Litigatingdeepfakes: verifyingevidenceintheaiera
Technology Section
continuedfrompage58
explicability— and establishes expert testimony as the“ presumptive” requirement for admissibility. An expert who provides the foundation for admissibility would also need to be available for cross-examination at trial so the jury can hear how the system reached its conclusion in the specific case. The draft includes an“ exceptional circumstances” exception permitting admissibility through“ other proof” such as validation studies, though commentators have cautioned that without clear guardrails, this exception could swallow the rule. Notably, the draft does not address how Rule 707 interacts with Rule 703, raising concerns that a party could bypass the new reliability requirements entirely by having an expert simply“ adopt” AI output as the basis of an opinion.
Draft Rule 901( c) addresses deepfake challenges through a two-step burden-shifting framework. First, an opponent claiming fabrication must present“ evidence sufficient to support a finding of fabrication”— a threshold designed to prevent bare accusations from derailing trials. Once that showing is made, the burden shifts to the proponent to prove authenticity by a preponderance of the evidence under Rule 104( a), a higher standard than the prima facie showing traditionally required under Rule 901( a). Key open questions include what types of evidence satisfy the threshold( expert forensic analysis, circumstantial evidence, or both), whether the preponderance standard is workable for litigants who lack resources for forensic experts, and whether the rule’ s limitation to“ generative AI” is too narrow, because deceptive editing and splicing raise identical authenticity concerns but fall outside the definition.
What Practitioners Should Do Now
Even before new rules take effect, litigators should preserve chain-of-custody documentation for all digital evidence, prepare to address authenticity challenges through forensic analysis or expert testimony, and familiarize themselves with deepfake detection tools and provenance metadata standards. Practitioners involved in arbitration and mediation should note that several organizations have already published AI guidelines addressing fabricated evidence, even though evidentiary guardrails in ADR remain far less formalized.
Whether or not Rules 707 and 901( c) are ultimately adopted, the authentication challenges they address are already here. Litigators who understand these issues now will be better positioned to protect their clients, and the integrity of the proceedings, in the AI era. n
Author: Chris Bach – Phelps Dunbar LLP
N O V- D E C 2 0 2 6 | H C B A L A W Y E R
5 9