The Risk to Fair Trial: The Challenge of Authentication with AI

By Mariel Barnes '29

---

Imagine a criminal trial where the prosecution presents video and audio that appear to show the defendant committing a crime. The evidence could seem so real that a jury may not think to question whether it is authentic. But if the recording was generated or altered by AI, it would show how easily this technology can threaten a defendant’s right to a fair trial.

Current authentication methods may no longer be sufficient to ensure the reliability of digital evidence because artificial intelligence can now generate highly convincing videos, images, and audio. (1) To reduce the risk of fake media influencing criminal convictions and undermining defendants’ fair trial rights, state and federal courts should adopt stronger methods for verifying digital evidence.

AI-generated media creates a major evidentiary problem because video, audio, and images can strongly influence how jurors view a case. Truthful and reliable evidence is crucial to a fair outcome. When jurors rely on video or audio to understand a case, that evidence can shape their view of the facts in powerful ways. This can mislead the court and wrongfully influence the jury to decide a case based on false evidence. As this technology continues to advance, federal and state courts need stricter verification processes to protect fair trials.  

Federal Rule of Evidence 901 governs authentication by requiring the party offering evidence to produce sufficient proof that the item is what the party claims it is. (2) This rule is one of the main evidentiary tools courts use to determine whether evidence is authentic. Witness testimony, analysis of specific characteristics, chain-of-custody documentation, and other supporting evidence can all help establish authenticity. (3) But because AI can manipulate videos, audio, and images so convincingly, people may not always be able to tell when evidence has been altered or generated. As a result, standard witness testimony and basic review may no longer be enough in every case.  

However, AI creates new problems for Rule 901 by making digital evidence harder to authenticate reliably. A witness may be persuaded by false audio or video that appears familiar, especially when memory is imperfect. Digital footage and audio may still satisfy basic verification standards in some cases, but courts need stronger protective measures when there is reason to believe the material may have been altered or fabricated through AI. (4)  

Federal Rule of Evidence 702 is also important because it governs expert testimony, which can be especially useful when evidence may have been generated or altered by AI. (5) Judges and juries are not always equipped with the technical knowledge needed to evaluate this kind of evidence, which makes forensic experts more important in reviewing metadata, digital history, and signs of AI manipulation. As a result, courts may increasingly rely on expert analysis to determine whether digital evidence is trustworthy enough to be used at trial. 

Additionally, Daubert requires judges to act as gatekeepers by deciding whether expert testimony is based on reliable methods. (6) In cases involving possible AI-generated or AI-altered evidence, judges should evaluate whether an expert used reliable forensic methods before allowing that testimony to reach the jury. As technology advances, manipulated media becomes easier to spread and harder to detect with certainty. Under the Daubert standard, judges must assess whether an expert’s methods of verification are reliable.  

When a party presents digital evidence that may have been generated or altered by AI, courts should require a more reliable authentication process. (7) AI makes it harder for the party offering the evidence to prove that it is what it appears to be under Rule 901, since a video, image, or audio recording may look genuine even after alteration. As a result, courts should require more specific proof from the offering party, such as data analysis, records showing the origin and handling of the evidence, and proof of the method used to verify the file.

Courts should also require forensic expert testimony under Rule 702 when authenticity is reasonably questioned. (8) They should then rely on Daubert to ensure that the expert’s methods are trustworthy. (9) Evidence that may have been altered by AI should be thoroughly reviewed before it is presented to the jury, but that does not mean digital evidence should be automatically excluded. A strengthened authentication process would better protect accurate decision-making and a defendant’s right to a fair trial as AI-generated media continues to advance.  

Endnotes

1. Jonathan D. Uslaner and Matthew Goldstein, “Proposed AI Evidence Rule Highlights New Challenges for Federal Practitioners,” Reuters, April 23, 2026, https://www.reuters.com/legal/legalindustry/proposed-ai-evidence-rule-highlights-new-challenges-federal-practitioners–pracin-2026-04-23/.

2. Cornell Law School, “Rule 901. Authenticating or Identifying Evidence,” Legal Information Institute, 2011, https://www.law.cornell.edu/rules/fre/rule_901.

3. Ibid.

4. United States v. Vayner, 769 F.3d 125 (2d Cir. 2014).

5. Cornell Law School, “Rule 702. Testimony by Expert Witnesses,” Legal Information Institute, 2011, https://www.law.cornell.edu/rules/fre/rule_702.

6. Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993).

7. Jonathan D. Uslaner and Matthew Goldstein, “Proposed AI Evidence Rule Highlights New Challenges for Federal Practitioners.”

8. Cornell Law School, “Rule 702. Testimony by Expert Witnesses.”

9. Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993).