AI Evidence and Deepfakes in Massachusetts Criminal Cases: Authentication, Daubert-Lanigan, and the Defense Playbook

By Attorney Joe Serpa | Georgetown University Law Center | 30 Years Massachusetts Criminal Defense

Artificial intelligence has entered Massachusetts criminal courtrooms from two directions at once. Digital evidence offered against a defendant, a screenshot, a voice recording, a video, can now be fabricated with consumer software, which puts authenticity in play in cases where it was once assumed. And Massachusetts now prosecutes the creation and distribution of AI-generated intimate images under the criminal harassment statute as amended by Chapter 118 of the Acts of 2024. Both directions run through the same evidence law, and that law is regularly misstated online, sometimes in ways that would sink a defense built on it. This post sets out what the Massachusetts Guide to Evidence provides, where deepfake fights are really won, and how the analysis changes by charge, by court, and by defendant.

What the Guide to Evidence Says About Digital Evidence

Authentication is governed by Mass. G. Evid. § 901, and the standard is deliberately low. The proponent must produce evidence sufficient to support a finding that the item is what the proponent claims it is. Authentication is a preliminary question of conditional relevance under § 104(b), which means the judge screens whether a reasonable jury could find the item genuine, and the jury then decides how much weight it deserves. Commonwealth v. Meola, 95 Mass. App. Ct. 303, 307 (2019), and Commonwealth v. Sargent, 98 Mass. App. Ct. 27, 30 (2020).

For electronic and digital communications, the Guide permits authentication by confirming circumstances that would allow a reasonable fact finder to conclude the evidence is what its proponent claims, and it states expressly that neither expert testimony nor exclusive access is necessary to authenticate the source. The confirming-circumstances approach comes from the case law on emails and social media authorship, including Commonwealth v. Purdy, 459 Mass. 442 (2011), where the surrounding facts, the account, the contents, the context, supplied the foundation. Digital content is not on the § 902 list of self-authenticating evidence, so some foundation is always required, but the foundation can be circumstantial and modest.

Section 1119 of the Guide, often misdescribed online as an AI authentication code, does something different. It defines digital evidence and personal electronic devices and directs judges to facilitate the admission and preservation of digital evidence in courtroom and virtual proceedings, including for self-represented parties. Its practical importance for the defense is preservation. When a complainant shows a judge a message thread on a phone at a hearing, § 1119 is the basis for insisting that the digital original be captured for the record, because the version preserved is the version an appellate court will see, and the version a forensic expert can later examine.

The firm’s comprehensive guide to admitting and objecting to digital evidence in Massachusetts criminal cases addresses authentication together with authorship, screenshots, phone extractions, hearsay, completeness, suppression, and trial objections.

Two consequences follow from this framework, and they cut in opposite directions. Because the bar is low, the defense cannot expect a judge to exclude a screenshot merely because screenshots can be faked. And because the bar is low, a conviction can rest on digital evidence that no expert ever validated, which is exactly why the defense work described below has to be done rather than assumed.

Deepfakes and the Authenticity Fight

Massachusetts has no AI-specific authentication rule, and the 2026 edition of the Guide, adopted effective February 1, 2026, added none. The federal rules committees are further along than the states. A proposed Federal Rule of Evidence 707 would apply expert-reliability standards to machine-generated evidence, and it was published for public comment in a period that closed on February 16, 2026. The Advisory Committee on Evidence Rules has also drafted a possible new Rule 901(c) that would create a burden-shifting procedure for claims that audio or visual evidence was generated or altered by AI, but that draft was held for further study and has not been published for comment. Neither proposal binds Massachusetts courts. Together they map where the law is heading, and they supply the vocabulary a Massachusetts judge will recognize when the defense raises a fabrication claim under the existing rules.

Under current Massachusetts law, a deepfake challenge is an authenticity challenge under § 901 and § 104(b), pressed through the confirming circumstances. The defense does not win by saying the video could be fake. The defense wins by showing the specific reasons to doubt that this video is what the Commonwealth claims. Those reasons include metadata that does not match the alleged capture date or device, the absence of an original file, a provenance chain that begins with a forward from an interested party rather than an extraction from a device, compression and artifact patterns inconsistent with the claimed source, and the availability and motive of a person to fabricate. Where the item is the core of the case, the defense retains a digital forensics expert to examine the file itself, and the expert’s findings feed both the § 104(b) argument to the judge and the weight argument to the jury.

Daubert-Lanigan: Challenging the Tools Themselves

The second front is the reliability of the forensic methods, on either side. Massachusetts governs expert methodology under Commonwealth v. Lanigan, 419 Mass. 15 (1994), which adopted the core reasoning of Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993). The judge acts as gatekeeper and asks whether the methodology is reliable, looking to testing, peer review, error rates, standards, and general acceptance.

Deepfake detection is a young science with published error rates that vary sharply by generation tool, compression level, and training data. When the Commonwealth offers an examiner to testify that a recording is authentic, or that it is synthetic, the defense files a motion in limine and demands a Daubert-Lanigan hearing on the detection methodology. The same challenge runs against attribution testimony, the claim that a particular device or account created the content, when it rests on proprietary extraction tools whose validation the defense has never been allowed to test. Discovery of the tool, its version, its validation studies, and the examiner’s protocol precedes the hearing, and the absence of that documentation is itself the argument.

The Prosecution Side: Massachusetts Deepfake Charges

The second direction is the defendant charged with creating or distributing AI-generated intimate imagery. Chapter 118 of the Acts of 2024, approved June 20, 2024, closed the gap that had left Massachusetts without a dedicated nonconsensual intimate image crime. The act rewrote the criminal harassment statute, G.L. c. 265, § 43A, which now expressly covers visual material produced by digitization, a term the statute defines to cover computer-generated images that would falsely appear to a reasonable person to be an authentic representation of the person depicted. A first offense under the image provision carries up to two and a half years in a house of correction and a fine of up to $10,000, and the same act created a diversion path for minors through a new G.L. c. 272, § 29D. At the federal level, the TAKE IT DOWN Act criminalizes knowingly publishing nonconsensual intimate imagery, defined to include realistic computer-generated images depicting identifiable real people, and requires platforms to remove such content within 48 hours of a valid removal request. For the full statutory treatment, see AI-Generated Imagery and Deepfake Criminal Defense in Massachusetts and Massachusetts Revenge Porn Law under Chapter 118 and the TAKE IT DOWN Act.

In these prosecutions the contested proof is usually identity twice over. The Commonwealth must show that the image depicts the complainant, and it must show that the defendant is the person who created or distributed it. On the first, AI generation tools produce composites, and where the Commonwealth’s proof of depiction rests on the complainant’s own recognition, an independent forensic comparison is a legitimate line of defense. On the second, the confirming-circumstances doctrine that helps prosecutors authenticate messages cuts both ways. Possession of a device is not authorship, accounts are shared and compromised, and files arrive on phones by AirDrop, group chat, and forwarded thread without any act of creation by the phone’s owner.

The Defense Playbook

Across both postures, the pretrial work falls into five lines, each concrete.

Attribution. The first question is whether the Commonwealth can tie the content to the defendant’s hands rather than the defendant’s device. Account access records, device sharing, the interval between creation and the defendant’s demonstrable use of the device, and the alternative senders the forensic image itself reveals all bear on that question.

Provenance and metadata. Creation dates, modification history, device identifiers, and geolocation are checked against the Commonwealth’s narrative. So is the difference between an original file and a re-saved, re-compressed, or screenshotted copy, because copying strips the metadata that would have answered the question.

Chain of custody. In Massachusetts, gaps in the chain ordinarily go to weight rather than admissibility, so the defense uses them at both stages. Severe gaps join the § 104(b) confirming-circumstances argument to the judge, and lesser gaps become cross-examination at trial.

Suppression. Digital evidence arrives through device searches, and device searches are where constitutional challenges live. A motion to suppress tests the particularity of the warrant, the scope of the extraction against the scope authorized, and the treatment of passcodes and biometrics. Our overview of phone and computer searches in Massachusetts and our digital search FAQ lay out the constitutional framework. See also Digital Search Warrants in Massachusetts and Can You Refuse to Give Police Your iPhone Passcode?

Lanigan motions. These run against detection tools, attribution tools, and any examiner whose method cannot show testing, error rates, and validation. They are filed early enough that exclusion reshapes the Commonwealth’s case rather than trims it.

How the Stakes Change by Defendant

Students. Campus deepfake allegations produce two proceedings at once. The criminal case runs alongside the university proceeding, which applies a preponderance standard and does not wait for the criminal outcome. A student charged criminally should not give a university interview without counsel coordinating both tracks. See College and University Student Criminal Defense and the student criminal charge FAQ for families sorting out the first week.

Licensed professionals. An arraignment on an image-based charge can trigger board reporting obligations before any adjudication, which raises the value of resolving the case at the earliest stage of the criminal process available, including the clerk-magistrate hearing where the charge arrives by summons. Our clerk-magistrate hearing FAQ explains how that stage works. See Criminal Defense for Licensed Professionals.

Non-citizens. Offenses involving intimate images can carry serious immigration consequences, and the arraignment record alone is visible to immigration authorities. Defense strategy is chosen with the immigration analysis in hand, not after the plea. See Immigration Consequences of Massachusetts Criminal Charges.

Domestic and harassment contexts. Image allegations frequently arrive alongside 209A or 258E proceedings, where the same digital evidence appears first in a civil hearing with looser process. What is said and admitted there follows the criminal case, so the two proceedings are defended as one matter. See Defending 209A and 258E Violations.

Courts Where Serpa Law Office Defends These Cases

Digital evidence and deepfake cases are charged where the conduct occurred, which for distribution offenses is typically the court covering where the defendant lives or where the images were received. Serpa Law Office defends these cases, and the clerk-magistrate hearings that often begin them, in the courts listed below.

• Boston Municipal Court (all divisions, including Central, Brighton, Charlestown, Dorchester, East Boston, Roxbury, South Boston, and West Roxbury)
• Cambridge District Court
• Somerville District Court
• Newton District Court
• Brookline District Court
• Waltham District Court
• Malden District Court
• Woburn District Court
• Concord District Court
• Framingham District Court
• Dedham District Court
• Quincy District Court
• Hingham District Court

For the full list, see Courts We Serve Across Greater Boston.

Key Takeaways

• Authentication of digital evidence in Massachusetts runs through Mass. G. Evid. § 901. The proponent needs only evidence sufficient to support a finding, the judge decides the question preliminarily under § 104(b), confirming circumstances suffice for digital communications, and neither expert testimony nor exclusive access is required. Commonwealth v. Purdy, 459 Mass. 442 (2011), and Commonwealth v. Meola, 95 Mass. App. Ct. 303 (2019).
• Section 1119 of the Guide governs the admission and preservation of digital evidence in courtroom and virtual proceedings. It is not an AI authentication code, and Massachusetts has no AI-specific evidence rule as of the 2026 edition of the Guide. The proposed federal rule on machine-generated evidence and the draft deepfake procedure still under committee study show where the law is heading.
• A deepfake challenge is an authenticity challenge built on specifics. Metadata, provenance, the missing original, and motive to fabricate carry the argument, supported by an independent forensic examination.
• The reliability of detection and attribution tools is tested under Commonwealth v. Lanigan, 419 Mass. 15 (1994), through a pretrial motion in limine and a demand for the tool’s validation record.
• Massachusetts prosecutes AI-generated intimate imagery under the criminal harassment statute, G.L. c. 265, § 43A, as amended by Chapter 118 of the Acts of 2024 to cover digitized and computer-generated images, alongside the federal TAKE IT DOWN Act. Identity, of the person depicted and of the person who created or sent the file, is where these cases are fought.
• Chain-of-custody gaps ordinarily go to weight rather than admissibility in Massachusetts, so they are used both in the § 104(b) argument and in cross-examination.
• For students, licensed professionals, and non-citizens, the arraignment itself carries separate consequences, which makes the earliest stage of the case, often a clerk-magistrate hearing, the most valuable one.

Serpa Law Office has defended Massachusetts criminal cases for 30 years and brings that experience to the newest category of evidence in the courtroom. Contact Serpa Law Office at 617.936.0201 for a confidential consultation.

Related Serpa Law Office Resources

• AI-Generated Imagery and Deepfake Criminal Defense in Massachusetts
• Massachusetts Revenge Porn Law under Chapter 118 and the TAKE IT DOWN Act
• Digital Search Warrants in Massachusetts for Phones and Computers
• Can You Refuse to Give Police Your iPhone Passcode in Massachusetts?
• Illegal Searches and Seizures in Massachusetts
• Defending 209A and 258E Restraining Order Violations
• College and University Student Criminal Defense
• Criminal Defense for Licensed Professionals
• Immigration Consequences of Massachusetts Criminal Charges
• Clerk-Magistrate Hearings in MassachusettsCourts We Serve Across Greater Boston