FBI: US Cybercrime Losses Rise to $21 Billion, With AI-Enabled Schemes Counted Separately

The FBI estimates cybercrime losses in the US for 2025 at almost $21 billion, a record high in the history of IC3 reports. For the first time, the document treats crimes in which attackers used AI tools as a separate category.

Record losses in the US

According to the Internet Crime Complaint Center, Americans filed 1,008,597 complaints about internet crime in 2025, compared with 859,532 a year earlier. Total confirmed losses came to $20.877 billion, up about 26% year over year.

Phishing and spoofing, extortion and investment schemes topped the list by number of complaints. For the FBI this is more than statistics: the bureau receives almost 3,000 complaints a day, and online fraud has long since grown into a large-scale economy.

The report also includes, for the first time, a section on crimes involving AI. In 2025, IC3 received 22,364 such complaints, with total losses of $893.3 million. It is an important shift: until recently, AI in cybercrime was described as an amplifier, and now it is treated as a separate dimension of risk. This is not one type of attack but a set of techniques that make old schemes more convincing, cheaper and faster to launch.

Who lost the most

Criminals took the most money from crypto schemes. According to the FBI, crypto-related complaints caused more than $11 billion in losses across 181,565 reports. Investment fraud overall remained the main driver of losses: it accounted for more than $8 billion, almost half of all losses from online fraud. To counter it, the FBI is promoting Operation Level Up: by December 2025, the initiative had helped notify 8,103 potential victims, 77% of whom did not realize they were already caught in a scheme.

• Crypto schemes: 181,565 complaints and more than $11 billion in losses
• Investment fraud: more than $8 billion, almost half of all losses from online fraud
• People over 60: $7.748 billion in losses and 201,266 complaints
• Phishing and spoofing: the most common category, about 191 thousand complaints
• Extortion: about 89 thousand complaints, followed by investment schemes and business email compromise

The 60+ age group stands out: its members reported $7.748 billion in losses, the heaviest toll among all civilian groups. In critical infrastructure, healthcare, manufacturing, financial services, IT and government organizations were hit most often. In practice, this means attacks hurt not only private wallets but also sectors where a digital failure quickly turns into an operational and reputational crisis.

How AI helps scammers

In the new AI category, the FBI lists fairly down-to-earth scenarios: voice clones for calls posing as relatives or bosses, fake social media profiles, forged documents, and deepfake videos of public figures. Generative models help write convincing emails for business email compromise schemes, run thousands of slightly different conversations with potential victims, and automate the part of social engineering that once required a lot of manual work.

Investment schemes were the costliest AI subcategory: complaints showing signs of AI use caused more than $632 million in losses. Next come business email compromise with losses above $30 million, tech support scams with $19.5 million, romance scams with $19 million, and employment fraud with almost $13 million. In effect, AI has not created a new genre of fraud. It has made trust cheaper to manufacture: criminals can now fake a voice, assemble a plausible backstory and scale an attack without a large team.

What this means

The FBI report's main takeaway is simple: AI has become part of fraudsters' standard toolkit rather than a rare curiosity. For users, that means verifying identity and payment requests outside the chat or call. For companies, it means revisiting how they approve transfers, handle hiring and manage corporate email.