AI Regulation Tracker  /  State law

New Jersey deepfake law A3540 carries criminal and civil penalties for deceptive AI media

Governor Phil Murphy signed A3540/S2544 on April 2, 2025, making it a third-degree crime to create or share deceptive AI audio or video for an unlawful purpose, and opening the door to civil damages. Anyone deploying synthetic media in New Jersey now operates under a statute that carries criminal exposure.

By Anthony Guerriero  ·  Reviewed by The Leveraged Years Editorial Desk  ·  Published July 7, 2026  ·  Last updated July 7, 2026

New Jersey has moved deepfakes out of the gray zone. On April 2, 2025, Governor Phil Murphy signed A3540, the Assembly version paired with Senate bill S2544, enacting it as P.L.2025, c.40. The law establishes civil and criminal penalties for the production and dissemination of deceptive audio or visual media, the category commonly known as deepfakes. The measure passed with bipartisan sponsorship, and it converts what had been a patchwork of privacy and fraud theories into a direct statutory offense.

The signing statement from the Governor's office is specific about intent. "I am proud to sign today's legislation and take a stand against deceptive and dangerous deepfakes," Murphy said, according to the Governor's office release. "While artificial intelligence has proven to be a powerful tool, it must be used responsibly." Attorney General Matthew Platkin framed the enforcement posture in similarly targeted terms, saying the law "is aimed at those who would misuse this powerful technology to defraud or hurt others, and provides carefully tailored criminal and civil safeguards." Both quotations come from the state's own announcement of the signing.

What the statute actually prohibits

The core criminal provision reaches a person who produces deceptive audio or visual media, or who distributes such media, for an unlawful purpose or to further another criminal act. That conduct is graded as a crime of the third degree. Under New Jersey's sentencing structure, a third-degree crime is punishable by a term of imprisonment, and the reported penalty for this offense runs up to five years, with a fine reported at up to $30,000. The gravity of the offense turns on purpose. The statute is built around the use of synthetic media to deceive, defraud, harass, or otherwise commit or advance an unlawful act, not around the mere fact that content was generated or altered by a machine.

That distinction matters for anyone who works with synthetic media lawfully. A studio that produces an AI voice for a licensed audiobook, a marketer who uses a synthetic presenter with the presenter's consent, or a newsroom that labels a reconstructed image is operating in a different lane from a person who fabricates a video to extort or defraud. The law's design puts the unlawful objective at the center of criminal exposure.

The consent and privacy dimension

Public discussion of the bill centered heavily on nonconsensual sexual imagery. Reporting on the law traces its momentum to Francesca Mani, a New Jersey high school student who was targeted by AI-generated explicit images, and much of the coverage frames the statute as a response to nonconsensual intimate deepfakes. Readers should treat the exact interaction between this statute and New Jersey's existing invasion-of-privacy offenses as a point to confirm in the enacted text, because the precise grading of sexually explicit deepfakes can depend on how the new law layers onto prior criminal statutes. What is clear and verified is that the legislation was written to give prosecutors and victims a defined tool against deceptive synthetic media used to harm.

What the law does not do

The statute is not a ban on generative media, and it carves out uses that a free-expression framework protects. Coverage of the enacted law reports exemptions for satire, parody, legitimate news reporting, teaching, and research. It also reportedly does not reach internet service providers, broadcasters, or platforms that carry such content inadvertently rather than by design. That last point is important for platform operators. The statute is aimed at creators and knowing distributors acting for an unlawful purpose, not at intermediaries that transmit content without the requisite intent. Operators should still treat that as a floor rather than a shield, because notice, moderation practices, and knowledge can all bear on whether a distributor acted knowingly.

The civil route and the litigator angle

The word civil in the law's title is not decorative. Alongside the criminal offense, the statute is described by the Governor's office as establishing civil penalties, which gives a person harmed by a deceptive deepfake a path to a remedy that does not depend on a prosecutor bringing charges. For litigators, that is the practical opening. A client who has been defamed, defrauded, extorted, or sexually exploited through synthetic media now has a New Jersey statute written for exactly that harm, rather than a strained analogy to older privacy, fraud, or right-of-publicity theories. The precise measure of civil recovery, and how it interacts with existing tort and statutory claims, should be read directly from the enacted text before a complaint is filed, because the availability and calculation of damages will shape strategy. What is verified is that the Legislature intended the law to arm victims as well as prosecutors, and counsel advising either plaintiffs or defendants should treat the statute as a live source of exposure and relief.

The broader New Jersey AI enforcement picture

A3540 does not sit alone. New Jersey's enforcement apparatus has signaled that it will apply existing law to AI-driven deception and discrimination well beyond the deepfake statute. The Attorney General's office has used the New Jersey Consumer Fraud Act as a vehicle for deceptive-practice enforcement, and that statute is significant for businesses because it can support treble damages and fee-shifting in private and state actions. Separately, the New Jersey Division on Civil Rights has issued guidance addressing how the New Jersey Law Against Discrimination applies to automated decision tools, including AI used in hiring and other employment decisions. For an operator, the practical reading is that a single AI system can implicate multiple New Jersey regimes at once. A synthetic-media campaign can raise A3540 questions, a deceptive claim about that campaign can raise Consumer Fraud Act exposure, and an AI hiring tool can raise questions under the Law Against Discrimination as interpreted by the Division on Civil Rights. The specifics of the Consumer Fraud Act and the Division on Civil Rights guidance should be confirmed against the state's own materials before relying on them in a compliance memo, and they are cited here as the surrounding enforcement context rather than as parts of A3540 itself.

Why this reaches beyond New Jersey

For firms based elsewhere, the reach is a function of where the media lands. A deepfake produced in another state but distributed into New Jersey, or aimed at a New Jersey resident, can fall within the state's interest in prosecuting the conduct. National advertisers, political operations, and platforms therefore cannot treat this as a purely local matter. State-level deepfake statutes are also accumulating across the country, and New Jersey's model, criminal penalties tied to unlawful purpose plus a civil route for those harmed and express creative-use exemptions, is one of the templates other legislatures are drawing from. A team that builds its controls to New Jersey's standard will likely find itself prepared for much of what neighboring states adopt next.

The operating standard for teams

The compliance question for anyone deploying synthetic audio or video is not whether the technology is allowed. It is. The question is whether the organization can show a lawful purpose and, where a real and identifiable person is depicted, documented consent. That means recording why a piece of synthetic media was made, who authorized any use of a real person's image or voice, and how the output is labeled or disclosed where disclosure is appropriate. For platforms, it means being able to distinguish inadvertent transmission from knowing distribution. For HR and hiring-technology teams, it means recognizing that AI tools touching New Jersey applicants sit inside the state's civil-rights framework as well. The statute rewards documentation and intent, and it penalizes deception. Building for that split is the durable response.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.