FBI logged 22,000 AI-linked fraud complaints in its first year of counting them

The two subtotals inside the FBI's count do not fit inside its total. Investment fraud accounts for $632 million of the $893 million reported ^[3][2], and victims over 60 account for $352 million ^[4]. Added, that is $984 million, about $91 million more than the whole ^[18]. An over-60 investment-fraud victim sits in both categories, so these are overlapping cuts of one pool. Investment fraud on its own is 71% of the reported money ^[21].

Divide the losses by the complaints and the average reported loss is about $40,600 ^[19]. A mean sitting behind an investment-fraud total is pulled up by the largest victims. And the $40,600 is an upper bound anyway, since the count is "more than 22,000" ^[2].

2025 is the first year the Internet Crime Complaint Center tracked the AI connection at all ^[1], so the $893 million is the whole series ^[23]. The counting rules matter more than the size of the total. Only victims who reported are in it, and only cases where AI's role could be identified ^[5]. If the number rises next year, that could be more AI fraud or better recognition of it.

Cloning a voice takes a few seconds of audio and cheap consumer tools ^[11]. In one study, listeners identified an AI-generated voice about 60% of the time ^[7]. The task design decides what that rate means, and the article gives the rate alone ^[22].

The best-documented losses in the piece belong to companies. A finance employee at the design firm Arup wired about $25 million in 2024 after a video meeting staffed by deepfakes of the chief financial officer and several colleagues ^[8]. Anthropic disrupted an espionage campaign last fall in which an AI agent performed 80% to 90% of the intrusion work against roughly 30 targets, financial institutions among them ^[10]. Those come from a firm's own accounting and a vendor's incident report. The $893 million comes from complaint forms ^[5].

The growth claim rests on a projection. Deloitte expects US fraud losses of $40 billion by 2027, up from $12.3 billion in 2023, with AI helping push them there ^[6]. That is 3.25 times in four years, about 34% a year compounded ^[20].

The author of the Conversation piece, a finance professor who studies household finance and how people use AI to make money decisions, wrote: "I believe the most consequential AI security story right now isn't unfolding in server rooms, but at kitchen tables" ^[13]. The defense recommended there is a procedure. "Banks defend their own wire rooms with procedures, not vigilance," the author wrote ^[14]. The household version is to hang up on the suspicious call and dial a number you already know, such as your bank's fraud hotline ^[15].