Faces for Hire: How Synthetic Media Has Become the Counterfeit Currency of Modern Fraud

For most Americans, the word "deepfake" conjures images of politicians saying things they never said, or celebrities inserted into videos without consent. Those concerns are legitimate. But inside law enforcement offices from San Francisco to Manhattan, a more financially motivated deepfake crisis has been quietly escalating — one that targets corporate treasuries, personal bank accounts, and the biometric systems that were supposed to make all of it more secure.

Synthetic media, meaning video, audio, or imagery generated or manipulated by artificial intelligence, has become a foundational instrument in a new class of fraud. The technology is no longer the exclusive domain of well-resourced nation-state actors or sophisticated research labs. Commercially available tools, some free, some available for a modest subscription, have democratized the production of convincing forgeries. What was once a technical novelty is now a criminal commodity.

The Anatomy of a Deepfake Fraud Operation

The mechanics of a synthetic media fraud campaign vary depending on the target, but several models have emerged as dominant.

The most financially damaging variant is known as a business email compromise attack augmented by deepfake video — sometimes called "BEC 2.0" within the cybersecurity community. In early 2024, a finance employee at a multinational firm based in Hong Kong was summoned to a video conference call that appeared to include his company's chief financial officer and several senior colleagues. The call was entirely fabricated. Every face on screen was a deepfake, constructed from publicly available footage. The employee was instructed to authorize a series of wire transfers totaling approximately $25 million. He complied. The real CFO knew nothing about it.

While that case unfolded overseas, U.S. investigators have documented similar schemes targeting American companies. The FBI's Internet Crime Complaint Center flagged a sharp increase in AI-assisted fraud complaints in its most recent annual report, with business email compromise remaining the costliest category of cybercrime reported by American victims.

A second major vector involves voice cloning. Fraudsters harvest audio — from earnings calls, YouTube interviews, podcast appearances, or social media — and use it to train models that can replicate a person's voice with unsettling accuracy. These synthetic voices are then deployed in phone calls to bank representatives, family members, or employees. In some documented cases, criminals have impersonated corporate executives to authorize emergency fund transfers. In others, they have impersonated distressed family members to extract money from elderly relatives — a digital evolution of the so-called "grandparent scam."

Biometric Systems Under Siege

Perhaps the most alarming development for the financial services industry is the documented ability of deepfake technology to defeat biometric authentication systems. Many banks and financial platforms have invested heavily in liveness detection — systems that require users to blink, turn their heads, or speak a phrase to confirm they are a living human being, not a photograph.

Researchers and, increasingly, criminal actors have demonstrated that these systems can be fooled. Sophisticated injection attacks, in which a pre-rendered deepfake video is fed directly into the camera input stream of a device rather than displayed in front of a physical camera, have bypassed several commercial liveness detection products. The implications for remote identity verification — a process that became ubiquitous during the pandemic and has remained standard for account opening at many institutions — are significant.

The financial identity verification firm iProov published findings in 2023 noting a 255 percent increase in deepfake-related attacks against its clients over the preceding year. That figure has continued to climb.

Romance Fraud at Industrial Scale

Beyond the corporate environment, deepfakes have dramatically amplified the efficiency of romance scams. Traditionally, romance fraud required a human operator to maintain ongoing text-based correspondence with victims — a labor-intensive process. Synthetic media has changed that calculus.

Operators running so-called "pig butchering" schemes — long-con investment frauds that cultivate emotional relationships with victims before stealing their savings — have begun incorporating AI-generated video personas to accelerate trust-building. A victim who might have grown suspicious of a purely text-based relationship is far more likely to believe in the authenticity of a person they have seen and heard on video calls, even if those calls are staged using synthetic faces and voices.

The Department of Justice has prosecuted several large pig-butchering networks with ties to Southeast Asian criminal organizations, and investigators have noted the increasing use of AI-generated media within those operations.

Extortion: The Darker Application

Synthetic media has also fueled a surge in non-consensual intimate image fraud — a category that legal advocates have fought to address through legislation, with several states now having enacted specific criminal statutes. Criminals use AI image generation tools to create fabricated explicit imagery of real individuals, then threaten to distribute it unless payment is made. Victims are frequently teenagers and young adults, though the FBI has documented cases involving adults across all age groups.

This form of extortion requires no prior relationship with the victim. A perpetrator needs only a collection of publicly available photographs — the kind most Americans post routinely to social media — to manufacture credible threats.

Detection and Defense

The defensive landscape is evolving, though practitioners caution that it remains a persistent arms race.

For organizations, the most effective countermeasure remains procedural rather than technological: mandatory out-of-band verification for any financial transaction above a defined threshold. If an executive appears on video requesting an urgent wire transfer, policy should require a separate, independently initiated phone call to a known number to confirm the request. No deepfake can intercept a call placed by the intended recipient to a verified contact.

On the technical side, several companies are developing and deploying deepfake detection tools that analyze video for artifacts — unnatural eye blinking patterns, inconsistent lighting on facial surfaces, subtle warping at the edges of the face, or audio-visual synchronization anomalies. These tools are imperfect but improving. Content provenance standards, including the Coalition for Content Provenance and Authenticity (C2PA) framework, are being adopted by some camera manufacturers and media platforms to cryptographically certify the origin of authentic recordings.

For individuals, the guidance is more straightforward. Treat any unexpected video call requesting money or sensitive information with immediate skepticism, regardless of how familiar the face or voice appears. Establish verbal code words with family members for use in emergency contact situations — a low-tech safeguard that has already thwarted documented fraud attempts. And review your public social media footprint with an awareness that the audio and imagery you post can be harvested and repurposed.

The Infrastructure Behind the Market

The deepfake fraud ecosystem does not operate in isolation. Dark web forums and encrypted marketplaces have developed dedicated sections for the trade of synthetic media services — deepfake-as-a-service offerings that allow criminals without technical expertise to commission convincing forgeries for a fee. Pricing structures documented by threat intelligence researchers range from a few hundred dollars for a simple voice clone to several thousand for a fully interactive video persona.

Law enforcement has begun to treat this infrastructure as it would any other criminal market. The FBI, in coordination with international partners, has targeted deepfake service providers in several recent operations, though the decentralized and pseudonymous nature of these markets makes sustained disruption difficult.

The synthetic media threat is not a future concern. It is an active and expanding criminal economy, and the gap between what fraudsters can produce and what victims can reliably detect has never been wider.