Artificial intelligence has changed how fraud is committed and how it is investigated. It has not changed the statutes. Fraud involving voice cloning, deepfake video, chatbots, or AI-generated records is charged as wire fraud, health care fraud, identity theft, and money laundering, and the government increasingly argues that AI should add time at sentencing. Wire fraud alone carries up to 20 years per count.
There is no general federal crime of "AI fraud." Prosecutors use the statutes they already have. 18 U.S.C. § 1343 covers any scheme to defraud that uses interstate wires, and the statute expressly reaches transmitted "sounds" and "pictures," so a cloned voice or a deepfake video call fits without any new law. Conspiracy under § 1349, money laundering under § 1956, and aggravated identity theft under § 1028A follow close behind.
The only federal offense whose definition names artificial intelligence is the TAKE IT DOWN Act, 47 U.S.C. § 223(h), which covers publishing intimate digital forgeries. Its first conviction came in 2026. Congress is considering more: the AI Fraud Accountability Act would create a standalone crime for using a realistic digital impersonation to defraud. Until something passes, AI cases are ordinary fraud cases with a technology story attached.
The headlines say voice clones are everywhere. The charging documents mostly do not. Our review of the public record found no federal indictment that alleges AI voice cloning in a grandparent, family-emergency, or executive-impersonation scam, even though the FBI warns about the technique. Many of those stories rest on what a victim believes she heard.
Where DOJ does put AI on the record, it matters. In 2026, four Minnesota defendants pleaded guilty to wire fraud after using ChatGPT to fabricate Medicaid records requested by insurers, and DOJ called AI use "a burgeoning trend" in health care fraud. A federal romance fraud indictment in Ohio alleges the use of AI-driven video personas. When the government says AI was used, the claim has to be proven, and a claim that is not proven should not drive a sentence.
The organizers of large AI-enabled schemes are often overseas and rarely in custody. The people agents reach first are here: couriers who picked up cash, account holders whose accounts received transfers, call-center workers, lead-list vendors, business owners whose marketing or calling tools made false claims, and health care providers whose records do not match the care delivered. Many of them did not know they were part of a fraud. Knowledge is the element that decides most of these cases, and what a person says in the first conversation with agents often decides knowledge.
Impersonation and elder fraud. Voice cloning, deepfake video calls, and AI-scripted calls targeting older adults. Exposure includes up to 10 additional years under 18 U.S.C. § 2326 when schemes target people over 55.
Health care documentation. AI scribes and templated notes are lawful. Records that describe care that never happened, or that are created after an auditor asks for them, create exposure under § 1347, § 1035, and the records-falsification statute, § 1519, which carries 20 years.
AI calling and marketing. The FCC treats AI-generated voices as robocalls under the TCPA, and Florida's Telephone Solicitation Act adds its own consent rules. Those problems are mostly civil. They become criminal when an AI agent's statements are false and material to why the customer paid.
Synthetic identities and deepfaked IDs. AI-generated identification used to open bank or crypto accounts supports bank fraud under § 1344 and identification document charges under § 1028.
Not automatically. The Sentencing Guidelines contain no AI enhancement. The Justice Department asked the Sentencing Commission for one in 2024, and the 2026 amendments did not adopt it. Prosecutors instead argue AI through sophisticated means, mass-marketing, vulnerable victim, and special skill enhancements. DOJ's own letter to the Commission acknowledged that sophisticated means does not reach AI use that is not itself sophisticated. Each of these enhancements is a fact question, and each can be fought.
We start before charges. In AI cases the government's first reading of logs, metadata, and vendor records is often wrong and hard to unwind once it is in a charging memo. Pre-indictment defense puts the real workflow and the client's actual role in front of the prosecutor early.
We make the government prove the technology. Victim perception is not forensic evidence. We test whether AI was actually used, who used it, and what it did, before it becomes an enhancement.
We fight knowledge and role. Couriers, account holders, and employees are often charged with the full scheme under conspiracy law. We hold the case to what the client knew and could foresee, which also drives the loss amount.
We protect the record. We preserve prompts, logs, and communications that show good faith, and we keep clients from creating obstruction exposure after contact.
Fraud committed with AI is charged under existing federal statutes, mainly wire fraud under 18 U.S.C. § 1343, which carries up to 20 years per count and covers transmitted sounds and pictures. Conspiracy, money laundering, and aggravated identity theft are frequently added. The only federal offense that names AI is the TAKE IT DOWN Act, which covers intimate digital forgeries.
Not to convict. AI is not an element of wire fraud or conspiracy. It matters at sentencing, where the government may argue it supports sophisticated means or other enhancements. At that stage the government must prove the facts, and a victim's belief that a voice sounded cloned is not the same as forensic proof.
Yes. Account holders and couriers are commonly charged with conspiracy under 18 U.S.C. § 1349 and money laundering under § 1956. The central question is whether you knew the money came from fraud. Do not explain yourself to agents without counsel, because those statements are usually the government's proof of knowledge.
There is no AI enhancement in the Sentencing Guidelines. DOJ asked for one in 2024 and the 2026 amendments did not adopt it. Prosecutors argue AI through sophisticated means, mass-marketing, vulnerable victim, and special skill enhancements, each of which depends on the facts and can be contested.
No. AI documentation tools are lawful. The exposure is a record that describes care that did not happen, or a note signed without review. Creating or completing records after an auditor or payer asks for them can add obstruction exposure under 18 U.S.C. § 1519, which carries up to 20 years.
Usually the exposure is civil, under the TCPA and Florida's Telephone Solicitation Act. It becomes criminal when the agent makes material misrepresentations to obtain money, such as posing as a bank, agency, or licensed professional, or lying about price or product. That can be wire fraud.
Be polite, do not answer substantive questions, and ask for the agent's contact information. Do not delete apps, prompts, messages, or accounts, and do not move money. Contact a federal criminal defense attorney, who can find out whether you are a target, subject, or witness and handle any contact with the prosecutor.
Wire fraud under 18 U.S.C. § 1343 is one of the most frequently charged federal crimes, and one of the most flexible tools in a federal prosecutor's toolkit. Any scheme to defraud that uses a wire communication crosses into federal jurisdiction. The penalty is up to 20 years per count.
Federal aggravated identity theft under 18 U.S.C. § 1028A carries a mandatory two-year consecutive sentence that stacks on top of the underlying offense, with no judicial discretion to run it concurrently. When paired with wire fraud or healthcare fraud, a single case can produce 15 to 20 years in mandatory federal time.
Federal healthcare fraud cases are built from claims data before anyone is interviewed. By the time HHS-OIG or FBI agents knock, the government usually has months or years of billing analysis, and often a cooperating insider. The defense has to start where the government started: the data, the medical records that support or undercut medical necessity, and the financial relationships behind the referrals.
Federal cybercrime charges under the Computer Fraud and Abuse Act carry up to 10–20 years per count. If the FBI has executed a search warrant or reached out to you, the investigation is already advanced. You need defense counsel who understands both the law and the technology.
Federal money laundering charges under 18 U.S.C. § 1956 carry up to 20 years per count. When federal prosecutors target financial transactions in South Florida, you need defense counsel who knows exactly how they build these cases, and how to dismantle them.
A target letter from a United States Attorney's Office means the government has already decided you are someone it wants to prosecute. The investigation is not beginning. It is ending.