FBI's 2025 Internet Crime Report Creates Its First AI Category, Logging 22,364 Complaints and $893 Million in Losses
For the first time in 25 years, the FBI's IC3 annual report carves out a standalone artificial intelligence section, formally recognizing AI-enabled fraud as a distinct policy concern.
Overview
The FBI’s flagship cybercrime document has, for the first time in its 25-year history, treated artificial intelligence as its own category of harm. The Bureau’s 2025 Internet Crime Complaint Center (IC3) annual report, released April 7, devotes a dedicated section to AI-enabled fraud, counting 22,364 complaints and $893.3 million in adjusted losses tied to generative tools, voice clones, and synthetic media. The disclosure is modest in dollar terms compared with the $20.87 billion total that Americans reported losing to cybercrime last year, but it marks a formal turning point in how federal law enforcement catalogues the technology.
What We Know
According to the FBI, overall cybercrime losses reported to IC3 climbed 26 percent in 2025, from $16.6 billion the year prior to a record $20.87 billion, as complaint volume crossed one million for the first time. The agency now averages nearly 3,000 complaints per day, a figure cited in the IC3 report itself as it opens with a retrospective on a quarter-century of cybercrime reporting.
The newly created AI subsection documents how generative systems enter the fraud economy. The IC3 report describes AI-enabled synthetic content as “becoming increasingly difficult to detect and easier to make,” enabling criminals to mass-produce social profiles, phishing emails, and conversational scripts. The Bureau breaks the $893 million figure into specific typologies: AI-nexus investment scams accounted for $632 million in losses across 4,356 complaints; business email compromise schemes involving AI tools drove more than $30 million in losses; romance and “distress” scams using voice cloning to impersonate loved ones produced over $19 million and $5 million respectively; and AI-involved employment lures, often featuring candidates whose on-camera lip movements fail to match their voices, cost victims almost $13 million.
Despite the new category, the FBI cautions that the numbers almost certainly understate AI’s real footprint. The IC3 report notes that overall investment fraud losses exceeded $8.6 billion in 2025 — meaning the officially tagged AI share represents less than 8 percent of that category, even as investigators acknowledge many victims never identify the synthetic content they encountered.
Other headline numbers in the report reinforce the scale of cyber-enabled fraud. CyberScoop reports that investment fraud led all crime categories at $8.65 billion in losses, followed by business email compromise at $3.05 billion and tech support scams at $2.1 billion, and that Americans aged 60 and older filed 201,000 complaints accounting for roughly $7.75 billion in losses — about 37 percent of the 2025 total. The Register adds that phishing and spoofing led by complaint count at 191,561 reports, and quotes FBI Criminal and Cyber Branch Operations Director Jose Perez saying: “It has never been more important to be diligent with your cybersecurity, social media footprint, and electronic interactions.”
What We Don’t Know
The report does not assign losses to any particular model, vendor, or foreign actor. The IC3 report flags AI complaints as those where the victim’s submission “contains a reference to artificial intelligence,” a self-reporting standard that depends on whether a defrauded individual recognized synthetic content at all. That definition almost guarantees undercounting, a point acknowledged in the report’s own caveat that the $893 million total reflects only what victims were able to identify.
It also remains unclear how the new AI section will feed into enforcement priorities. The IC3 report highlights Operation Level Up, a Bureau-led initiative launched in January 2024 that the report says has reduced potential losses by more than $500 million by notifying victims of cryptocurrency investment fraud, and references an active Operation Winter Shield aimed at hardening organizational defenses. Whether the AI data will now drive a dedicated task force, distinct prosecutorial guidance, or new disclosure requirements for AI vendors has not been announced.
Analysis
Carving out an AI category is, in effect, a definitional move. Until 2025, AI misuse had been scattered across the IC3 report’s existing buckets — investment scams, BEC, confidence fraud — where it registered only as a tactic. The IC3 report now treats “AI Related” as a cross-cutting descriptor alongside long-standing ones such as “Cryptocurrency” and “Crimes Against Children,” producing the first publicly released federal dataset that quantifies how generative tools are surfacing in reported fraud.
That framing matters for policy debates already under way. AI governance proposals at the state and federal level frequently invoke deepfake impersonation, voice cloning, and synthetic identity fraud as motivating harms, but until now there was no authoritative federal baseline beyond anecdote. The IC3 figures give legislators and regulators a concrete, if deliberately conservative, number to work with — one the Bureau itself says is likely to grow as detection improves and as more victims learn to recognize the tells of synthetic content.
The gap between the $893 million AI tag and the $8.6 billion investment-fraud total also hints at where AI governance discussions are likely to focus next: not on whether AI is involved in large-scale fraud, but on how much of existing fraud it is already silently powering. The Register notes that the report explicitly warns cyber threats will continue evolving “as the world embraces emerging technologies such as artificial intelligence,” a framing that positions AI not as a discrete threat vector but as a diffuse accelerant across the entire cybercrime economy.