AI can help scammers imitate a trusted person, produce convincing fake documents, or tailor a message that pressures you to act. It does not have to break into your bank account: a common danger is being manipulated into sharing credentials or authorizing a payment yourself. Banks and finance teams are responding with layered checks, monitoring, and staff training, but no single detector can make every request safe.
This article focuses on U.S. consumers and financial institutions. Available figures do not establish a comprehensive total for losses caused by AI-enabled fraud.
As an Amazon Associate I earn from qualifying purchases.
How AI can turn familiar fraud into a more convincing scam
Impersonation can sound or look like someone you trust
Generative AI can help create or adapt scam messages, scripts, voice recordings, and video impersonations. A call or video that seems to come from a family member, executive, or financial institution is not independent proof of identity. Treat it as a claim to verify, not as authentication.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →In a 2025 speech, Federal Reserve Governor Michael S. Barr described a reported case in which an employee was deceived by a deepfake video meeting and transferred $25 million to thieves. That is an individual case, not a typical loss or a measure of how often such attacks succeed. Its practical lesson is narrower: seeing a person on screen does not prove who is behind the request.
#1 Best Overall
Fake documents and synthetic identities can target onboarding checks
AI can also help produce altered or fabricated identity documents and other evidence used to impersonate someone or create a synthetic identity. FinCEN’s November 13, 2024 alert described increased suspicious activity reporting involving suspected deepfake media, particularly fraudulent identity documents used to evade identity checks. The alert is a warning about reported suspicious activity, not proof that every identity fraud attempt involves AI.
The money often moves because a real person is persuaded to move it
In many payment scams, the account holder is legitimate, but a criminal persuades that person to disclose access information or approve a transfer. AI may make the story, evidence, or impersonation more believable; the financial event is still the victim’s action. That distinction matters when assessing both prevention and the possibility of recovering funds.
What the reported figures do—and do not—show
Several official figures illustrate the broader fraud environment, but they measure different things. None should be read as an estimate of money stolen because of AI.
Recommended Free Tools
| Figure | What it measures | Important qualification |
|---|---|---|
| $3.31 billion in 2022 and $4.57 billion in 2023 | FBI-reported losses from fake investment opportunity scams, as reported by the U.S. Government Accountability Office (GAO) in 2024. | These figures cover one scam type and complaint-reported losses; they are not an estimate of AI-caused losses or all fraudulently induced payments. |
| 23% | Share of surveyed institutions reporting account takeover by authorized parties. | Federal Reserve Financial Services’ survey of more than 400 risk professionals at institutions using its services was conducted in Q4 2025 and reported in 2026. This is an institution survey finding, not the share of consumers affected or an AI-attributed rate. |
| 75% | Share of surveyed institutions reporting debit-card fraud attempts. | This is from the same Q4 2025 survey and likewise does not measure consumer incidence or AI’s contribution. |
| $25 million | Amount in the reported Arup deepfake-video payment fraud case discussed by Barr in 2025. | It is one reported case, not a typical consumer exposure or a frequency estimate. |
GAO says there is no complete estimate of overall losses from fraudulently induced payments. Combining these numbers into a single AI-fraud total would confuse unlike measures and imply a causal link the sources do not establish.
How to check an urgent call or payment request
Use a channel you already trust to verify both the person and the payment details. Do not rely solely on a number, link, voice, or video included in the unexpected request. A caller ID can appear familiar, and a convincing voice is not a second verification factor.
- Stop the interaction. Do not share a password, one-time code, or account information, and do not approve a transfer while under pressure. Urgency, secrecy, or a last-minute change to payment instructions is a reason to pause.
- Contact the person or institution independently. Use a number printed on your bank card or shown in its official app, or use a contact method you already have for the person. Do not call back using the number in the message or let the caller keep you on the line while you verify.
- Confirm the exact transaction. Ask whether the request is genuine and independently check the recipient and payment details. A real relationship or familiar voice does not establish that new instructions are legitimate.
- Enable multi-factor authentication. Turn it on for financial accounts where available. A hardware security key may be an option if your institution supports it; check compatibility with the institution before relying on one.
- Report suspected exposure quickly. Contact your financial institution promptly if you shared credentials, approved a suspicious transaction, or believe someone has access to your account. Ask what steps it can take; a transfer may not be reversible.
Will a bank refund money sent to a scammer?
There is no blanket promise of reimbursement for a payment that a customer personally authorized after being manipulated. GAO’s 2024 report says federal law generally does not require financial institutions to reimburse consumers for fraudulently induced authorized payments, and recovery can be difficult. That is different from an unauthorized account transaction, which can involve different protections. The outcome depends on the circumstances; contact the institution immediately and do not assume either that a transfer can be reversed or that a particular claim has a guaranteed result.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What banks and finance leaders are doing
The response is layered rather than dependent on a single AI detector. In his April 17, 2025 speech on banking cybersecurity, Barr pointed to identity verification, multi-factor authentication, account monitoring, review of recipient details, additional scrutiny of suspicious transactions, employee training, and threat-information sharing. These measures aim to make it harder to impersonate an account holder or move money unnoticed; they cannot guarantee that every scam will be caught.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minuteReviewing identity, recipient, and transaction risk
Institutions can check whether identity evidence is credible, monitor account activity, and scrutinize unusual transactions or suspicious recipients. Extra review can add time before a high-risk payment proceeds. Barr’s framing is useful for customers as well: “When it comes to protecting our money, we ought to expect and appreciate a little friction.”
Using AI as an additional signal, not a verdict
Federal Reserve Financial Services describes AI-assisted analysis as a way to look for anomalies across transaction records and information such as text, audio, and images, alongside existing rules and predictive tools. An AI system should not be treated as a reliable detector of every deepfake or scam. It can help prioritize review, while people and other controls remain part of the decision.
Governing the tools and sharing threat information
U.S. Treasury’s 2024 financial-services AI report calls for collaboration, AI-specific information sharing, regulatory coordination, and checking AI use cases against existing laws. It also identifies privacy, bias, and third-party-provider risks. Federal Reserve Financial Services likewise emphasizes governance around data privacy, model transparency, misuse, and staff training. Finance leaders need to assess not just whether a model flags anomalies, but what data it uses, how its output is reviewed, and how errors or misuse are handled.
Quick Recap
A quick verification checklist before money moves
- Is the request unexpected, urgent, secret, or changing payment details? Pause.
- Have you verified the request through a separate, already trusted contact route?
- Have you independently checked the recipient and exact payment instructions?
- Are multi-factor authentication and account alerts enabled where available?
- If you may have exposed access or sent money, have you contacted the institution promptly?
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Free tools Windows power users keep installed
One-click scans. No signup required.
