Senate Bill S9390

                                 
BILL NUMBER: S9390

SPONSOR: CANZONERI-FITZPATRICK
 
TITLE OF BILL:

An act to amend the civil practice law and rules, the criminal procedure
law and the family court act, in relation to the rules of evidence and
artificial intelligence

 
PURPOSE:

To outline and clarify the rules of evidence with respect to admissibil-
ity of evidence alleged to be fabricated through generative artificial
intelligence.

 
SUMMARY OF PROVISIONS:

Section 1: This bill adds a new rule 4540-b to the civil practice law
and rules outlining admissibility of evidence alleged to be fabricated
by generative artificial intelligence. Provides that a court shall only
conduct an inquiry into an allegation that evidence has been fabricated
by means of generative artificial intelligence if the party making such
              

claim makes a showing sufficient to support a reasonable inference that
the evidence may not be authentic. Further stipulates that the
provisions of this section apply to all evidence offered in a civil
practice proceeding subject to authentication under law, including self-
authenticating evidence.

Section 2: This bill adds a new section 60.80 to the criminal procedure
law outlining admissibility of evidence alleged to be fabricated by
generative artificial intelligence. Provides that a court shall only
conduct an inquiry into an allegation that evidence has been fabricated
by means of generative artificial intelligence if the party making such
claim makes a showing sufficient to support a reasonable inference that
the evidence may not be authentic. Further stipulates that the
provisions of this section apply to all evidence offered in a criminal
proceeding subject to authentication under law, including self-authenti-
cating evidence.

Section 3: This bill adds a new section 164-a to the family court act
outlining admissibility of evidence alleged to be fabricated by genera-
tive artificial intelligence. Provides that a court shall only conduct
an inquiry into an allegation that evidence has been fabricated by means
of generative artificial intelligence if the party making such claim
makes a showing sufficient to support a reasonable inference that the
evidence may not be authentic. Further stipulates that the provisions of
this section apply to all evidence offered in a family court proceeding
subject to authentication under law, including self-authenticating
evidence.

Section 4: Provides the effective date.

 
JUSTIFICATION:

A recent New York Court of Appeals decision in Matter ofMS. (M.H.)
(2026 NY Slip Op 00825), reversed prior rulings to limit contact between
two children and their mother in relation to a child sex abuse case
stemming from alleged repeated contact between the mother's former
boyfriend and her daughter. The Court held that evidence used in Family
Court was inadmissible due to a lack of authentication and establishment
of the video's reliability. In the majority's opinion, concerns regard-
ing the "increasing prevalence of `deepfake' videos" were cited as at
least a partial basis for the high standard of authenticating evidence.
However, as pointed out in one of the dissenting opinions, "the majority
creates new and perhaps insurmountable hurdles for future authentication
of video evidence.  As a result, children will be harmed and abusers
will escape accountability."

The video in question was uncovered through a separate case a 2022 FBI
investigation into child pornography where a suspect claimed to have
hacked into security web cameras of various homes, and had videos that
were found to reveal the sexual abuse at the heart of Matter of MS. The
evidence allegedly showed graphic video of a M.H.'s daughter being
repeatedly sexually abused by an adult man identified as the mother's
live-in-boyfriend. The video was corroborated by an FBI agent specializ-
ing in child pornography and by matching details from the video with the
layout of M.H.'s living room, as well as placement of several items in
the video. While the majority opinion pointed out that "deepfake" videos
can utilize genuine details to fabricate content, disturbingly, some of
the items identified in the video were sex toys the boyfriend allegedly
used on the child in the video that matched specific sex toys later
found to be owned by the boyfriend and that were kept in a fingerprint-
locked bedroom that did not have cameras. The Court of Appeals contended
that law enforcement did not do enough to verify the authenticity of the
video presented as evidence, however, again, the dissenting opinions
highlighted the incredible, and unrealistic obstacles the Court is
imposing to holding dangerous offenders accountable because the specter
of "deepfake" videos exists.

In her dissent, Judge Singas warned that "it is hard to imagine what
circumstantial evidence could ever suffice to authenticate a video" as a
result of this ruling. One challenge in authenticating such videos is
that there are good reasons not to require child victims to testify to
their abuse, for one, it can be deeply traumatic for them to relive
these horrors. Further, anyone captured on video where there is on-going
sexual abuse would almost certainly be culpable in such abuse, so courts
cannot simply rely on defendants acknowledging the accuracy of video
demonstrating their criminal acts. In ruling that such strongly
supported circumstantial evidence, in conjunction with testimony from an
FBI agent who testified that he did not see anything "that led (him) to
believe that the video footage had been tampered with or altered in any
way," is inadequate to even be admitted into evidence despite the fact
that the theory of the video being a "deepfake" was not raised before
appeal jeopardizes not only the safety of victims, but it could be weap-
onized against defendants. Judge Singas noted that the government could
similarly establish an impossibly high burden against a defendant who
offers video evidence to establish an alibi.

The dissenting judges also raised the practical hurdles this ruling
creates for both defendants and prosecutors. If anyone can simply allege
evidence is "deepfaked," even if there is nothing to support that
assertion, will every piece of evidence ranging from videos to documents
and signatures require experts in "deepfake" technology to rule out such
possibility? Further, it is not even clear that an adequate number of
such experts exist to testify in every proceeding in civil courts, crim-
inal courts, and family courts. Professor Rebecca Delfino, who studies
"deepfakes" and their challenges, remarked that "(o)ne of the fundamen-
tal tenets of the American legal system is that the trier of fact either
the judge or the jury is best equipped to find the truth based on the
evidence presented." While there is no doubt that emerging technologies
will always pose both opportunities and challenges to our justice
system, it is essential that we have established guidelines to ensure
justice is served.  Without prudent measures to outline a process for
handling claims that evidence may have been fabricated with the use of
generative artificial intelligence, our judicial system runs the risk of
being paralyzed through appeals or insurmountable h urdles that lack
basis in any true justification or reasonable inference.

This legislation would tackle these challenges by incorporating the
draft rules of evidence from the federal Advisory Committee on Evidence
Rules to address claims of "deepfaked" evidence. By requiring contending
parties to demonstrate sufficient reasonable inference of such a claim,
it would ensure that alleging "deepfake" without merit is not a viable
legal strategy to escape accountability. This bill is essential to main-
taining a functional judicial system one that can effectively take on
criminals, and more importantly, keep our most vulnerable safe.

 
LEGISLATIVE HISTORY:

New bill.

 
FISCAL IMPLICATIONS:

To be determined.

 
EFFECTIVE DATE:
This act shall take effect immediately.