Does Using AI Make Your Federal Sentence Longer? What the Guidelines Say, What DOJ Wants, and How to Fight It

If you are facing federal fraud charges and the government says you used artificial intelligence, you will hear some version of this at sentencing: the AI made the scheme more sophisticated, reached more victims, and was harder to detect, so the sentence should be longer.

The short answer to whether that works is this. There is no AI enhancement in the Sentencing Guidelines. The Justice Department asked for one and did not get it. So prosecutors argue AI through enhancements written for other things, and each of those can be fought. Here is how.

There is no AI enhancement in the Sentencing Guidelines. Prosecutors argue AI through enhancements written for other things, and each of those can be fought.

Key Takeaways

• No guideline adds levels for using AI. The November 1, 2026 amendments did not create one, and the 2026-2027 priorities do not list one.
• DOJ asked the Sentencing Commission in July 2024 for an AI enhancement, conceding sophisticated means under USSG § 2B1.1(b)(10) does not cover ordinary AI use.
• Prosecutors argue AI through sophisticated means, mass-marketing under § 2B1.1(b)(2), vulnerable victim under § 3A1.1(b), and special skill under § 3B1.3.
• In the Eleventh Circuit, sophisticated means looks at the whole scheme, so the fight is over what the defendant did, not the software.
• Wire fraud under 18 U.S.C. § 1343 carries the same 20-year maximum whether or not AI was used. The difference is in the guideline range.

What DOJ Said It Would Do

In February 2024, the Deputy Attorney General announced that prosecutors would seek stiffer sentences for offenses made significantly more dangerous by the misuse of AI (remarks). She compared it to the way firearms increase sentences, and said DOJ would seek reforms if existing enhancements fell short.

Five months later, the Criminal Division followed through. Its July 15, 2024 letter to the Sentencing Commission recommended a Chapter Three enhancement for any defendant who used AI during an offense, in preparing for it, or in trying to avoid apprehension. The letter is candid about why. DOJ wrote that the new enhancement "would differ from the sophisticated-means enhancement" because it would apply "not just those where the use of artificial intelligence was sophisticated," and would be "distinct from the special-skill enhancement" because it would apply "regardless of whether special skill was required to use the AI."

"...would differ from the sophisticated-means enhancement... not just those where the use of artificial intelligence was sophisticated..."— DOJ Criminal Division, letter to the U.S. Sentencing Commission, July 15, 2024

Read that closely. The Department itself told the Commission that the existing enhancements do not automatically reach ordinary AI use. That is a useful sentence to put in front of a sentencing judge.

What the Sentencing Commission Actually Did

Nothing, on AI. The 2026 amendments, effective November 1, 2026, address simplification, inflation adjustments to the loss table, drug offenses, multiple counts, and sentencing options. None adds an AI enhancement. The Commission's final priorities for 2026-2027, published September 3, 2026, focus on a broad evaluation of the guidelines and do not list AI.

So the policy lives in charging decisions and sentencing memos, not in the Guidelines Manual. That matters. When a prosecutor argues AI deserves more punishment, the argument has to fit an existing guideline or become a request for an upward variance under 18 U.S.C. § 3553(a), which the defense can answer on the merits.

The Four Enhancements Prosecutors Use Instead

With no AI guideline to cite, the government reaches for four that already exist.

Sophisticated means, USSG § 2B1.1(b)(10)(C): two levels, with a floor of level 12, for especially complex or especially intricate conduct in executing or concealing the offense. Offshore conduct under (b)(10)(B) applies regardless.

Victims and mass-marketing, § 2B1.1(b)(2): two levels for ten or more victims or for mass-marketing, and more for substantial financial hardship. AI robocalls and chatbot campaigns fit mass-marketing on their face.

Vulnerable victim, USSG § 3A1.1(b): two levels if the defendant knew or should have known a victim was unusually vulnerable, and two more for a large number of vulnerable victims. Elder-targeted voice scams draw this argument.

Abuse of trust or special skill, § 3B1.3: two levels. Special skill means a skill not possessed by the general public, usually requiring substantial education, training, or licensing.

How to Fight Each One

Sophisticated means is the main battleground. The Eleventh Circuit looks at the totality of the scheme, not whether each step was clever. That cuts both ways. The government will argue that deepfake video or voice cloning made the whole scheme intricate. The defense answer is that typing a prompt into a consumer app is not especially complex, and that the scheme itself, a lie told over the phone or in an email, is the same fraud people have run for decades. DOJ's own letter supports that position.

Special skill is the weakest argument for the government when the AI was a consumer product. Using a subscription voice tool or a chatbot takes no training or licensing. Building or fine-tuning a model may be different, and that line will be litigated.

Mass-marketing and vulnerable victim turn on facts, not technology. The question is how many people were actually targeted and whether the defendant knew their vulnerability. A courier who never spoke to a victim is in a very different position from the person who scripted the calls. Role matters, and so does relevant conduct: a defendant is responsible for reasonably foreseeable conduct within the scope of what he agreed to, not everything the organization did.

A victim's belief that a voice sounded cloned is not forensic evidence. Assumptions about technology often do not survive a hearing.

Critical Mistakes That Increase Exposure

The most expensive errors in these cases are made by the defendant, often before a lawyer is involved.

• Describing your own tools to agents. Statements about what software could do become the government's sophisticated-means proof.
• Deleting apps, prompts, or AI accounts after contact. That invites an obstruction increase under § 3C1.1 and can cost acceptance of responsibility.
• Signing a plea agreement that stipulates to sophisticated means or special skill without a fight. Many enhancements are agreed to, not proven.
• Waiting until the presentence report to challenge the AI narrative. By then it is in the probation officer's draft.

Strategic Defense Approach

The best sentencing work in an AI case is pre-indictment defense. If you are the subject of a federal investigation, received a target letter, or are considering a proffer session, counsel should be shaping the factual record about what technology was used, by whom, and to what effect. That record drives the guideline calculation later, which is why federal investigation defense and sentencing strategy are the same project.

Waiting until the presentence report to challenge the AI narrative is a mistake. By then it is in the probation officer's draft.

After charges, the fight is in plea negotiation and in PSR objections. A federal criminal defense attorney should negotiate the guideline stipulations carefully, contest AI-based enhancements that are not proven, and present 3553(a) arguments that focus on what the defendant actually did. A white collar defense attorney should also track the loss amount dispute closely, because the inflation-adjusted loss table that takes effect November 1, 2026 can change the baseline for anyone sentenced on or after that date.

For how the new table plays out in healthcare fraud cases, see what the 2026 loss table means for Florida telemedicine sentencings.

Why This Matters Now

DOJ has already called AI use a trend in health care fraud (Minnesota Medicaid pleas), and prosecutors are starting to put AI into sentencing memos even where it is not in the indictment. With no guideline written for it, the arguments are being made case by case. Defendants in the Southern District of Florida and the Middle District of Florida who get this right early can keep a technology story from adding years to a sentence.

Common Questions

Facing Federal Fraud Charges Where the Government Claims AI Was Used?

Sentencing exposure in these cases is shaped early, often before charges are filed. As a Boca Raton federal criminal lawyer and federal wire fraud attorney, Aaron M. Cohen represents people in federal investigations, plea negotiations, and sentencings in Florida and nationwide. Consultations are confidential. Call 561.542.5494 or request a consultation at amcdefenselaw.com.

In an AI case, the factual record about what technology was used, by whom, and to what effect drives the guideline calculation. Counsel should be shaping it before charges are filed.

If you or your loved ones have been arrested or are under federal investigation, call Aaron M. Cohen, 24 hours a day to get help.

About the author. Aaron M. Cohen is the founder and principal attorney of AMC Defense Law, a criminal defense firm in Boca Raton, Florida, with more than 30 years of experience in state and federal courts. He is admitted to practice in Florida, New York, New Jersey, and the District of Columbia, and before the United States District Courts for the Southern and Middle Districts of Florida, the Southern, Eastern, and Western Districts of New York, the District of New Jersey, and the Northern District of Texas. He appears pro hac vice in other federal districts nationwide. The firm represents clients in federal investigations and prosecutions involving healthcare fraud, Anti-Kickback Statute matters, DME and telemedicine fraud, peptide and compounded-drug enforcement, controlled-substance and drug conspiracy cases, financial crimes, and complex federal litigation, in Florida and nationwide.

This article is for general informational purposes only and does not constitute legal advice. Reading it does not create an attorney-client relationship. Every case turns on its own facts, and guideline calculations depend on the Manual in effect and the specific conduct. Past results do not guarantee a similar outcome. Consult a qualified attorney about your specific situation.