and over filed 20.0% of IC3 complaints in 2025 but carried 37.1% of reported losses, an average of $38,500 per complaint, per FBI data.
- Investment scam complaints rose 52.3% in 2025 while the average loss fell 13.6% to $118,500 — more victims at smaller tickets, the signature of scam operations scaling with automation, Axis Intelligence Research finds.
Here’s what actually happened in 2025. The FBI’s IC3 logged 1,008,597 complaints and $20.877 billion in losses, up 26% in a year, according to the FBI’s 2025 IC3 Annual Report. No new exploit class drove it. The same three schemes that led in 2023 — investment fraud, business email compromise and tech-support impersonation — led again, just with more victims on the other end of the chat window.
The AI question is whether 2026 and 2027 bend that curve upward. The honest answer from the ledger: AI is visible in the method, barely visible in the money, and nearly impossible to separate from the scams it now sits inside.
How Much Will Americans Lose to Scams in 2027?
About $27.8 billion reported to the FBI in the Axis base case, measured as the sum of IC3’s crime-type loss table. On IC3’s headline basis, which also counts losses not assigned to a crime type, that scales to roughly $28.6 billion — an Axis Intelligence Research calculation. The 2026 base case is $23.6 billion, which the FBI’s 2026 report will confirm or break in spring 2027.
Reported is the operative word. These are losses victims filed with the FBI, not the cost of fraud. The Federal Trade Commission estimates that, adjusted for underreporting, fraud cost consumers as much as $195.9 billion in 2024, per its March 25, 2026 testimony to the Joint Economic Committee. The forecast here tracks the measurable floor, because the floor is the only number anyone will publish a check against.
The Axis Base Case formula
Most scam forecasts take one growth rate and apply it to a total. This one splits IC3 losses into five segments and each segment into two drivers that the FBI publishes every year:
Losses 2027 = Σ segments [ Complaints 2025 × (1 + complaint growth)² × Average loss 2025 × (1 + average-loss growth)² ]
- Complaint growth — how many new victims file. Base case: half of each segment’s 2023–2025 compound annual rate. Enforcement against scam compounds, the FBI’s victim-notification program and the slowdown in Australia (below) argue that volume growth cools; nothing in the data says it stops.
- Average-loss growth — how much each victim loses. All scenarios hold each segment’s 2023–2025 compound rate.
- All other categories (22 smaller crime types, including personal data breach and non-delivery) are modeled on losses alone: their 2025 growth of 12.7% in the base case.
| Segment (IC3 crime types) | 2025 complaints | 2025 losses | Avg. loss 2025 | Complaint growth / yr (base) | Avg.-loss growth / yr | 2027 base case |
|---|---|---|---|---|---|---|
| Investment | 72,984 | $8.65B | $118,500 | 17.9% | 1.3% | $12.34B |
| Call-center impersonation (tech support + government) | 80,218 | $2.93B | $36,558 | 12.3% | 19.8% | $5.30B |
| Business email compromise | 24,768 | $3.05B | $123,005 | 3.7% | −5.3% | $2.94B |
| Confidence/romance | 23,159 | $0.93B | $40,126 | 7.0% | 4.7% | $1.17B |
| All other (22 types) | 567,323 | $4.74B | $8,352 | — | — (losses +12.7%) | $6.02B |
| Total | $20.30B | $27.76B |
Source: Axis Intelligence Research calculation from FBI IC3 2023–2025 crime-type tables. Inputs and every intermediate value are in the dataset.
Investment fraud carries 44.4% of the 2027 base case. That is the whole forecast in one number: whatever happens to crypto investment scams happens to US scam losses.
Base, upside and downside scenarios
| Scenario | 2026 | 2027 | What separates it | What would confirm it |
|---|---|---|---|---|
| Downside | $20.9B | $21.6B | Complaint counts flat in every segment; “all other” flat | IC3 2026 complaints at or below 1,008,597 |
| Base | $23.6B | $27.8B | Complaint growth at half its 2023–25 pace; severity trend held | IC3 2026 crime-type losses between $22B and $25B |
| Upside | $26.6B | $35.2B | Complaint growth at its full 2023–25 pace | Investment complaints up 30%+ again in 2026 |
Source: Axis Intelligence Research projection, October 31, 2026. All values are IC3 crime-type loss sums.
Even the upside lands below Deloitte’s $40 billion. Getting there needs the 2023–2025 complaint surge to repeat at full speed and something extra on top.
Which assumption matters most?
The investment segment, on both drivers. Moving investment complaint growth 10 points a year either way shifts the 2027 total by +$2.2 billion or −$2.0 billion. Moving investment average-loss growth 10 points shifts it by +$2.6 billion or −$2.3 billion. The same 10-point move on call-center complaints shifts it by about $1 billion. Read that as a monitoring instruction: watch the investment line in the FBI’s 2026 report before anything else.
Source: Axis Intelligence Research, AI Scam Trends 2027 (CC BY 4.0)
Is the $40 Billion AI Fraud Forecast Still Realistic?
Not on the current track. Deloitte’s Center for Financial Services forecast in May 2024 that generative AI could enable US fraud losses of $40 billion by 2027, from $12.3 billion in 2023, a 32% compound annual growth rate, per Deloitte Insights. Deloitte says it built the forecast by scoring each of the 26 fraud types in the FBI’s IC3 report for generative-AI risk.
That method is what makes the forecast gradable. The 2023 IC3 crime-type table sums to $12.35 billion — the same base Deloitte used. So the honest comparison is IC3’s crime-type sum against the path Deloitte’s endpoints imply:
| Year | Deloitte path (interpolated by Axis) | IC3 crime-type loss sum (actual) | Gap |
|---|---|---|---|
| 2023 | $12.3B (Deloitte base) | $12.35B | — |
| 2024 | $16.5B | $16.09B | −2.6% |
| 2025 | $22.2B | $20.30B | −8.5% |
| 2026 | $29.8B | Axis base: $23.6B | — |
| 2027 | $40.0B (Deloitte forecast) | Axis base: $27.8B | −30.6% |
Source: Deloitte (May 29, 2024) endpoints; Axis Intelligence Research interpolation at the endpoint-implied 34.3% annual rate (Deloitte rounds its stated rate to 32%); FBI IC3 crime-type tables. Interpolated years are Axis arithmetic, not Deloitte figures.
The gap opened in 2025 and the arithmetic now closes the door: reaching $40 billion from $20.3 billion takes 40.4% growth in each of 2026 and 2027. IC3 crime-type losses grew 26.2% in 2025.
One caveat cuts in Deloitte’s favor. Its forecast measures fraud “enabled” by generative AI across all 26 types, and IC3 reports only what victims file. If underreporting worsens — victims who never learn they were scammed don’t complain — the reported series would understate a real acceleration. That is a hypothesis, though, not a measurement. The forecast that went viral was a number; the number is behind.
AI fraud forecast reconciliation
| Firm | Published | Scope | Base value | Horizon value |
|---|---|---|---|---|
| Deloitte Center for Financial Services | May 29, 2024 | US fraud losses enabled by gen AI, 26 IC3 fraud types | $12.3B (2023) | $40B (2027), base |
| Deloitte Center for Financial Services | May 29, 2024 | US gen-AI email fraud (business email compromise) | — | about $11.5B (2027), aggressive scenario |
| Federal Trade Commission | Mar 25, 2026 | US consumer fraud cost adjusted for underreporting | — | up to $195.9B (2024 estimate, not a forecast) |
| Axis Intelligence Research | Oct 31, 2026 | IC3 crime-type losses, all fraud | $20.3B (2025) | $27.8B (2027), base |
Other “AI scam” forecasts circulating in search results could not be traced to a producer that published its method; they were excluded rather than repeated.
How Much of Scam Loss Is Actually AI?
On the FBI’s books, $893.3 million across 22,364 complaints in 2025 — 4.4% of crime-type losses. That number is a floor with a known leak.
IC3 applies an “AI Related” descriptor when a complaint mentions artificial intelligence. The FBI’s report names six crime types where AI enhances the scheme: investment, business email compromise, romance and distress scams, tech support, government impersonation and employment. Those six carried $15.9 billion in 2025 — 78.4% of crime-type losses. The AI flag covers 5.6% of that money, an Axis Intelligence Research calculation.
The FBI says why, in its own words: investment complaints with an AI nexus passed $632 million, but overall investment losses exceeded $8 billion, which shows many victims do not realize how far AI was involved. The flag captures 7.3% of investment losses. Victims report the script they saw, not the tooling that wrote it.
| AI-related losses by crime type, 2025 | Amount | Source |
|---|---|---|
| Investment | $632.0M | FBI IC3 |
| Business email compromise | $30.3M | FBI IC3 |
| Tech/customer support | $19.5M | FBI IC3 |
| Confidence/romance | $19.0M | FBI IC3 |
| Employment | $12.6M | FBI IC3 |
| Government impersonation | $7.1M | FBI IC3 |
| Distress (“grandparent”) voice-clone scams, within romance | over $5M | FBI IC3 |
Source: FBI IC3 2025 Annual Report, AI section.
Two consequences for anyone citing AI scam growth. First, there is no FBI growth rate for AI scams: IC3’s three-year comparison tables mark the AI descriptor as not captured before 2025, so any “AI scams up X%” attributed to the FBI is arithmetic on a series that does not exist. Second, the AI-flagged complaint is expensive: an average of $39,946, 1.93 times the all-complaint average of $20,699, an Axis Intelligence Research calculation.
Blockchain data shows the same effect from the money side. Chainalysis estimates that scam operations with on-chain links to AI vendors — sellers of face-swap tools, deepfake software and language models — took in $3.2 million per operation in 2025, against $719,000 for operations without such links, 4.5 times more, per its 2026 Crypto Crime Report scams chapter. Chainalysis sells fraud-prevention tools; its on-chain series is still the field’s reference.
Trend 1: Will Investment Scams Keep Driving Losses in 2027?
Yes. Investment fraud was $8.65 billion of IC3 losses in 2025, the largest category for the third straight year in the FBI’s three-year table, and $7.228 billion of it was cryptocurrency investment fraud.
The shape of 2025 matters more than the size. Complaints rose 52.3% to 72,984. The average loss fell 13.6% to $118,500. That is what industrial scam operations look like when they scale: more conversations, run in parallel, landing smaller first deposits. The FBI describes the same mechanics — AI used to generate thousands of conversations that read differently to each target, and deepfaked celebrities and executives selling “investment clubs.”
Other countries’ banks see the same line move. UK Finance recorded £221.5 million in investment scam losses in 2025, up 40%, the largest category of authorised push payment (APP) fraud, per its Annual Fraud Report 2026. APP fraud is a scam in which the victim authorizes the transfer themselves. In Australia, investment scams were the top category at A$837.7 million, per the National Anti-Scam Centre.
The front door is the feed. Investment scams that started on social media cost Americans $1.1 billion in 2025 — more than half of all social-media scam losses — per the FTC. Deepfaked endorsements need distribution, and paid social reach is the cheapest distribution there is; our digital marketing trends outlook covers where that ad inventory is heading.
What would prove it wrong: investment complaints flat or falling in the FBI’s 2026 report. Who it affects: retail investors approached on social platforms and messaging apps; exchanges and banks that receive the first deposit. Track the payment rails on our stablecoin statistics.
Trend 2: Are Impersonation Scams Getting Worse?
Yes, and faster than investment fraud in percentage terms. The FBI groups tech-support and government impersonation as call-center fraud. Together they drew 80,218 complaints in 2025, up 50.3%, and $2.93 billion in losses, up 56.8%. Government impersonation alone went from $405.6 million to $797.9 million.
The Federal Trade Commission’s series shows the same direction. Imposter scams were nearly one in three fraud reports in 2025, with $3.5 billion in losses. Government-impersonation losses reached about $920 million, up from $789 million in 2024, and business impersonation came to nearly $1 billion, up from $866 million, per the FTC’s June 15, 2026 release.
The phone channel turned back up. Do Not Call complaints about robocalls rose to 1,601,611 in fiscal 2025 from 1,099,229 in fiscal 2024, up 45.7%, after falling every year since fiscal 2021, per the FTC’s FY 2025 Do Not Call Data Book. Imposters were the second-largest complaint topic, with 243,628 complaints. Automated calls are where synthetic voices scale cheapest; the reversal is the leading indicator to watch, even if the Data Book cannot say how many of those calls used AI.
Chainalysis measured impersonation scams growing more than 1,400% on-chain in 2025, and documents the infrastructure behind it: Chinese-language phishing-kit vendors selling toll-text and postal templates, some kits for under $500.
This is the category where cheap automation pays best. The script is short — your toll is overdue, your account is compromised, a federal agent needs to verify your funds — and the delivery is SMS, robocall or a spoofed number at near-zero marginal cost. In the base case it is the fastest-growing segment, up from $2.93 billion to $5.30 billion by 2027.
What would prove it wrong: call-center complaints falling in 2026 after the FBI and India’s Central Bureau of Investigation’s joint raids (about 175 arrests through 13 operations in 2025, per IC3). Who it affects: older adults first (see below), then anyone with an account at a bank, exchange or toll agency worth impersonating. Read the entry-point view on our phishing statistics.
Trend 3: How Big Are Voice-Cloning and Deepfake Scams Really?
Smaller in reported dollars than the headlines suggest, and growing inside other categories rather than as their own. IC3 attributes over $5 million in 2025 to distress scams — the voice-cloned family emergency — and over $30 million to business email compromise involving AI. AI-involved employment scams, including deepfaked job interviews, reported almost $13 million.
The FBI’s own description of the employment pattern is the most practical detection tell on record: lip movement that doesn’t match the audio, coughs and sneezes that don’t line up with the video. The goal in those interviews, IC3 notes, is usually network access rather than money — which is why the dollar loss stays small while the risk to employers doesn’t. Hiring-side exposure is covered in our future of work trends.
The Axis reading: the deepfake is a closer, not a category. A cloned voice or a fake CEO video shows up as the moment that converts an investment, BEC or romance scam — and gets filed under that crime type. Expect the FBI’s separately reported voice-clone line to stay in the millions.
Vendor telemetry tells a louder story, and should be read as what it is. Pindrop, which sells deepfake detection, reports that AI attacks across its own customer base rose 1,210% from Q4 2024 to Q4 2025, and 1,680% by the end of June 2026, per its Deepfake Readiness Index report (September 2026). In the same report, a Wakefield Research survey of 250 US security leaders (June 11–22, 2026, margin of error ±6.2 points) found 74% reported deepfake attacks, 10% used purpose-built detection tools, and about 25% of affected organizations put the total cost at $1 million or more. Those are company-reported and survey figures: they measure attempts seen by one vendor and costs estimated by respondents, not losses filed with law enforcement. The Axis reading: the 1,210% and the FBI’s $5 million describe different things, and both are true. Attempt volume is exploding; reported dollar losses from pure voice-clone schemes are not — yet.
What would prove it wrong: IC3’s 2026 report showing distress-scam losses above $50 million, a tenfold jump. Who it affects: finance teams approving payments on voice or video instruction; families with public audio of a relative online.
Who Loses the Most to Scams by Age?
Adults 60 and over, by a wide margin, in the FBI data. Younger adults report scams often and lose less per complaint.
| Age group | IC3 complaints, 2025 | Reported losses | Avg. loss per complaint | Share of all IC3 losses |
|---|---|---|---|---|
| Under 20 | 31,254 | $67.1M | $2,147 | 0.3% |
| 20–29 | 112,069 | $563.1M | $5,025 | 2.7% |
| 30–39 | 153,293 | $1.7B | $11,090 | 8.1% |
| 40–49 | 167,066 | $2.957B | $17,700 | 14.2% |
| 50–59 | 124,820 | $3.7B | $29,643 | 17.7% |
| 60+ | 201,266 | $7.748B | $38,500 | 37.1% |
Source: FBI IC3 2025 Annual Report; averages and shares are Axis Intelligence Research calculations. 218,829 complaints gave no age.
Losses reported by people 60 and over rose 59% in 2025 while their complaints rose 37%, and 12,444 of them lost more than $100,000. The average loss climbs with every decade of age. That gradient — not scam frequency — is why elder fraud dominates the dollar totals.
Three rails concentrate the damage. Cryptocurrency ATMs drew $389 million in losses in 2025, up 58%, and 66.2% of that money came from people 60 and over. Romance and confidence scams cost the 60+ group $584.0 million, 62.8% of that category. AI-flagged complaints from people 60 and over averaged $112,153 — 2.81 times the all-ages AI average, an Axis Intelligence Research calculation.
The FTC’s age data shows the other half of the picture. In 2024, 44% of fraud reports from people aged 20–29 included a money loss, against 21% for people 80 and over — but the median loss was $417 for people in their 20s and $1,650 for those 80 and over, per the FTC Consumer Sentinel Data Book 2024 [older data — the 2025 Data Book was not available for this edition]. Younger adults lose money more often; older adults lose more when they do.
| Age group | FTC fraud reports, 2024 | Share reporting a loss | Median loss |
|---|---|---|---|
| 19 and under | 30,330 | 51% | $189 |
| 20–29 | 155,346 | 44% | $417 |
| 30–39 | 203,764 | 40% | $450 |
| 40–49 | 191,508 | 38% | $500 |
| 50–59 | 175,898 | 35% | $520 |
| 60–69 | 208,896 | 29% | $691 |
| 70–79 | 159,550 | 24% | $1,000 |
| 80 and over | 51,713 | 21% | $1,650 |
Source: FTC, Consumer Sentinel Network Data Book 2024. Only 45% of 2024 fraud reports included age.
Social media cuts across every age band. Nearly 30% of people who reported losing money to a scam in 2025 said it started on social media, for $2.1 billion in losses, and every age group except 80 and over lost more to social media scams than to any other contact method, per the FTC’s April 27, 2026 release. For people 80 and over, phone calls still ranked first.
Australia shows the same skew on a different reporting system: people 65 and over make up about 17.1% of the population and 26.5% of losses reported to Scamwatch, per the National Anti-Scam Centre.
What would change this in 2027: state limits on crypto-kiosk transactions and bank holds on suspicious transfers by older customers. Who it affects: families and the banks, credit unions and kiosk operators who see the transaction before it leaves.
Trend 4: Are Romance Scams Increasing?
Yes, on every national dataset checked for this outlook. IC3 confidence/romance losses rose 38.3% to $929.3 million in 2025, with complaints up 29.3% to 23,159. The FBI category includes grandparent and distress scams, so part of this line is the voice-clone story from Trend 3.
Nearly 60% of people who reported losing money to a romance scam in 2025 said it started on social media, per the FTC. UK Finance recorded £39.2 million in romance scam losses in 2025, up 23%. Australia’s combined data put romance scams third at A$139.9 million. The FBI says scammers now use AI-generated profiles and scripts to make conversations more believable; the AI-flagged slice was $19.0 million, 2% of the category — the same attribution gap as investment fraud.
Romance is also the on-ramp to the biggest losses. Relationship-led crypto “investment” schemes, often called pig butchering, get filed as investment fraud once money moves. Our romance scams statistics break the category down further.
What would prove it wrong: IC3 romance complaints below 23,159 in 2026. Who it affects: people 60 and over (62.8% of category losses) and dating-app and social-platform trust teams.
Trend 5: Will Banks Have to Refund Scam Victims?
In the UK, they already do — and it is the clearest test of what liability shifts do to losses. Since mandatory reimbursement rules took effect in October 2024, UK banks returned £354.3 million to APP fraud victims in 2025, 61% of losses, per UK Finance. APP losses still rose 19% to £576.4 million, and 66% of cases began online.
The US has no equivalent rule, but the plumbing is moving. The FBI’s Financial Fraud Kill Chain, which asks banks to freeze fraudulent transfers, froze $679.0 million of $1.164 billion in attempted theft in 2025, a 58% success rate. Operation Level Up notified 3,780 crypto-investment victims in 2025; 78% did not know they were being scammed. Executive Order 14390 of March 6, 2026 directed federal agencies to target transnational crime groups running cyber-enabled fraud, per the FTC’s testimony.
Australia offers the counter-example every forecast should weigh. Combined reported losses there fell 29.7% from the A$3.1 billion peak in 2022, to A$2.18 billion in 2025, after banks, telecoms and platforms were pulled into a shared anti-scam regime. Losses ticked up 7.8% in 2025, and Scamwatch reports fell 17.8% year on year in the first quarter of 2026, per the National Anti-Scam Centre’s Q1 2026 update. Reported losses can bend when the ecosystem is made to pay. That is the downside scenario’s logic.
What would prove it wrong: UK APP losses rising again in 2026 despite reimbursement. Who it affects: US banks and payment apps, whose fraud-liability posture is the variable most likely to move the 2027 number. Payment-rail context is on our payment trends 2027 outlook.
Trend 6: Will Crypto Remain the Main Scam Payment Rail?
Yes. Complaints with a cryptocurrency nexus carried $11.37 billion of losses in 2025 — 54.4% of everything reported to IC3, an Axis Intelligence Research calculation. Chainalysis estimates scams received at least $14 billion on-chain globally in 2025 and projects the year’s total could exceed $17 billion as more wallets are identified. The average scam payment it traced rose from $782 to $2,764.
Recovery fraud is the second bite: IC3 logged $1.4 billion in losses to scams posing as fund-recovery firms or officials in 2025. A victim of one scam is a lead list for the next.
What would prove it wrong: IC3 crypto-nexus losses below $11 billion in 2026. Who it affects: exchanges, kiosk operators and stablecoin issuers asked to freeze funds faster.
Why Do FBI and FTC Scam Numbers Differ?
Different collectors, overlapping data. The FTC’s Consumer Sentinel Network logged 3 million fraud reports and $15.9 billion in losses in 2025, investment scams $7.9 billion of it. The FBI’s IC3 logged $20.877 billion. The FTC lists the FBI’s IC3 among Sentinel’s data contributors — so adding the two double-counts. Use one series per claim, and say which.
| Measure, 2025 | Value | What it counts | Source |
|---|---|---|---|
| IC3 reported losses | $20.877B | Complaints filed with the FBI, mainly US | FBI IC3 |
| FTC Sentinel reported fraud losses | $15.9B | Reports to the FTC and ~180 partners, including IC3 | FTC |
| FTC underreporting-adjusted cost (2024) | up to $195.9B | Modeled estimate | FTC |
| Chainalysis on-chain scam revenue (global) | at least $14B | Traced crypto inflows to scams | Chainalysis (estimate) |
| UK APP fraud losses | £576.4M | Bank-reported push-payment scams | UK Finance |
| Australia combined scam losses | A$2.18B | Five reporting bodies combined | National Anti-Scam Centre |
Source: producers as linked above. Broader totals across fraud types are on our online fraud statistics and cybercrime statistics pages.
What Could Change This Outlook?
| Signpost | Expected date | Pushes toward | Watch it on |
|---|---|---|---|
| Chainalysis 2027 Crypto Crime Report, scams chapter | expected January 2027 (2026 edition: Jan 13) | Trend 6; AI-vendor revenue gap | stablecoin statistics |
| FTC 2026 Consumer Sentinel totals | expected March 2027 | Cross-check on the IC3 base case | online fraud statistics |
| National Anti-Scam Centre Targeting Scams 2026 | expected March 2027 (2025 edition: Mar 30) | Downside logic: do shared-liability regimes keep bending losses? | — |
| FBI IC3 2026 Annual Report | expected spring 2027 | The first grade of this base case: $23.6B for 2026 | AI scam statistics |
| UK Finance Annual Fraud Report 2027 | expected June 2027 | Trend 5: reimbursement vs losses | — |
| FBI IC3 2027 Annual Report | expected spring 2028 | Final grade of the $27.8B base case and Deloitte’s $40B | — |
Expected dates are inferred from each publisher’s previous release. Rules governing AI-generated content and impersonation are tracked on our AI regulation tracker; the agentic-AI side of fraud risk sits in our AI agents trends and cybersecurity trends 2027 outlooks.
How Axis Intelligence Research Built This Outlook
Inputs. Every figure was fetched from its producer on October 31, 2026 and logged with URL and retrieval date in the dataset. No figure was taken from a published Axis page. FBI IC3, FTC, the National Anti-Scam Centre and UK Finance are official or industry-reported datasets; Chainalysis and Deloitte figures are estimates; Pindrop figures are company-reported telemetry and a vendor-commissioned survey. FTC age figures are from the 2024 Data Book, the latest edition obtained. All IC3 figures are complaint-based: losses victims chose to report, adjusted by FBI analysts.
Why this scope. The base case models the sum of IC3’s crime-type loss table ($20.30 billion in 2025) rather than the headline ($20.877 billion), because the crime-type table is where the drivers live and because Deloitte’s forecast was built on the same table. The headline-basis figure ($28.6 billion) scales the projection by the 2025 ratio of headline to crime-type losses.
Driver choice. Complaint counts and average loss are the two observable levers behind every IC3 line. Average-loss growth uses each segment’s 2023–2025 compound rate in all scenarios; complaint growth is the scenario lever (zero, half, full 2023–2025 rate). The “all other” segment is modeled on losses because its 22 crime types have unrelated drivers, and a 2024 jump in personal-data-breach losses would otherwise inflate its average-loss trend.
Limitations. Two years of history is a short base for compound rates. IC3 can reclassify complaints between crime types year to year. The AI descriptor exists for one year only, so no AI-specific growth driver could be modeled. Reporting behavior itself changes: public campaigns raise complaint counts without raising crime. None of the projections are adjusted for inflation.
Grade-back. We will compare the 2026 base case ($23.6 billion) with the FBI’s 2026 report in spring 2027, and the 2027 base case ($27.8 billion) — alongside Deloitte’s $40 billion — with the FBI’s 2027 report in spring 2028. The comparison will open the next edition of this page, right or wrong. This is the first edition of this outlook.
AI Scams in 2027: Questions People Ask
How much money is lost to AI scams?
Complaints the FBI flagged as AI-related reported $893.3 million in losses in 2025, across 22,364 complaints. That is a floor: the flag covers only 5.6% of losses in the six crime types the FBI says AI enhances, so most AI-assisted fraud is filed under investment, impersonation or romance.
Are AI scams going to get worse in 2027?
Reported scam losses are, but more slowly than the most-cited forecast. Axis Intelligence Research’s base case puts FBI-reported losses at about $27.8 billion in 2027, up from $20.3 billion in 2025; Deloitte’s $40 billion would require growth of 40.4% a year.
What is the most common AI scam?
By money, AI-assisted investment fraud: $632.0 million of the $893.3 million in AI-flagged FBI losses in 2025. By reach, impersonation — toll, bank, tech-support and government-agency scripts, which Chainalysis saw grow more than 1,400% on-chain.
Can scammers really clone a family member’s voice?
Yes. The FBI reports victims lost over $5 million in 2025 to “distress” scams that use voice cloning to mimic a relative in trouble, and says the scheme is expanding to friends and other family roles. Agree on a family code word and call back on a known number.
Who is most targeted by scams, older or younger people?
It depends on what you measure. Younger adults lose money more often: in FTC data for 2024, 44% of fraud reports from people in their 20s involved a loss, versus 21% for people 80 and over. Older adults lose far more: people 60+ filed 20.0% of FBI complaints in 2025 but carried 37.1% of losses, averaging $38,500 each.
How can you spot a deepfake on a video call?
The FBI’s tell from deepfaked job interviews: lip movement that doesn’t match the audio, and coughs or sneezes that don’t line up with the picture. For money requests, the stronger test is procedural — end the call and verify through a channel you already trust.
Is the “$40 billion AI fraud by 2027” figure real?
It is a real Deloitte forecast from May 2024, not a measurement. It started from $12.3 billion in 2023, the same base as the FBI’s crime-type losses. FBI-reported losses reached $20.3 billion in 2025, 8.5% below the path to $40 billion.
Where do most scams start now?
On social media, measured by money. Nearly 30% of people who lost money to a scam in 2025 said it began on social media, for $2.1 billion in reported losses, per the FTC — the costliest contact method for every age group except 80 and over, who lose most through phone calls.
What should you do if you’ve been scammed?
Call your bank or payment provider first and ask for a recall — the FBI says time is critical. Then report at ic3.gov with full transaction details; the FBI’s Recovery Asset Team froze 58% of attempted theft in the cases it acted on in 2025. See our guide on how to report an online scam.
About This Dataset
ai-scam-trends.csv holds 418 rows: IC3 losses and complaints by all 27 crime types for 2023–2025, losses by age group, AI-related and cryptocurrency descriptors, the FTC (Sentinel, Do Not Call, age and social-media data), Pindrop, Chainalysis, UK Finance and National Anti-Scam Centre figures cited above, Deloitte’s forecast with the Axis interpolation, and every base-case driver, scenario and sensitivity result with its formula. Each row carries source organization, URL, retrieval date, a primary-source flag and, for Axis rows, the method. License: CC BY 4.0.
How to cite this page
- APA: Axis Intelligence Research, & Chen, M. (2026). AI scam trends 2027: Reported losses head toward $28 billion, and the $40 billion forecast is off track. Axis Intelligence. https://axis-intelligence.com/ai-scam-trends/
- MLA: Axis Intelligence Research, and Marcus Chen. “AI Scam Trends 2027.” Axis Intelligence, 31 Oct. 2026, axis-intelligence.com/ai-scam-trends/.
- Chicago: Axis Intelligence Research and Marcus Chen. “AI Scam Trends 2027: Reported Losses Head Toward $28 Billion, and the $40 Billion Forecast Is Off Track.” Axis Intelligence, October 31, 2026. https://axis-intelligence.com/ai-scam-trends/.
Review trigger: the FBI IC3 2026 Annual Report, or the FTC’s 2026 Sentinel totals, whichever comes first.
More outlooks in Cybersecurity & Digital Trust.
Found an error? [email protected] — corrections are dated and published at the bottom of this page.