FBI reports AI scams stole hundreds of millions in 2025
AI Fraud Now Tracked as Distinct Crime Category
The Federal Bureau of Investigation's Internet Crime Complaint Center (IC3) documented 22,364 complaints explicitly marked as involving artificial intelligence in 2025, with $893,119,903 in associated financial losses. The agency's annual Internet Crimes Report, released in May 2026, represents the first time the FBI has tracked AI-related complaints as a separate descriptor, signaling the emergence of autonomous fraud as a measurable criminal phenomenon.
The scale of losses—nearly $900 million in a single year—reflects how rapidly AI-generated fraud has penetrated U.S. crime ecosystems. These are not hypothetical threats: they are documented financial transfers from victims to perpetrators, validated through formal complaint channels and loss reporting mechanisms.
North Korean Operatives Weaponizing Deepfakes
The FBI has previously issued warnings about North Korean operatives deploying deepfake and fabricated identities to infiltrate U.S. companies posing as remote IT workers. This tactic exploits the shift to distributed workforces and relies on synthetic media—AI-generated video, audio, and identity documentation—to bypass hiring and security vetting processes.
Once inside company networks, such operatives can access sensitive systems, exfiltrate data, or install malware. Unlike traditional phishing or social engineering, deepfake-based infiltration scales across multiple targets simultaneously and bypasses the human intuition that might catch inconsistencies in text-only communication.
What the Numbers Mean
The FBI's 2025 data captures only reported complaints. Industry estimates suggest actual losses run higher, as many victims—particularly businesses that keep breaches confidential—never file formal reports. The $893 million figure therefore represents a confirmed floor, not a ceiling.
The decision to track AI crimes separately reflects institutional recognition that traditional fraud categories (phishing, business email compromise, confidence schemes) increasingly involve AI-generated components. Separating them enables the FBI to allocate resources more precisely and provides law enforcement with clearer patterns for prosecution.
