Voice cloning, deepfakes drive $893M in AI scams

The 2025 FBI Internet Crime Report records $893,346,472 in losses tied to 22,364 complaints involving AI-based scams reported by Americans. The report identifies voice cloning, deepfake images and video, and AI-generated scripts as the main technologies used in these cases.

Those tools have been used to enhance established frauds such as romance scams, extortion and impersonation. Scammers combine cloned voices, doctored video and realistic written messages to create multi-channel attacks that appear authentic and can reach many people.

Government impersonation schemes have become more complex. Incidents that once involved simple phone calls demanding gift cards have developed into operations using spoofed caller ID, stolen agency logos and AI-created audio or video of public officials to persuade victims they are dealing with legitimate authorities.

Companies have also seen losses. Business email compromise cases that incorporate AI techniques have resulted in losses in the tens of millions of dollars. Criminals use automated tools to research targets, craft tailored messages and impersonate executives or vendors to prompt urgent payments.

Michael Machtinger, deputy assistant director of the FBI Cyber Division, said: “AI-created fraudulent communications can look very official and very legitimate to even the most trained individuals.”

The FBI count includes only incidents reported to the agency, and the agency says the figure likely underestimates total losses because many victims do not report fraud. Authorities and private security teams are urging quicker reporting and improved detection to better track and stop these schemes.

Agencies and financial institutions recommend that people verify identities through official contact channels before responding to unexpected requests, be wary of urgent payment demands-especially those requesting cryptocurrency or gift cards-and limit how much voice and video content they share publicly. Victims are advised to report incidents to their banks and to the FBI’s Internet Crime Complaint Center at IC3.gov.

Security vendors are increasingly using AI to scan large amounts of telemetry, detect anomalies and block attacks that progress faster than people can respond. The report notes that defensive tools are being developed to analyze suspicious links, messages and attachments and to help users check content before responding.