Artificial Intelligence in Maryland Courts

Half-Baked Motion to Compel Was Not Prompt, Not Ripe, Not Complete, and Not Likely to Succeed

August 22, 2025

What is a “Keyword Warrant”?

September 1, 2025

Published by Michael Berman at August 24, 2025

Mooney v. State, 487 Md. 701 (2024).

“Video footage, like social media evidence, is susceptible to alteration, and the increased availability of new technology, particularly the advent of image-generating artificial intelligence, may present unique challenges in authenticating videos and photographs. As we have noted, “(p)hotographic manipulation, alterations and fabrications are nothing new, nor are such changes unique to digital imaging, although it might be easier in this digital age.” Id. at 734-35 (citation omitted).

…

“I write separately to express one additional thought. Justice Gould begins his thoughtful dissent with a reference to “the age of artificial intelligence” and the growing “risk of fabricated or altered evidence.” The evidentiary concerns associated with the growth and proliferation of artificial intelligence, especially generative artificial intelligence, are real and pressing. Courts should be alert to claims that evidence has been altered by the use of artificial intelligence, and artificial intelligence technology may ultimately require us to adjust our rules and procedures for authenticating electronic evidence. But the record in this case does not contain any hint that artificial intelligence may have played a role, nor was there any suggestion that the video may have been altered in any way. We can expect to need to tackle issues associated with artificial intelligence soon, but this is not the case.” Id. at 735 (Fader, C.J., concurring).

…

“I respectfully dissent to the Majority’s well-written and thorough opinion. In the age of artificial intelligence, the risk of fabricated or altered evidence has never been greater, and that risk will only increase as technology advances.” Id. at 736 (Gould, J., dissenting).

…

For more on Mooney, please see Can a witness authenticate a video if the video contains images that the witness did not see?

Brasse v. State, 264 Md. App. 740 (2025).

“In 2019, this Court noted that the computer-generated language in CR § 11-208 was added to address technology that had advanced to allow computer-generated images referred to as “deepfakes.” In re S.K., 466 Md. at 56 n.22. This process is more advanced than systems like Photoshop; “deepfakes are the product of artificial intelligence, relying on neural networks to generate ‘realistic impersonations out of digital whole cloth.'” See Emily Pascale, Deeply Dehumanizing, Degrading, and Violating: Deepfake Pornography and the Path to Legal Recourse, 73 Syracuse L. Rev. 335, 337 (2023) (footnote omitted). “The process involves inputting hours of video footage of a specific individual to train the neural network to ‘understand’ the nuances of that person’s face.” Id. (footnote omitted). “Once the network is trained, it can digitally graft one person’s face onto another person’s body.” Id. (footnote omitted).”

…

“Generative artificial intelligence models are used to create photorealistic images that are virtually indistinguishable from real images. Internet Watch Found., How AI is being abused to create child sexual abuse imagery 7 (2023). “In short, you type in what you want to see; the software generates the image.” Id. These systems are “trained by a process called deep learning, which is a type of machine learning that is loosely modelled on the human brain — using artificial neural networks. These deep learning systems are trained on huge datasets scraped from the internet.” Id. at 10. Diffusion models are trained via “a vast dataset of images that are scraped from the internet and then labelled with descriptive words or phrases — the type of text that will later be used for prompting new generations.” Id. at 12. These models are “large-scale” with the ability to generate “detailed, high-quality images.” Id. at 13.

The most realistic AI child sexual abuse material is created by fine-tuned models “that are well-known among AI CSAM communities – reputed for enabling realistic generation of certain CSAM scenarios, children, or child characteristics.” Id. at 22. These models often use “datasets that feature a particular child individual — usually a known victim of child sexual abuse, or a famous child. This is because, for both these categories, large enough image sets exist to train AI models.” Id.

To the extent that images are created by artificial intelligence using an actual child’s face, the same concerns involved with morphing exist. Such images, to the extent that they are indistinguishable from an actual and identifiable child, implicate the interests of an actual child, who is subject to the type of reputational and emotional harm that justifies their exclusion from protection under the First Amendment.

It may be possible, however, that an image could be created virtually without reference to a specific person, but the image nevertheless could look indistinguishable from an actual minor that the State can identify at trial. In that circumstance, there is a concern that the statute covers virtual child pornography that Free Speech Coalition held was protected by the First Amendment.”

Johnson v. State, 2025 Md. App. Lexis 660 (Aug. 6, 2025)(unreported).

“The State’s discovery violation carries greater significance, and inflicted greater prejudice on Mr. Johnson, in the context of the FRT technology that the State admits to using here. There is ample reason to question the reliability of evidence generated by FRT and artificial intelligence (“AI”) more broadly, and we would send exactly the wrong message if we allowed the State to rely on an FRT-generated identification without accountability. FRT has produced unreliable results in multiple instances across the country, including here in Maryland. In 2022, Alonzo Sawyer was arrested after a facial recognition program identified him as the perpetrator of an assault. Khari Johnson, Face Recognition Software Led to His Arrest. It Was Dead Wrong, WIRED, (Feb. 28, 2023, 7:00 AM), https://www.wired.com/story/face-recognition-software-led-to-his-arrest-it-was-dead-wrong/, archived at https://perma.cc/Q7QW-RPX5 . The true perpetrator was seven inches shorter and twenty years younger than Mr. Sawyer, and Mr. Sawyer spent nine days in jail before the error was fixed. Id. According to the Innocence Project, at least six others (as of February 2024) had been accused of crimes wrongfully due to misidentification through FRT. Alyxaundria Sanford, Artificial Intelligence Is Putting Innocent People at Risk of Being Incarcerated, Innocence Project, (Feb. 14, 2024), https://innocenceproject.org/news/artificial-intelligence-is-putting-innocent-people-at-risk-of-being-incarcerated/, archived at https://perma.cc/3WUB-HN8A . Our courts must, and will, recognize the power and opportunity AI tools can offer. But the very real prospect that AI could hallucinate evidence, as it does text and citations when it can’t find an answer, see Joe Patrice, Trial Court Decides Case Based On AI-Hallucinated Caselaw, Above the Law (July 1, 2025, 12:48 PM), https://abovethelaw.com/2025/07/trial-court-decides-case-based-on-ai-hallucinated-caselaw/, archived at https://perma.cc/PC6C-WX5Y, places all the greater imperative on allowing FRT- and AI-generated evidence to be tested appropriately, and we cannot give the State a pass here where it failed even to identify the technology it used to identify the suspect it pursued and prosecuted.”

For more on Johnson, please see Criminal Conviction Reversed After State Failed to Timely & Fully Disclose its Use of a Type of Artificial Intelligence – E-Discovery LLC and Maryland’s Facial Recognition Technology Statute – E-Discovery LLC.

McCaden v. State, 2025 Md. App. Lexis 598 (Jul. 15, 2025)(unreported).

“Maryland Rule 5-901(a) provides that authentication of evidence is “a condition precedent to admissibility.” “[M]ovies and tapes are easily manipulated, through such means as editing and changes of speed, to produce a misleading effect.” Washington v. State, 406 Md. 642, 651, 961 A.2d 1110 (2008) (quoting 5 Lynn McLain, Maryland Evidence § 403.6 at 592 (2001)). Moreover, as the Supreme Court recently noted, “the advent of image-generating artificial intelligence, may present unique challenges in authenticating videos.” Mooney, 487 Md. at 734-35. Nevertheless, a video can be admissible if it is properly authenticated. Id. at 735.”[1]

Md. Rule 2-504.3 – Computer-Generated Evidence

(a) Definition–Computer-Generated Evidence. “Computer-generated evidence” means (1) a computer-generated aural, visual, or other sensory depiction of an event or thing and (2) a conclusion in aural, visual, or other sensory form formulated by a computer program or model. The term does not encompass photographs merely because they were taken by a camera that contains a computer; documents merely because they were generated on a word or text processor; business, personal, or other records or documents admissible under Rule 5-803 (b) merely because they were generated by computer; or summary evidence admissible under Rule 5-1006, spread sheets, or other documents merely presenting or graphically depicting data taken directly from business, public, or other records admissible under Rules 5-802.1 through 5-804.

(b) Notice.

(1) Except as provided in subsection (b)(2) of this Rule, any party who intends to use computer-generated evidence at trial for any purpose shall file a written notice within the time provided in the scheduling order or no later than 90 days before trial if there is no scheduling order that:

(A) contains a descriptive summary of the computer-generated evidence the party intends to use, including (i) a statement as to whether the computer-generated evidence intended to be used is in the category described in subsection (a)(1) or subsection (a)(2) of this Rule, (ii) a description of the subject matter of the computer-generated evidence, and (iii) a statement of what the computer-generated evidence purports to prove or illustrate; and

(B) is accompanied by a written undertaking that the party will take all steps necessary to (i) make available any equipment or other facility needed to present the evidence in court, (ii) preserve the computer-generated evidence and furnish it to the clerk in a manner suitable for transmittal as a part of the record on appeal, and (iii) comply with any request by an appellate court for presentation of the computer-generated evidence to that court.

(2) Any party who intends to use computer-generated evidence at trial for purposes of impeachment or rebuttal shall file, as soon as practicable, the notice required by subsection (b)(1) of this Rule, except that the notice is not required if computer-generated evidence prepared by or on behalf of a party-opponent will be used by a party only for impeachment of other evidence introduced by that party-opponent. In addition, the notice is not required if computer-generated evidence prepared by or on behalf of a party-opponent will be used only as a statement by a party-opponent admissible under Rule 5-803 (a).

(c) Required Disclosure; Additional Discovery. Within five days after service of a notice under section (b) of this Rule, the proponent shall make the computer-generated evidence available to any party. Notwithstanding any provision of the scheduling order to the contrary, the filing of a notice of intention to use computer-generated evidence entitles any other party to a reasonable period of time to discover any relevant information needed to oppose the use of the computer-generated evidence before the court holds the hearing provided for in section (e) of this Rule.

(d) Objection. Not later than 60 days after service of a notice under section (b) of this Rule, a party may file any then-available objection that the party has to the use at trial of the computer-generated evidence and shall file any objection that is based upon an assertion that the computer-generated evidence does not meet the requirements of Rule 5-901 (b)(9). An objection based on the alleged failure to meet the requirements of Rule 5-901 (b)(9) is waived if not so filed, unless the court for good cause orders otherwise.

(e) Hearing and Order. If an objection is filed under section (d) of this Rule, the court shall hold a pretrial hearing on the objection. If the hearing is an evidentiary hearing, the court may appoint an expert to assist the court in ruling on the objection and may assess against one or more parties the reasonable fees and expenses of the expert. In ruling on the objection, the court may require modification of the computer-generated evidence and may impose conditions relating to its use at trial. The court’s ruling on the objection shall control the subsequent course of the action. If the court rules that the computer-generated evidence may be used at trial, when it is used, (1) any party may, but need not, present any admissible evidence that was presented at the hearing on the objection, and (2) the party objecting to the evidence is not required to re-state an objection made in writing or at the hearing in order to preserve that objection for appeal. If the court excludes or restricts the use of computer-generated evidence, the proponent need not make a subsequent offer of proof in order to preserve that ruling for appeal.

(f) Preservation of Computer-Generated Evidence. A party who offers or uses computer-generated evidence at any proceeding shall preserve the computer-generated evidence, furnish it to the clerk in a manner suitable for transmittal as a part of the record on appeal, and present the computer-generated evidence to an appellate court if the court so requests.

Committee note: This section requires the proponent of computer- generated evidence to reduce the computer-generated evidence to a medium that allows review on appeal. The medium used will depend upon the nature of the computer-generated evidence and the technology available for preservation of that computer-generated evidence. No special arrangements are needed for preservation of computer-generated evidence that is presented on paper or through spoken words. Ordinarily, the use of technology that is in common use by the general public at the time of the hearing or trial will suffice for preservation of other computer- generated evidence. However, when the computer-generated evidence involves the creation of a three-dimensional image or is perceived through a sense other than sight or hearing, the proponent of the computer-generated evidence must make other arrangements for preservation of the computer-generated evidence and any subsequent presentation of it that may be required by an appellate court.

Cross reference: For the shortening or extension of time periods set forth in this Rule, see Rule 1-204.

Source: This Rule is new.

Credits

[Adopted Feb. 10, 1998, eff. July 1, 1998. Amended Sept. 10, 2009, eff. Oct. 1, 2009.]

Proposed Fed.R.Evid. 707

preliminary-draft-of-proposed-amendments-to-federal-rules_august2025.pdf

Machine-Generated Evidence
When machine-generated evidence is offered without an expert witness and would be subject to Rule 702 if testified to by a witness, the court may admit the evidence only if it satisfies the requirements of Rule 702(a)-(d). This rule does not apply to the output of simple scientific instruments.

____

[1] Emphasis added to “AI” and “artificial intelligence.”