FBI Documents Show Crypto and AI Fraud Surge

The Federal Bureau of Investigation, through its IC3 annual report for 2025, documents a major escalation in online fraud driven by cryptocurrency and artificial intelligence. Victims filed more than 1,000,000+ complaints, reporting combined losses of about $20.9 billion. Cryptocurrency-related complaints numbered 181,565, with losses totaling $11.366 billion, and the report logs 22,364 AI-enabled scam complaints that generated approximately $893 million in losses. Older adults bore a heavy share of the damage, with Americans aged 60 and over losing about $4.4 billion to crypto scams. The average reported loss per crypto victim was near $62,604.

Scammers are combining cryptographic rails and generative AI to scale confidence fraud and make traditional red flags harder to detect. Noticeable tactics include:

• •synthetic voice cloning to impersonate relatives or executives
• •fabricated identity documents and "proof-of-life" videos built with deepfakes
• •coordinated social-media persona creation and fake documentation to build trust over time
• •rapid onchain transfers and obfuscated mixing to remove recoverability

These patterns change the feature space defenders must model. Signal sources that matter now include cross-modal similarity scores between voice and video, provenance metadata on media, temporal graph features for payment flows, and onchain clustering signals from blockchain analytics providers.

Agency and industry response: The FBI highlights proactive initiatives that reduced losses and routed investigations; examples cited include Operation Level Up, which alerted thousands of potential victims and prevented hundreds of millions in losses, and newer actions under Operation Winter SHIELD. Public-private collaboration and tooling improvements are emphasized:

• •increased partnership with blockchain analytics firms such as Chainalysis
• •targeted outreach campaigns and victim notifications
• •expanded reporting mechanisms inside IC3 and enriched data sharing with banks and exchanges

These responses reflect a shift from reactive takedowns to coordinated prevention and cross-domain signal fusion.

This is the first IC3 report to dedicate a full section to AI-enabled scams, marking a structural change in the threat landscape. For ML and security practitioners, the implications are practical: fraud detection models must absorb multimodal inputs, defenses must handle adversarially generated content, and onchain monitoring needs to integrate with synthetic-media detectors. The economics are material. Cryptocurrency schemes now account for the largest share of reported financial damage, and AI-related losses, while smaller in dollar terms, grow rapidly and erode traditional behavioral heuristics used by rule-based detection.

Expect accelerated investment in multimodal detection, tighter exchange KYC/AML practices, and more vendor offerings that fuse media provenance, voice biometrics, and onchain analytics. Teams should prioritize synthetic-media watermarking, graph-based transaction features, and incident response playbooks that include rapid onchain freezing and outreach to vulnerable demographics.