Matter of M.S. (M.H.)

Issue

In a New York Family Court child-protective proceeding, what foundation must a county social services agency lay to authenticate videos that it did not create, that were never recovered from the family's own devices, and that reached law enforcement years later through a third party who never testified? And does the availability of deepfake technology change that foundation?

Rule

The New York Court of Appeals applied People v Patterson, 93 NY2d 80, 84 (1999), under which video evidence may be authenticated in one of two ways: by testimony of a witness to the recorded events, or of an operator, installer or maintainer of the equipment, that the recording accurately represents the subject matter depicted; or by testimony, expert or lay, establishing that the recording truly and accurately represents what was before the camera. The proponent carries the burden of making a sufficient threshold showing of reliability (People v Price, 29 NY3d 472, 482 [2017]), which can be met by proof that the evidence is genuine and that there has been no tampering with it (People v McGee, 49 NY2d 48, 59 [1979]). Zegarelli v Hughes, 3 NY3d 64, 69 (2004), applies the Patterson standard in civil proceedings, and the Court of Appeals held in this case that the rules of evidence apply in Family Court just as they apply in any other New York court.

Application

Erie County's only evidence of abuse was three timestamped videos, and the foundation it offered for them came from two witnesses. FBI Agent Martin Baranski described how the videos surfaced: while executing a search warrant on B.W., a Syracuse man suspected of trading child sexual abuse material, agents were told by B.W. that he had been hacking into security cameras and had saved footage from one such camera in 2019. Baranski recovered three video files from a copy of B.W.'s computer and passed identifying information to New York State Police. Investigator Gary Mahoney testified that the living room he saw when he searched the family home matched the room in the videos, that he observed cameras in the house, and that objects of the kind shown in the videos were recovered there. The mother identified the two people in screenshots. Family Court credited Baranski's technical experience and expertise, found no signs the videos had been manipulated or altered, and gave the videos great weight. The New York Court of Appeals measured that foundation against Patterson and found it weaker on every axis. In Patterson the Court held a grocery store surveillance tape inadequately authenticated even though the store owner had himself inserted the tape, gave it to police two weeks after the robbery, and an officer testified that the video accurately showed the store's physical layout. Here, D.K., who allegedly created the recordings, offered no evidence or testimony, and B.W., the man who claimed to have hacked the camera, never testified either; his account came in only as hearsay through Agent Baranski. There was no continuous feed, only excerpts B.W. selected from footage he never explained selecting, and Baranski did not testify that B.W. had told him the files were unaltered. The store owner in Patterson was a businessperson with no apparent ulterior motive; B.W. faced criminal prosecution when he supplied the information. And roughly two and a half years separated the recording of the videos from their recovery by the FBI, a gap the Court described as sixty-five times as long as the two-week gap in Patterson. On Agent Baranski, the Court's objection was about qualification and questioning rather than about his character. The county never sought to qualify him as an expert in video authentication or to establish his experience in it, and broad familiarity with child sexual abuse material does not make a witness learned in detecting tampering. He was never asked whether he examined the videos for tampering, whether he used any forensic tool, or whether he could offer an affirmative opinion that the videos were untampered. The single question put to him was whether he made any observations that led him to believe the footage had been tampered with or altered in any way, and he answered no, without elaboration. On Investigator Mahoney, the Court held that matching a room's layout and contents to personal observation was the same showing rejected in Patterson, and it was here that the majority reached deepfakes: because fabricated videos typically leverage real details of real places and people and alter only what the maker wants altered, corroborating identifying features in a video says little about whether the depicted events are real. The majority drew limits that a reader should not skip. It stated that failure to authenticate does not mean the evidence was false, that it was not requiring child victims to testify, and that Agent Baranski, B.W., or some other witness could have supplied an adequate foundation had the right questions been asked. It also said the likelihood that these particular videos were altered was beside the point, because the question is whether the proponent carried its burden. The decision was four to three, and the dissents attack the deepfake reasoning directly. Judge Troutman, joined by Judges Garcia and Singas, wrote that no one uttered the word deepfake in Family Court, that the mother raised the theory for the first time on appeal, and that no evidence of fabrication was ever presented. Her dissent sets out the detail Family Court found matching between the videos and the home, including the couch, painting, afghan, end table and lamp, and specific objects recovered under a search warrant. She argued the majority's rule invites a party to posit a deepfake with no evidentiary support and then force the proponent to disprove it, and she questioned whether deepfake-detection experts exist in sufficient numbers, or at prices a local child protective agency can pay. Judge Singas, joined by Judge Garcia, wrote separately to argue that the fabrication the majority feared was not technically possible in 2019 or 2022, citing published surveys of deepfake and video generation research, and that the majority's approach departs from other jurisdictions, including the draft Federal Rule of Evidence 901(c) prepared by the federal Advisory Committee on Evidence Rules, which would require the opponent of an exhibit to present evidence sufficient to support a finding of fabrication before any inquiry is triggered. One further fact bears on any remand: a footnote records that the videos were destroyed by Family Court order in August 2024, a month after the mother moved for leave to appeal.

Conclusion

The New York Court of Appeals reversed the Appellate Division orders, without costs, and dismissed the article 10 petitions, holding that Family Court erred in admitting the three videos because Erie County had not authenticated them. Because the videos were the county's sole evidence of abuse, the Court did not reach the mother's arguments about whether Family Court Act section 1012(e)(iii)(A) requires actual knowledge, whether the derivative abuse finding as to her son could stand, or whether the dispositional order served his best interests. Chief Judge Wilson wrote for the majority, joined by Judges Rivera, Cannataro and Halligan. Judge Troutman dissented in an opinion joined by Judges Garcia and Singas, and Judge Singas dissented separately in an opinion joined by Judge Garcia.

the increasing prevalence of "deepfake" videos has only rendered the method of matching circumstantial details in a video to personal observations a more suspect form of authentication

Majority, Wilson, Ch. J., for a four-judge majority

The failure to authenticate evidence sufficiently does not mean the evidence was false, but only that it was not properly authenticated according to the rules of evidence.

Majority, Wilson, Ch. J., limiting the reach of the holding

Yet no one uttered the word "deepfake" during the Family Court proceedings below, nor was there any evidence presented that the portions of the videos depicting sexual abuse were fabricated.

Troutman, J., dissenting, joined by Garcia and Singas, JJ.

It seems that in the future, a party opposing the introduction of video evidence need only posit, with no evidentiary support whatsoever, that the video might be a deepfake.

Troutman, J., dissenting

The majority's novel burden flips the authentication standard on its head, requiring the proponent to disprove unsupported theories of fakery as a prerequisite to mere admissibility.

Singas, J., dissenting, joined by Garcia, J.

Family Court, criminal, and civil litigants alike will surely raise a deepfake defense, however baseless, when faced with unfavorable evidentiary exhibits of all kinds, and the most vulnerable among us will suffer the consequences.

Singas, J., dissenting, closing prediction

Why it matters. This is the first state high court decision to fold deepfake risk into a video authentication holding and then reverse on it, and it does so in a jurisdiction, New York, whose Patterson standard many states track. But it is a four to three decision in which three judges say the deepfake theory appeared for the first time on appeal with nothing in the record behind it, so a practitioner who cites the majority without the dissents will overstate what New York now requires of a proponent.

Read the full opinion on CourtListener

This summary was written from the opinion itself. Every quotation above was located in the text of the decision it is attributed to. It is a summary for public education and is not legal advice.

Related reading