FBI: $893M in AI fraud losses during 2025
The FBI's Internet Crime Complaint Center (IC3) documented $893 million in losses from AI-enabled fraud during 2025, the first year the agency broke artificial intelligence out as its own standalone crime category, according to the FBI's 2025 Internet Crime Report released in April 2026.
The tally encompasses 22,364 individual complaints where victims or reporting parties cited AI involvement, yet the FBI emphasized the $893 million figure is "almost certainly higher" because many victims do not recognize when AI tools—including voice cloning and deepfake impersonation—have been deployed against them.
Investment Fraud and Government Impersonation Drive Losses
Within the AI-fraud category, investment fraud accounted for $632 million in losses, making it the largest driver. Business Email Compromise (BEC) caused $30 million in AI-linked losses, while romance scams generated $19 million. The broader impersonation category, where voice cloning and deepfake technology are most frequently deployed, saw government impersonation complaints nearly double year-over-year to 32,500 complaints, resulting in $797 million in losses across all impersonation types.
This shift reflects a fundamental change in how criminals operate. The FBI noted that AI has evolved from being the *subject* of fraud schemes to becoming the *machinery* behind them, dramatically amplifying traditional tactics like business email compromise, romance scams, and credential theft. Agents and AI-generated personas now execute these attacks at scale.
Context Within Broader Cybercrime Landscape
The $893 million in AI-enabled fraud represents a significant slice of the $20.877 billion in total cyber-enabled crime losses that IC3 recorded across all categories in 2025. The agency's decision to separately track AI-driven crime for the first time underscores law enforcement's recognition that artificial intelligence has become a material and measurable vector for financial crime.
The report arrives as federal agencies, tech companies, and the private sector grapple with rapid proliferation of accessible voice synthesis, deepfake generation, and large language model tools that lower the barrier to executing sophisticated impersonation attacks. Government impersonation—often involving AI-generated voice calls or synthetic video—has proven particularly lucrative for criminals, with losses nearly doubling within a single year.
The IC3 data provides concrete evidence of the agent economy's darker dimension: autonomous and semi-autonomous systems deployed not to streamline legitimate business processes, but to deceive, defraud, and steal at speed and scale. The fact that the reported total likely understates reality suggests the true cost of AI-enabled fraud in 2025 may be substantially higher than the documented $893 million.
