A criminal might impersonate someone you trust, send convincing payment instructions, or operate thousands of fake online profiles at once. The message, voice, or person may appear authentic—even when the entire situation has been manufactured.
This week, AI Digital Money examined three recent warnings involving investment fraud, real-estate wire fraud, and deceptive dating applications.
We separated confirmed information from claims that still require caution.
Scam Warning 1: FBI Reports $893 Million in Complaints Referencing AI
The FBI’s Internet Crime Complaint Center received 22,364 complaints referencing AI during 2025, with $893,346,472 in reported losses.
According to the FBI:
Investment-fraud complaints reporting an AI connection accounted for approximately $632 million in losses.
Complainants age 60 and older reported approximately $352.5 million in losses involving the FBI’s AI-related descriptor.
Criminals are using AI-generated content to support investment fraud, impersonation, phishing, distress scams, and other schemes.
These numbers are based on information people reported to the FBI. They do not necessarily mean that investigators independently confirmed AI use in every complaint.
Why It Matters
AI allows criminals to create polished messages, convincing images, and realistic voices without needing a large operation.
That means the old advice to simply look for spelling mistakes is no longer enough. A scam can look and sound professional while still being fraudulent.
Scam Warning 2: A Homebuyer Lost $66,000 Before Closing
CNN reported that Michigan homebuyers Brian and Wendy VanDoeselaar lost $66,000 after receiving fraudulent communications shortly before their home closing.
According to CNN, the couple received what appeared to be legitimate wiring instructions and a telephone call they believed sounded like their mortgage professional. The incident was reported as a suspected voice-cloning scam, but public reporting does not establish conclusively that investigators proved an AI-cloned voice was used.
The broader FBI numbers require an important distinction:
All reported real-estate fraud produced approximately $275.1 million in losses during 2025.
That was about 58.5% higher than the $173.6 million reported in 2024.
Within the 2025 data, only 115 real-estate complaints—representing approximately $2.7 million in losses—were reported as having an AI connection.
Therefore, the $275.1 million figure covers real-estate fraud broadly. It should not be presented as the amount lost specifically to AI voice-cloning scams.
Why It Matters
Home closings involve large transfers, strict deadlines, and numerous emails. Criminals exploit that pressure by impersonating lenders, attorneys, title companies, and real-estate professionals.
Never assume new payment instructions are authentic just because the email address, caller ID, or voice appears familiar.
Scam Warning 3: Thousands of AI Personas Operated on Dating Apps
Anthropic disclosed that it disrupted an operation involving a China-based app studio that reportedly created a network of more than 20 dating applications.
During a two-week period in April 2026, Anthropic identified:
More than 4,700 AI personas
At least 25,000 people who interacted with them
Approximately 2.36 million messages
A ratio of roughly three AI personas for every real person
Users reportedly saw a feed containing approximately 75% AI personas and 25% real people, without being told which profiles were automated.
The operator’s instructions reportedly told the AI personas not to reveal that they were automated, to avoid requests for photographs or live video, and to move users through predetermined conversational stages.
The applications used paid in-app coins and metered messaging, meaning users could spend real money continuing conversations with profiles they believed were human.
Anthropic also reported that the operation used real gig workers for tasks the AI could not perform, including live video calls and social-media follow-backs. The applications were allegedly designed to hide certain functionality during app-store reviews.
Why It Matters
A friendly conversation, realistic photograph, or occasional video interaction no longer guarantees that an online relationship is genuine.
The person appearing briefly on video might not be the same entity handling the rest of the conversation.
Warning Signs to Watch For
Be cautious when:
An unexpected caller requests money or sensitive information.
Someone suddenly changes payment or wiring instructions.
A caller creates urgency and discourages you from contacting anyone else.
An online match repeatedly avoids spontaneous live video.
Conversations feel unusually smooth, repetitive, or scripted.
A platform requires coins or credits to continue a personal conversation.
Someone you met online requests money, cryptocurrency, gift cards, or financial information.
You are told to keep a payment or emergency secret.
One warning sign does not automatically prove fraud, but several warning signs together should make you stop and verify.
What You Should Do Today
Create a family security word.
Require relatives to use it during unexpected financial emergencies.
Call back using a trusted number.
Do not use a telephone number supplied in the suspicious message.
Verify wiring instructions independently.
Contact the lender, attorney, or title company using information collected before the payment request arrived.
Require a second person to review large transfers.
A spouse, adult child, trusted friend, or financial professional may notice something you missed.
Pause before unexpected payments.
Criminals create urgency because slowing down gives you time to uncover the deception.
Question paid online conversations.
Be cautious when someone quickly pushes you toward coins, credits, subscriptions, or another paid platform.
Never send money to an online romantic interest you haven’t verified independently.
Report suspected internet crime promptly.
Submit a report at IC3.gov. Rapid reporting may improve the possibility of stopping or recovering a fraudulent transfer, although recovery is never guaranteed.
What the Data Does Not Prove Yet
The available evidence does not establish:
That every complaint categorized as AI-related involved independently confirmed AI use.
That AI caused the overall increase in real-estate fraud.
That the Michigan homebuyer case conclusively involved a cloned voice.
That the dating-app operation was the first scheme of its kind.
That a convincing voice can always be cloned for a specific advertised price.
These distinctions matter. Consumer warnings should be serious without exaggerating what the evidence proves.
Learn • Earn • Protect
Learn: AI can help criminals imitate voices, create identities, and operate deceptive conversations at scale.
Earn: Every dollar protected from fraud remains available for your family, savings, business, and future goals.
Protect: When a request involves money, secrecy, or urgency, stop and verify it through a communication channel you control.
Stay informed. Slow down. Verify before you trust.
Mark Turner
AI Digital Money
Learn • Earn • Protect