Criminals running romance scams have added a new weapon to their playbook: real-time deepfake video calls that let them pose as someone else on camera, pass visual “proof of identity” checks, and keep victims sending money for months. The UK’s financial intelligence unit has confirmed that deepfake video is already appearing in romance fraud cases reported through suspicious activity reports, with at least one victim losing substantial savings after repeated video verification. Across the Pacific, Hong Kong police coordinated a cross-border operation with nine countries and regions targeting organized fraud networks that use similar impersonation tactics, while the U.S. Federal Trade Commission recorded $12.5 billion in total reported fraud losses for 2024.
Live deepfake video calls defeat the oldest fraud safeguard
For years, the standard advice to anyone suspicious of an online love interest was simple: insist on a live video call. A still photo can be stolen, but a moving, speaking face on camera was supposed to be hard to fake. That assumption is now outdated. The UK National Crime Agency’s financial intelligence unit, in its SARs in Action Issue 26 bulletin, documented that deepfake technology is being used in romance frauds, including video calls designed to deceive victims, and the bulletin is available through the agency’s publications page. In one referenced case, a victim continued transferring funds precisely because video calls appeared to confirm the scammer’s claimed identity.
The mechanics are straightforward. Scammers obtain photos or short clips of an attractive person, feed the material into face-swapping software, and run the output through a webcam in real time. Lighting glitches and audio lag still occur, but the quality has improved enough to fool people who are already emotionally invested. Europol’s public guidance on romance scam manipulation patterns notes that fraudsters rely on pressure tactics, manufactured urgency, and emotional leverage to manage verification moments, and deepfake tools now give them a way to clear the video-call hurdle entirely rather than dodge it.
The FBI describes a consistent pattern in which criminals build trust on dating platforms, then move targets to private messaging channels before requesting money. That playbook has not changed. What has changed is the scammer’s ability to survive the one moment most likely to expose the con. When the person on screen looks and sounds like the photos, victims lose their strongest reason to doubt, especially if the deepfake is combined with convincing backstories, stolen social media content, and scripted responses to common questions.
Cross-border enforcement and $12.5 billion in U.S. fraud losses
Organized scam operations increasingly span multiple countries, making enforcement difficult for any single jurisdiction. In May 2026, Hong Kong police conducted a coordinated crackdown known as Operation FRONTIER+, a cross-border anti-scam effort involving nine countries and regions that targeted industrialized fraud networks running romance cons, investment scams, and other impersonation schemes at scale, as outlined in an official government release. Arrests and seizures from the operation reflect the size of the criminal infrastructure behind these schemes, though the announcement does not isolate which cases specifically involved AI-generated video or other deepfake tools.
On the financial side, the scale of fraud in the United States alone is staggering. Reported losses to fraud reached $12.5 billion in 2024, according to the Federal Trade Commission’s Consumer Sentinel data, which aggregates consumer complaints across multiple categories. That figure covers imposter scams, including romance fraud, as well as investment schemes, business email compromise, and other tactics. The FTC does not currently break out losses caused specifically by deepfake-enabled romance cases from other subtypes, a gap that limits how precisely regulators can track the impact of AI-driven impersonation.
The FBI directs romance scam victims to file reports through IC3, its Internet Crime Complaint Center, and provides detailed prevention and reporting advice on its dedicated romance scam guidance. That reporting channel feeds into federal investigations and helps build cases against organized networks that may operate call centers or “scam farms” overseas. Yet the absence of a dedicated deepfake tag in most reporting systems means analysts cannot easily measure whether AI video impersonation is driving higher per-victim losses or simply adding a new layer to existing tactics that were already highly profitable.
No reliable public data yet links deepfakes to higher per-victim losses
A reasonable expectation is that deepfake-enabled romance scams produce larger individual losses because victims stay engaged longer when video calls appear to confirm the relationship. If that pattern holds, jurisdictions that track deepfake mentions in suspicious activity reports or IC3 filings should eventually show a measurable rise in average romance-scam loss amounts within a year or so of the first documented deepfake cases. But no public dataset currently tests that hypothesis in a systematic way.
The UK financial intelligence bulletin references deepfake use without publishing aggregate statistics on victim losses, reporting volume, or the proportion of romance scams that now involve AI-generated video. Similarly, the FTC’s Consumer Sentinel system aggregates imposter-scam losses but does not release data separating deepfake-enabled romance cases from other imposter frauds such as family-emergency or government-impersonation schemes. That leaves regulators and researchers relying on anecdotal law-enforcement reports and case studies rather than robust trend data.
This data gap has practical consequences. Without clear evidence that deepfake-enabled scams cause higher losses, it is harder to justify dedicated funding for AI-detection tools, specialized training for investigators, or new regulatory requirements for video platforms and dating apps. Banks and payment providers, which often serve as the last line of defense when victims attempt large transfers, also lack detailed typologies that would help them distinguish deepfake-driven romance fraud from more traditional variants when filing suspicious activity reports.
Some early signals may eventually emerge from how financial intelligence units categorize their cases. If analysts consistently flag deepfake use in narrative fields, researchers could mine those narratives to estimate prevalence and average loss. Cross-border operations like Hong Kong’s FRONTIER+ may also begin to publish more granular breakdowns of the technologies seized from scam syndicates, offering indirect evidence of how widely real-time face-swapping tools are being deployed in romance fraud.
What victims and platforms can do while the data catches up
In the absence of detailed statistics, prevention efforts still matter. Individuals can treat video calls as one data point rather than definitive proof of identity, watch for inconsistencies between voice, facial movement, and background, and refuse to send money or cryptocurrency to anyone they have not met in person. Requests that hinge on emergencies, investment opportunities, or secrecy remain strong red flags, whether or not a face appears on screen.
Platforms that host dating profiles and messaging should consider adding friction around high-risk behaviors, such as prompts when users are asked to move conversations off-platform quickly or when certain payment keywords appear in chats. They can also provide in-app education about deepfake risks, emphasizing that even live video is no longer a guarantee of authenticity. Payment providers, for their part, can refine monitoring rules to look for patterns associated with romance fraud, such as repeated transfers to new overseas accounts justified as “helping a partner.”
Law-enforcement agencies and regulators face a parallel challenge: updating guidance quickly enough to reflect emerging threats without waiting years for perfect data. Bulletins like the UK financial intelligence unit’s deepfake warning and cross-border crackdowns such as Operation FRONTIER+ show that authorities are beginning to take AI-enabled romance scams seriously. The next step will be to translate those early warnings into standardized reporting categories and public statistics that can reveal whether deepfake video is simply another tool in the scammer’s kit or a genuine multiplier of financial harm.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
