The Illusion of Certainty

For decades litigants and their attorneys searched through discovery for the quintessential “smoking gun” – an undeniable audio recording, a clear security video, or an internal string of text messages meant a case was won or lost. Not anymore.  Artificial intelligence has broken the traditional link between “seeing” and “believing” in the courtroom.  Where once juries were likely to be convinced by the paper trail, AI has introduced an expansive cause for doubt and hesitation.  It isn’t just that sophisticated adversaries can fabricate deepfakes; it is that they can now look at perfectly authentic, devastating evidence and plausibly cry “AI” to escape liability.

The Strategic Threat: Two Sides of the AI Coin

The ability of commercially available AI systems to produce authentic-sounding voice recordings, alter video footage, or manufacture written records creates the threat of fabricated evidence.   Until recently audio and video could help alleviate the risk of eyewitness misidentification, poor memory, or bias, settling the “he said, she said” dispute.  Indeed, the ubiquity of cell phone cameras and audio recording meant events in recent years have been better documented than ever before.  AI, however, risks dramatically undermining the believability of this entire category of proof.  If voicemails and emails can be fabricated convincingly, how is a jury of laypeople supposed to properly weigh and consider them?

Hesitancy to ascribe reliability to recorded evidence leads to the second threat from AI: the “Liar’s Dividend,” or the deepfake defense.  A party pleading for a jury to believe him rather than their own lying eyes once had a steep hill to climb.  Now, litigants can claim genuine historical statements or videos are AI-generated fabrications.  This risk is not hypothetical.  In Huang v. Tesla, the defense tried to disavow real, recorded statements by CEO Elon Musk about the safety of the Autopilot feature as potential deepfakes.  The California state court judge rejected this argument and ordered an apex deposition of Musk.  The case subsequently settled.

Changing the Rules of Engagement

To address these twin problems, federal and state court judges are looking for ways to move away from historically lax authentication standards, where a witness simply saying “that looks right” was good enough to render evidence admissible.  The Advisory Committee on Evidence Rules proposed two amendments to the Federal Rules of Evidence (which are closely aligned with most state’s rules of evidence) to address these concerns:

  • Proposed Rule 707: Intended to subject machine-generated evidence to the same strict reliability standards as human scientific expert testimony.
  • Proposed Rule 901(c): Designed to tackle the deepfake defense directly by putting the initial burden on the objecting party to provide concrete forensic evidence of an irregularity before a judge determines whether to exclude it.

However, the rulemaking process has already hit a massive snag. Proposed Rule 707 was recently shelved by the Committee due to major questions as to whether it would actually address the problem of surreptitious AI, and whether it risks sweeping in too much evidence in a burdensome process. Because Rule 707 only triggers when a party acknowledges they are offering machine-generated evidence, it is completely blind to hidden AI. If an adversary introduces a deepfake and lies about its origin, an expert reliability rule like 707 cannot catch them.

Finding Solid Footing in Shifting Sands 

Without clarity on whether and when the Federal Rules of Evidence will change to address these problems, litigants and their counsel will have to be proactive. Success requires ensuring you can both demonstrate the authenticity of your own evidence and successfully challenge unreliable evidence based on machine learning as well as pure deepfakes. It requires both an outstanding command of the evidentiary record and technical savvy to ensure that the credible and only credible evidence gets in front of the jury.

At Keller Anderle Scolnick, our tight teams of experienced trial lawyers know their records forward and back. They leverage their knowledge, skill, and tenacity in connection with advanced digital forensics experts to protect the integrity of digital evidence and the trial process. In high-stakes commercial and white-collar disputes, you cannot rely on the evidence to speak for itself. You need a trial team that understands how to make a judge listen, how to expose fabrications, and how to protect the truth.

Bring us your toughest case.  We’ll get the job done.