AI scams surge in 2025

AI scams got industrial in 2025. The Federal Bureau of Investigation said Americans filed 22,364 complaints that referenced artificial intelligence last year, and those complaints reported about $893 million in losses. (fbi.gov) That number matters because artificial intelligence does not invent new crimes so much as it makes old ones cheaper, faster, and harder to spot. The Federal Bureau of Investigation warned in December 2024 that criminals were already using generative artificial intelligence to make fraud “more believable” and to scale it with less time and effort. (ic3.gov) A scam used to require a person who could write a convincing email, fake a voice on a call, or keep up a long text conversation. Now software can draft the message, translate it, clone the voice, and keep the conversation going across thousands of targets at once. (ic3.gov) That changes the math for criminals. If one fake message fools one person out of $500, the scam stays small. If a machine sends 100,000 versions of that message and tunes each one to sound local, urgent, and personal, the same scam becomes a mass-production business. (fbi.gov) (ic3.gov) The Federal Bureau of Investigation’s 2025 Internet Crime Report put the broader damage even higher. Across all cyber-enabled crime categories, Americans reported nearly $21 billion in losses in 2025. (fbi.gov) (ic3.gov) Cryptocurrency sat near the center of that picture. The Federal Bureau of Investigation said Americans lost about $11.4 billion to cryptocurrency-related fraud in 2025, which means more than half of all reported cyber-enabled losses involved crypto rails somewhere in the transaction. (fbi.gov) (ic3.gov) That pairing is not accidental. Artificial intelligence helps criminals win trust at the front end, and cryptocurrency helps them move money at the back end without the reversibility people expect from a credit card charge. (ic3.gov 1) (ic3.gov 2) The Federal Bureau of Investigation has also been warning that the inbox is no longer the only front door. In March 2026, the Federal Bureau of Investigation and the Cybersecurity and Infrastructure Security Agency said phishing campaigns tied to Russian intelligence services were targeting commercial messaging application accounts. (ic3.gov) That warning followed a pattern the bureau had already described in 2025. In December 2025, the Federal Bureau of Investigation said attackers often start with a plain text message and then push the conversation onto Signal, Telegram, or WhatsApp, where the exchange feels more private and more credible. (fbi.gov) (ic3.gov) The same alert said some attackers used artificial-intelligence-generated voice messages while impersonating senior United States officials. That detail shows how the scam has evolved from fake words on a screen to fake people in your pocket. (fbi.gov) The result is a kind of fraud assembly line. One system writes the lure, another copies a real person’s voice, another handles replies, and cryptocurrency gives the victim a fast way to send money before doubt catches up. (ic3.gov) (fbi.gov) The Federal Bureau of Investigation’s numbers do not mean every complaint involved a science-fiction deepfake. They mean victims themselves identified artificial intelligence somewhere in the scheme, which makes the totals an important signal of adoption even if the exact tool varied from case to case. (fbi.gov) (ic3.gov) The practical lesson is simple and old-fashioned. A message that asks you to switch apps, act immediately, scan a code, or move money into cryptocurrency deserves the same suspicion as a stranger asking for your bank card and personal identification number in a parking lot. (ic3.gov) (fbi.gov) What changed in 2025 was not human nature. What changed was that fraud got software. (fbi.gov)