On the Potential Abrogation of Federal Rule of Evidence 901(b)(9)

Essay - by Online Edition - Volume 104

Introduction

One of the key questions related to the impact of artificial intelligence (AI) on the practice of law is how AI will change the introduction and assessment of evidence in the courtroom.[1] To begin to address this issue, the federal Advisory Committee on Evidence Rules has proposed new Federal Rule of Evidence 707, which would require AI-generated evidence offered without an expert witness to satisfy the reliability requirements of Rule 702 governing expert testimony.[2]

As the Committee awaited public comments on the proposed rule, its Reporter addressed recent related scholarship, including Professor Edward Imwinkelried’s work advocating for the evaluation of AI-generated evidence through standards of reliability rather than authentication.[3] In the piece, Professor Imwinkelried dismisses Rule 901(b)(9), which allows evidence to be authenticated by describing a “process or system” that produces an accurate result,[4] as misplaced within the authentication rules and therefore an inappropriate lens to evaluate AI-generated evidence.[5] Professor Imwinkelried then calls for Rule 901(b)(9)’s repeal.[6] In response, the Reporter’s October 2025 memo on proposed Rule 707 recommended that the Committee decline to consider the suggestion.[7] While acknowledging that Rule 901(b)(9)’s text improperly conflates authentication and reliability, the memo cautioned that abrogating the rule could have unintended consequences for the authentication of other evidence, such as “simple machine-generated evidence.”[8]

The Reporter’s memo’s commentary regarding Rule 901(b)(9) raises some practical questions: What role does Rule 901(b)(9) currently play in federal litigation? What consequences would result if it were abrogated? This essay offers a preliminary examination of that question through a study of federal court opinions issued between January 1, 2020 and April 1, 2026 that address authentication under Rule 901(b)(9). Part I describes the Advisory Committee’s proposal of Rule 707 and the October 2025 Reporter’s memo’s related commentary on Professor Imwinkelried’s call to eliminate Rule 901(b)(9). Part II discusses the results of the study: subpart II(A) analyzes the opinions in the sample set from a quantitative perspective, while subpart II(B) offers qualitative observations. The essay concludes with reflections on Rule 901(b)(9)’s future in the AI era.

I. Proposed Rule 707 and the Call for Rule 901(b)(9)’s Abrogation

Since October 2023, the Judicial Conference Advisory Committee on the Federal Rules of Evidence has been considering how to address the impact of AI on the presentation of evidence in court proceedings.[9] These efforts culminated in the proposal of a new Federal Rule of Evidence 707 dealing with evidence generated through machine learning.[10] The following prior version of proposed Rule 707 was submitted for public comment on August 15, 2025:

Rule 707. Machine-Generated Evidence

When machine-generated evidence is offered without an expert witness and would be subject to Rule 702 if testified to by a witness, the court may admit the evidence only if it satisfies the requirements of Rule 702(a)-(d).[11] This rule does not apply to the output of simple scientific instruments.[12]

The comment period was set to close on February 16, 2026.[13]

In October 2025, as the Committee awaited public comment on proposed Rule 707, the Reporter, Professor Daniel Capra, prepared a memorandum reviewing new developments related to the proposed rule, including recent scholarship touching on the AI issues with which the rule is concerned.[14] Among the works the memo addressed is Professor Edward Imwinkelried’s recent piece discussing how preliminary questions on the admissibility of AI-generated evidence should be determined.[15] In the article, Professor Imwinkelried argues that Federal Rule of Evidence 901(b)(9) is misplaced in Rule 901[16] governing authentication:

Rule 901(a) sets out the conditional relevance standard of proof;[17] Rule 901(b) indicates that that standard governs the listed authentication issues. One such provision is Rule 901(b)(9): “Evidence describing a process or system and showing that it produces an accurate result.” Even at first blush, the provision is an odd fit for Rule 901. Every other provision in Rule 901(b) describes a technique for authenticating a particular item of evidence. In contrast, Rule 901(b)(9) describes techniques—processes or systems—that can be used to generate items of evidence that in turn will require authentication; Rule 901(b)(9) is an apple among oranges. The content of 901(b)(9) is one step removed from the content of all of the other provisions of 901(b). The other provisions deal with the authentication of a particular writing, audio, or image, while 901(b)(9) deals with a system such as an AI tool that can generate a writing, audio, or image. There is a strong case that 901(b)(9) does not belong in Rule 901 and that the validation of a process or system should not be governed by Rules 104(b) and 901(a).[18]

As a result, Professor Imwinkelried concludes that Rule 901(b)(9) is not an appropriate mechanism to determine the admissibility of evidence created using AI.[19] He then advocates for its repeal.[20]

After summarizing these arguments, the memo responded, acknowledging that Rule 901(b)(9) is “problematic” because it conflates the question of authenticity with a reliability determination.[21] The requirement in Rule 901(b)(9) for evidence showing that the process or system “produces an accurate result” necessarily requires the proponent to demonstrate the validity of the mechanism used to generate the evidence—but the Federal Rules of Evidence assess the validity of scientific and other methodologies under Rule 702, not Rule 901.[22] Despite this mismatch, the memo expressed hesitation to abrogate Rule 901(b)(9):

[T]his does not necessarily mean that Rule 901(b)(9) should be abrogated. Abrogating a rule is a major step, and the consequences of abrogating a rule need to be seriously considered. The rule has a role outside of machine-learning, such as providing a ground of authenticity for simple machine-generated evidence, and so deleting it could have negative consequences.[23]

Later, the memo echoed this position, noting that the reliability aspect of Rule 901(b)(9) is “misplaced,” but suggesting that the Committee decline to consider abrogating the rule at this time.[24] If the Committee were to turn its full attention to the possibility of abrogating Rule 901(b)(9), the memo explained, “[c]onsideration would have to be given to whether it would negatively impact authentication for the kinds of machines that are not covered by Rule 707 (such as electric scales and blood pressure machines).”[25]

The memo did make clear, however, that the Committee does not intend for AI-generated evidence to be evaluated under Rule 901(b)(9).[26] Instead, proposed Rule 707 would require the offering party to establish that the evidence is reliable under the more stringent standards of Rule 702, even when the evidence is not introduced through an expert witness.[27] The memo suggested an Advisory Committee note to proposed Rule 707 to prevent confusion regarding the potential application of Rule 901(b)(9).[28]

At its November 2025 meeting, the Committee agreed with the Reporter’s distinction between Rules 707 and 901(b)(9), stating in its meeting report that “Rule 707 renders Rule 901(b)(9) irrelevant in the cases to which Rule 707 applies.”[29] Regarding abrogation, the Committee conceded that “Rule 901(b)(9) is an odd fit for an authenticity rule,” but determined that eliminating it was inadvisable without further consideration of the types of evidence not covered by Rule 707.[30] The Committee also accepted the Reporter’s suggestion to clarify the relationship between Rules 707 and 901(b)(9) in its notes on the proposed rule.[31]

In May 2026, after the public comment period had closed, the Committee convened again and made significant revisions to proposed Rule 707 in response to issues raised in the comments.[32] The current language of the proposed rule appears below:

Rule 707. Evidence Produced by Artificial Intelligence and Presented at Trial Without an Expert

(a) General Rules. If evidence is a product of artificial intelligence and is offered without an expert witness, but would be subject to Rule 702 if testified to by a witness, the proponent must establish that the evidence:

(1) will help the trier of fact;

(2) is based on sufficient facts or data;

(3) is the product of reliable principles and methods; and

(4) reflects a reliable application of the principles and methods to the facts of the case.

(b) Establishing Admissibility. Admissibility under this rule ordinarily requires the proponent to provide an expert to explain how the system of artificial intelligence reliably produced the evidence; but in exceptional circumstances, the court may rely upon other proof to find that the output satisfies the requirements of (a)(2)-(4).

(c) Notice. The evidence is admissible only if the proponent gives an adverse party reasonable notice of the intent to offer the evidence — so that the party has a fair opportunity to meet it.

(d) Judicial Notice. This rule does not apply to facts that may be judicially noticed under Rule 201.

(e) Definition. In this rule, “artificial intelligence” means a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments.[33]

The Committee retained the note on the relationship between Rules 707 and 901(b)(9) appended to the revised rule.[34] Given the substantial changes to the rule’s text, the Committee determined that the new version would require re-publication for comment before being adopted but elected to have the rule evaluated by experts at the Committee’s fall meeting before doing so.[35] As a result, the earliest that proposed Rule 707 could go into effect is in December 2029.[36] The Committee’s May report contained no further discussion of Professor Imwinkelried’s call to eliminate Rule 901(b)(9) or the Reporter’s concerns with doing so.[37]

II. Authentication Under Rule 901(b)(9) and the Impact of Abrogation: A Brief Study

The Reporter’s memo’s response to Professor Imwinkelried’s proposal to abrogate Rule 901(b)(9) raises the issue of whether the memo’s concerns about eliminating the rule are well-founded. What is the role of Rule 901(b)(9) outside of machine learning? What kinds of evidence are being authenticated under this rule? And what negative consequences would result from Rule 901(b)(9)’s deletion? This essay begins to address these inquiries.

Specifically, this essay explores the impact of the potential abrogation of Rule 901(b)(9) through a study of recent cases discussing authentication under that rule. The paper surveys federal court opinions available on commercial research platforms[38] issued between January 1, 2020 and April 1, 2026 that involve authentication of evidence under Federal Rule of Evidence 901(b)(9).[39] The resulting cases were analyzed from both a quantitative and a qualitative perspective.

A. Quantitative Findings

Searches yielded nineteen cases that fit the study’s criteria,[40] suggesting in itself an answer to one of the questions prompted by the Reporter’s memo: The role of Rule 901(b)(9) may be limited, at least in terms of its discussion in written judicial opinions.

Out of the nineteen opinions studied, twelve were issued in civil cases and seven were issued in criminal cases.[41] The type of proceeding ran the gamut, from admiralty to class action to employment on the civil side, and murder to drug trafficking to embezzlement of mail on the criminal side.[42] Federal circuit courts of appeals issued three of the opinions, while the other sixteen came from federal district courts.[43] In twelve cases, the court determined that the evidence at issue was admissible (or potentially admissible), while the evidence was or should have been excluded in the other seven.[44] These determinations were made at various procedural stages of the case, including in rulings on motions in limine, motions for summary judgment, and motions for new trial, as well as on appellate review.[45] These findings suggest, unsurprisingly, that Rule 901(b)(9) functions as a general-purpose authentication doctrine rather than a rule tied to a particular area of the law or procedural stage of the case.

What was perhaps surprising is that almost all of the cases studied involved the authentication of evidence derived from the use of a somewhat complex electronic device, such as a computer, a mobile phone, or a global positioning system (GPS).[46] In six cases, the evidence at issue was a document or exhibit created by a human author using information obtained from an electronic device, such as a declaration relying on electronic banking records, spreadsheets recording messages from an encrypted platform, and exhibits summarizing GPS data from a vehicle.[47] Another six involved the authentication of raw data or communications obtained directly from an electronic device, such as electronic transaction data on the use of federal SNAP benefits, cell phone contents, and emails.[48] Four cases involved images of information on an electronic device, including screenshots of file data on a computer and a cell phone video of a surveillance video from a laundromat.[49] One case focused on authenticating an arbitration agreement signed using an employer’s electronic document system,[50] and another dealt with a witness’s testimony on her use of an Apple AirTag to track missing mail.[51]

These findings highlight that the use of Rule 901(b)(9)—at least in the cases in the sample set—may be different than the Reporter’s memo contemplated. The memo expressed a concern with the impact of Rule 901(b)(9)’s abrogation on the authentication of “simple machine-generated evidence” such as information obtained from “electric scales and blood pressure machines.”[52] Some of the electronic devices involved in the cases studied might at least arguably fall into this category of simpler devices, such as a GPS or Apple AirTag. However, many of the cases involved evidence from devices that are decidedly more complex, such as a cell phone, computer, encrypted messaging platform, or electronic banking system.[53] The study therefore indicates that the primary use case for authentication under Rule 901(b)(9) may be for information derived from a computer, mobile device, or other electronic system of similar complexity.

When it came to how the Rule 901(b)(9) question arose, the cases studied generally fell into one of three categories. In ten cases, the rule was identified by the court as an appropriate authentication method for the evidence at issue, whether or not it was actually utilized by the proponent of the disputed evidence.[54] In another eight cases, the party opposing admission of the evidence cited Rule 901(b)(9) as a method of authenticating the item at issue, often in the context of arguing that the proponent should have relied on the rule but did not.[55] Finally, in only two cases, the proponent of the evidence actually used Rule 901(b)(9) to authenticate the evidence in question.[56]

These observations suggest that the potential disruptions caused by the abrogation of Rule 901(b)(9) may be minimal. Based on the cases in the sample set, authentication under that rule appears to be used relatively infrequently. Underscoring this point is the fact that, out of the seventeen cases in which Rule 901(b)(9) was mentioned but not relied on by the proponent as the basis for authentication, the court concluded that the disputed evidence was nonetheless admissible (or potentially admissible) in ten cases.[57] Other methods of authentication, then, can ostensibly provide alternatives to authenticate evidence that is currently being offered under Rule 901(b)(9).

Indeed, almost all of the cases within the sample set discussed authentication under Rule 901(b)(9) in conjunction with or as an alternative to one or more other authentication methods. Only three cases did not mention an alternative method of authentication (although one of those cases prefaced its reference to Rule 901(b)(9) with a “See, e.g.,” signal, suggesting it was not the exclusive authentication option).[58] The most common additional authentication method discussed was testimony of a witness with knowledge under Rule 901(b)(1)[59]: Thirteen of the nineteen cases either mentioned or discussed this method.[60] Five cases identified self-authentication under Rule 902(11)–(14)[61] as an alternative or companion method of authenticating the disputed evidence, while another four identified evidence of distinctive characteristics under Rule 901(b)(4) for that task.[62]

These findings indicate that the use of Rule 901(b)(9) as a standalone authentication method is uncommon. In addition, the frequency with which the testimony of a witness with knowledge authentication method was coupled with Rule 901(b)(9) positions it as a logical gap-filler if Rule 901(b)(9) were abrogated. In several of the cases studied, as might be expected, the evidence that would be used to satisfy 901(b)(9) (i.e., to demonstrate that the “process or system . . . produce[d] an accurate result”[63]) was the testimony of a witness with knowledge of that system and how it was used to obtain the disputed evidence.[64] In others, Rule 901(b)(9) was suggested by the opposing party as the (or a) proper method that the proponent should have used to authenticate disputed evidence, but the court concluded that authentication through testimony of a witness with knowledge was sufficient instead.[65] As a result, authentication in these cases might not look very different at all in the absence of Rule 901(b)(9). In providing testimony that the “item is what it is claimed to be”[66] when the item at issue is information obtained from an electronic device (or even a simpler instrument), the witness would likely need to describe, at least at some basic level, how the machine operated to produce that item. This would, of course, result in the elimination of the reliability aspect of Rule 901(b)(9), as the “testimony of a witness with knowledge” authentication method does not require evidence that the electronic device produces accurate results.[67] To the extent that this reliability check for the kinds of evidence currently authenticated under Rule 901(b)(9) is necessary or desired, other rules of evidence could potentially provide it, as discussed further in the next section.

B. Qualitative Observations

A qualitative analysis of the sample set revealed a number of observations relevant to the Reporter’s memo’s concerns. First, and perhaps most significantly, several of the cases studied demonstrate that the conflation of the authenticity and reliability determinations under Rule 901(b)(9) is not hypothetical. For example, in a case involving the authentication of Google Earth screenshots of the plaintiff’s roof, the Eastern District of Louisiana relied on the Ninth Circuit’s prior discussion of authenticating evidence from Google Earth.[68] There, the Ninth Circuit explained that, to authenticate this evidence, the proponent “would have to establish Google Earth’s reliability and accuracy,”[69] a much higher bar than just demonstrating that the images were indeed screenshots from that website. Likewise, the Eastern District of New York denied a motion in limine to categorically admit spreadsheets reflecting encrypted messages between the defendant and others because the Certificate of Authenticity provided by the government did “nothing to establish the origins or reliability of” the information.[70] The Eighth Circuit similarly considered whether a witness’s testimony had “demonstrate[d] the accuracy” of an Apple AirTag in addressing an argument that information from the tracking device had not been properly authenticated at trial.[71] And the Southern District of Mississippi concluded that data collected from a UPS truck documenting its movements was not self-authenticating because the certifications accompanying it did not “[speak] to the accuracy of the . . . process or system.”[72]

These courts were, of course, applying the language of Rule 901(b)(9), which requires evidence that the “process or system” at issue “produces an accurate result.”[73] But this misplaced language makes the more nuanced determination of whether the evidence is reliable a hurdle to meeting the separate authentication requirement, which carries only the lower conditional relevance standard.[74]

Also notable within the sample set were some courts’ expressed views on the role and appropriate use of Rule 901(b)(9), especially with regard to electronically-stored information. The Eastern District of New York cited Rule 901(b)(9) as the typical authentication method for forensic copies or extractions of data from mobile devices or online accounts.[75] Similarly, the Northern District of Alabama explained that “courts generally rely upon evidence of the procedures used to retrieve electronically-stored information to accomplish proper authentication.”[76] The Eastern District of Michigan also pointed to computer-generated information as “a familiar instance where Rule 901(b)(9) applies.”[77] These comments tie in with the empirical finding from the previous section reflecting that all of the studied cases dealt with (arguably) more complex electronic devices. [78] As discussed above, the Reporter’s memo emphasizes that outside of machine learning, a key role of Rule 901(b)(9) is “providing a ground of authenticity for simple machine-generated evidence” such as electronic scales and blood pressure machines.[79] While the absence of these simple machines in the studied cases does not indicate that Rule 901(b)(9) is not currently playing a role in the authentication of evidence obtained from them, it does suggest that its prevalence for this purpose is perhaps less than the Reporter’s memo contemplates.

Related to both of the above issues—conflating authenticity with reliability and the use of the rule primarily to authenticate information obtained from more complex electronic devices—was the recurring question of whether an expert witness was needed to authenticate evidence under Rule 901(b)(9). In several of the cases studied, the parties or the court either suggested or required the use of an expert to demonstrate the accuracy of the evidence in question.[80] Where an expert is used to authenticate evidence under Rule 901(b)(9), however, the expert’s testimony is also subject to the reliability standards for expert testimony set out in Rule 702.[81] This again means that a reliability issue with the evidence in question could prevent it from meeting the less demanding authenticity standard. If evidence offered under Rule 901(b)(9) must be reliable to even be “sufficient to support a finding” of authenticity, then Rule 901(b)(9) imposes a much stricter standard for authentication than the other authentication methods in Rule 901.[82] This mismatch certainly supports Professor Imwinkelried’s call for its abrogation. In addition, from an authentication perspective, the use of an expert witness familiar with the technology in question to authenticate it would presumably constitute testimony of a witness with knowledge, which once again points toward Rule 901(b)(1) as an apt alternative method in cases where Rule 901(b)(9) is currently being used.

A separate question is whether anything is lost if this additional reliability check on the kinds of electronically-generated evidence currently being authenticated under Rule 901(b)(9) is eliminated as a result of the rule’s abrogation. For machines whose functionality the layperson is familiar with and can easily grasp, no additional reliability showing would seem to be required.[83] And for computer-generated evidence and even some simpler devices, if their functionality requires specialized knowledge to understand, Rule 702 would seem to kick in to vet the evidence’s reliability.[84] However, as some cases in the sample set demonstrate, information obtained from electronic devices of varying complexity (but not generated using AI) is sometimes permitted without an expert witness.[85] In those situations, Rule 901(b)(9)’s requirement of evidence “showing that [the process or system] produces an accurate result,”[86] when applied properly, could function as a reliability backstop to prevent inaccurate data collected or presented by a layperson from being admitted at trial. While this reliability check on electronically-stored (but non-AI-generated) information surely does not belong in the authentication analysis, abrogating Rule 901(b)(9) could result in much of this digital data coming into the record unchecked.

The version of proposed Rule 707 circulated for public comment in August 2025 could have provided a solution to this potential problem. The text of that version made Rule 707 applicable to “machine-generated evidence.”[87] The Reporter’s memo acknowledged that this language swept more broadly than the Rule’s intended focus on machine learning,[88] and several of the public comments pointed out that, by its plain language, the prior version of the Rule would apply to data and information obtained from an electronic device without the use of AI.[89] Indeed, the disputed evidence in each case in the sample set would seem to have fit within the commentators’ broader, textualist conception of “machine-generated evidence.”[90] As a result, in each case, any evidence generated via a process requiring specialized knowledge to understand would have had to satisfy Rule 702’s stricter reliability standards, whether or not an expert witness was used. This would have rendered the separate reliability check under Rule 901(b)(9) superfluous, which would have suggested abrogation of Rule 901(b)(9) as a natural change to make in conjunction with the adoption of the prior version of proposed Rule 707.

The revised language of proposed Rule 707, however, eliminates the term “machine-generated evidence,” specifying instead that it applies to evidence that “is a product of artificial intelligence.”[91] While certainly clearer and more tailored to the Rule’s intended purpose, this revision leaves open the potential for non-AI-generated information obtained from electronic devices authenticated through the testimony of a layperson to dodge a reliability analysis. Under those circumstances, Rule 901(b)(9) could still have a role to play, although its reliability backstop would remain misplaced in an authentication rule. This possibility counsels against Rule 901(b)(9)’s abrogation, at least until the Committee is able to consider whether additional reliability rules or commentary are needed to address these cases that could otherwise slip through the cracks.[92]

Conclusion

The study this paper conducted, while limited and brief, begins to shed light on the questions raised by the Reporter’s memos about abrogating Rule 901(b)(9). On one hand, the rule appears relatively infrequently in reported federal decisions and is rarely used as the sole method of authenticating evidence. These findings suggest that eliminating Rule 901(b)(9) might not significantly disrupt existing authentication practice, especially given that Rule 901(b)(1) exists as a logical gap-filler.

However, the cases in which Rule 901(b)(9) is discussed illustrate the conceptual concern underlying the call for the rule’s abrogation: the requirement that a process or system produce an “accurate result” compels courts to treat reliability as part of the authentication inquiry. While clearly a mismatch with the lower conditional-relevance standard of authentication, this language in Rule 901(b)(9) does function to create a reliability backstop for non-AI-generated evidence obtained from more complex electronic devices that, in some cases, benefits from the stricter accuracy-focused evaluation. New proposed Rule 707 will certainly provide this reliability backstop for AI-generated evidence, but as revised, it does not apply to other computer-generated evidence that is currently being run through Rule 901(b)(9). Given the potential for that category of evidence to evade any reliability analysis, Rule 901(b)(9) should perhaps remain in place for now.

    • . See, e.g., Paul W. Grimm, Maura R. Grossman, Gordon V. Cormack, Artificial Intelligence as Evidence, 19 Nw. J. Tech. & Intell. Prop. 9, 13, 84–86 (2021). ↑
    • . Report of the Advisory Committee on Evidence Rules, in Comm. on Rules of Prac. & Proc. of the Jud. Conf. of the United States, Preliminary Draft of Proposed Amendments to the Federal Rules of Appellate, Bankruptcy, Civil, and Criminal Procedure, and the Federal Rules of Evidence 100, 102, 109–11 (Aug. 2025) [hereinafter “2025 Proposed Amendments”], https://www.uscourts.gov/sites/default/files/document/preliminary_draft_of_proposed_amendments_final.pdf [https://perma.cc/423N-AYS9]. ↑
    • . Reporter’s Memorandum on Machine-Learning and Proposed Rule 707, in Advisory Comm. on Evidence Rules, Agenda Book 102, 103–06, 146–47 (Nov. 5, 2025) [hereinafter “2025 Reporter’s Memo”] (citing Edward J. Imwinkelried, The Challenge that the Advent of Artificial Intelligence (AI) Tools Poses to the Procedures for Determining the Existence of the Preliminary Facts that Condition the Admissibility of Items of Evidence, 108 Marq. L. Rev. 621, 642–46, 651–52 (2025)), https://www.uscourts.gov/sites/default/files/document/2025-11_evidence_rules_commitee_agenda_book_final.pdf [https://perma.cc/RN9D-KPFP]. ↑
    • . Fed. R. Evid. 901(b)(9). ↑
    • . Imwinkelried, supra note 3, at 642–43. ↑
    • . Imwinkelried, supra note 3, at 651–52. ↑
    • . 2025 Reporter’s Memo, supra note 3, at 105, 147. ↑
    • . Id. at 105. ↑
    • . Advisory Comm. on Evid. Rules, Minutes of Meeting of October 27, 2023, at 3, https://www.uscourts.gov/sites/default/files/2023-10_minutes_of_meeting_of_evidence_rules_committee_fall_2023_final.pdf [https://perma.cc/L6VE-SXJN]. See also 2025 Reporter’s Memo, supra note 3, at 102. ↑
    • . 2025 Proposed Amendments, supra note 2, at 109–11. The Reporter’s Memo explains that “machine learning” is “broadly defined as the capability of a machine to imitate intelligent human behavior,” citing ChatGPT as the “most famous example.” 2025 Reporter’s Memo, supra note 3, at 133. ↑
    • Rule 702 is the rule of evidence for evaluating the reliability of expert testimony: A witness who is qualified as an expert by knowledge, skill, experience, training, or education may testify in the form of an opinion or otherwise if the proponent demonstrates to the court that it is more likely than not that:(a) the expert’s scientific, technical, or other specialized knowledge will help the trier of fact to understand the evidence or to determine a fact in issue;(b) the testimony is based on sufficient facts or data;(c) the testimony is the product of reliable principles and methods; and(d) the expert’s opinion reflects a reliable application of the principles and methods to the facts of the case.Fed. R. Evid. 702. ↑
  • . 2025 Proposed Amendments, supra note 2, at 109 (footnote added). ↑
  • . Id. at 1–2; 2025 Reporter’s Memo, supra note 3, at 102. ↑
  • . See generally 2025 Reporter’s Memo, supra note 3. ↑
  • . Id. at 103–06 (citing Imwinkelried, supra note 3, at 642–46, 651–52). ↑
  • . Rule 901 is the authentication rule, and it requires the proponent of a piece of evidence to present other, supporting evidence sufficient to support a jury finding that the “item is what is claimed to be.” Fed. R. Evid. 901(a). Subsection (b) of the rule lists examples of supporting evidence that could be used for this purpose. Fed. R. Evid. 901(b). ↑
  • . This refers to the standard of proof set out in Federal Rule of Evidence 104(b): “When the relevance of evidence depends on whether a fact exists, proof must be introduced sufficient to support a finding that the fact does exist. The court may admit the proposed evidence on the condition that the proof be introduced later.” Fed. R. Evid. 104(b). ↑
  • . Imwinkelried, supra note 3, at 642–43. ↑
  • . Id. at 651–52. ↑
  • . Id. at 652. ↑
  • . 2025 Reporter’s Memo, supra note 3, at 105. ↑
  • . See Fed. R. Evid. 702, 901. ↑
  • . 2025 Reporter’s Memo, supra note 3, at 105. ↑
  • . Id. at 146–47. ↑
  • . Id. at 147. ↑
  • . Id. at 105–06, 147. ↑
  • . Id. ↑
  • . Id. at 147–48. ↑
  • . Report of the Advisory Committee on Evidence Rules, in Comm. on Rules of Prac. & Proc., Agenda Book 282, 286 (Jan. 6, 2026), https://www.uscourts.gov/sites/default/files/document/2026-01_standing_committee_agenda_book_final_0.pdf [https://perma.cc/38ZN-A6GN]. ↑
  • . Id. ↑
  • . Id. at 285. ↑
  • . Report of the Advisory Committee on Evidence Rules, in Comm. on Rules of Prac. & Proc., Agenda Book 525, 525, 531–32 (June 3–4, 2026) [hereinafter “May 2026 Committee Report”], https://www.uscourts.gov/sites/default/files/document/2026-06_standing_committee_meeting_agenda_book_final.pdf [https://perma.cc/4A66-7Z9D]. ↑
  • . Id. at 531–32. ↑
  • . Id. at 534. ↑
  • . Id. at 535. ↑
  • . Ashley Skladany, Analysis: Proposed FRE 707 Stalls—But Is That Such a Bad Thing?, Bloomberg L. (May 15, 2026, at 04:00 CT), https://www.bloomberglaw.com/product/blaw/bloomberglawnews/bloomberg-law-news/XL3CENK000000 [https://perma.cc/RG2N-76XB]. ↑
  • . See May 2026 Committee Report, supra note 32, at 530–35. ↑
  • . Many uses of Rule 901(b)(9) will not be included in this study by virtue of their not being discussed in a written opinion available on these research platforms, so the sample analyzed here is inevitably underinclusive. The purpose of this study is not to undertake an exhaustive review of all recent cases in which this issue came up, but to explore the courts’ and the parties’ use of Rule 901(b)(9) in opinions that are readily available to begin to shed light on the questions raised in the Reporter’s memo. ↑
  • . The cases were collected from Westlaw and Lexis using the following search term: “901(b)(9),” with appropriate filtering to the noted parameters. ↑
  • . See Appendix (table of cases included in study). ↑
  • . See id. ↑
  • . See id. ↑
  • . See id. ↑
  • . See id. ↑
  • . See id. ↑
  • . See id. It is certainly possible that the cases involving authentication of evidence obtained from simple machines were not available on commercial research databases because the use of Rule 901(b)(9) in those situations is straightforward and non-controversial. This essay does not suggest that those cases do not exist. Instead, it focuses on what readily available written judicial opinions indicate about the role of Rule 901(b)(9). ↑
  • . See id. The messages from the encrypted platform were allegedly decrypted and organized using AI. United States v. Gogic, No. 22-CR-493, 2025 WL 3042350, at *3 (E.D.N.Y. Oct. 31, 2025). AI was not used to create the evidence at issue in any of the other cases studied. ↑
  • . See Appendix. ↑
  • . See id. ↑
  • . See id. ↑
  • . See id. ↑
  • . 2025 Reporter’s Memo, supra note 3, at 105, 147. ↑
  • . See Appendix. ↑
  • . See id. ↑
  • . See id. ↑
  • . See id. ↑
  • . See id. ↑
  • . See id. ↑
  • . This rule permits authentication through “[t]estimony that an item is what it is claimed to be.” Fed. R. Evid. 901(b)(1). ↑
  • . See Appendix. ↑
  • . See id. ↑
  • . See id. ↑
  • . Fed. R. Evid. 901(b)(9). ↑
  • . See, e.g., Marquette Transp. Co. Gulf-Inland LLC v. Navigation Mar. Bulgarea, 598 F. Supp. 3d 467, 473 (E.D. La. 2022), aff’d sub nom. Marquette Transp. Co. Gulf-Inland, L.L.C. v. Navigation Mar. Bulgare JSC, 87 F.4th 678 (5th Cir. 2023); Locke v. Jefferson Hills Manor, No. 2:18-CV-01260, 2020 WL 5363320, at *3 (W.D. Pa. Sept. 8, 2020). ↑
  • . See, e.g., Wiseman v. Tastefully Better, No. 19-CV-1441, 2023 WL 3200229, at *3 (N.D. Ill. May 2, 2023); United States v. Gafford, 149 F.4th 1002 (8th Cir. 2025). ↑
  • . Fed. R. Evid. 901(b)(1). ↑
  • . See id. ↑
  • . Baugh v. Voyager Indem. Ins. Co., No. CV 19-14275, 2020 WL 7042859, at *4 (E.D. La. Dec. 1, 2020) (citing United States v. Lizarrago-Tirada, 789 F.3d 1107, 1110 (9th Cir. 2015)). ↑
  • . Lizarrago-Tirada, 789 F.3d at 1110. ↑
  • . United States v. Gogic, No. 22-CR-493, 2025 WL 3042350, at *7, 9–10 (E.D.N.Y. Oct. 31, 2025) (emphasis added). See also Reporter’s Memorandum on Artificial Intelligence and Proposed Rule 707, in Advisory Comm. on Evidence Rules, Agenda Book 116, 117 (May 7, 2026) [hereinafter “2026 Reporter’s Memo”], https://www.uscourts.gov/sites/default/files/document/2026-05-evidence-rules-agenda-book.pdf [https://perma.cc/9ZRP-A82V] (discussing Gogic and noting that “it can be argued that the court erred, by treating a reliability question as one of authenticity”). ↑
  • . United States v. Gafford, 149 F.4th 1002, 1006 (8th Cir. 2025) (emphasis added). ↑
  • . Magee v. Noe, No. 2:20cv183, 2023 WL 116349, at *6 (S.D. Miss. Jan. 5, 2023) (emphasis added). ↑
  • . Fed. R. Evid. 901(b)(9). ↑
  • . See Fed. R. Evid. 901(a). ↑
  • . United States v. Gogic, No. 22-CR-493, 2025 WL 3042350, at *8 (E.D.N.Y. Oct. 31, 2025). ↑
  • . United States v. Ramos-Contreras, No. 3:23-CR-135, 2024 WL 6948305, at *17 n.26 (N.D. Ala. Aug. 28, 2024) (emphasis omitted), report and recommendation adopted in part, rejected in part sub nom. United States v. Contreras, No. 3:23-CR-135, 2025 WL 3298317 (N.D. Ala. Nov. 26, 2025). ↑
  • . United States v. Pancholi, No. CR 19-20639, 2023 WL 5706197, at *4 (E.D. Mich. Sept. 5, 2023), motion for reconsideration denied, 2023 WL 5938529 (E.D. Mich. Sept. 12, 2023); see also Lorraine v. Markel Am. Ins. Co., 241 F.R.D. 534, 549 (D. Md. 2007) (providing that Rule 901(b)(9) is “particularly useful in authenticating electronic evidence stored in or generated by computers”). ↑
  • . See supra subpart II(A). ↑
  • . 2025 Reporter’s Memo, supra note 3, at 105. ↑
  • . See, e.g., Marquette Transp. Co. Gulf-Inland, L.L.C. v. Navigation Mar. Bulgare JSC, 87 F.4th 678, 685 (5th Cir. 2023); Magee v. Noe, No. 2:20CV183, 2023 WL 116349, at *5–6 (S.D. Miss. Jan. 5, 2023); see also Loma Deli Grocery Corp. v. United States, No. 20 Civ. 7236, 2021 WL 4135216, at *8 (S.D.N.Y. Sept. 10, 2021) (suggesting that an expert report would be required to authenticate an algorithm’s output—but not the underlying data). ↑
  • . See Fed. R. Evid. 702; Victor J. Gold, Chapter 10: Authentication And Identification, in 31 Fed. Prac. & Proc. Evid. § 7114 (Wright & Miller eds., 2d ed. Apr. 2026) (“[I]t seems prudent to assume that authentication of evidence under Rule 901(b)(9) does not preclude consideration of additional admissibility issues under the rules and cases relating to expert testimony.”); Gold, supra, § 7147 (“If making the assertion regarding accuracy requires specialized knowledge about the technology involved, the certification should establish that the person signing the certification qualifies as an expert in such matters.”). ↑
  • . See Fed. R. Evid. 901. ↑
  • . See Fed. R. Evid. 901(b)(9) advisory committee’s note on proposed rules (“Example (9) does not, of course, foreclose taking judicial notice of the accuracy of the process or system.”); Fed. R. Evid. 902(13) advisory committee’s note to 2017 amendment (“Nothing in the amendment is intended to limit a party from establishing authenticity of electronic evidence on any ground provided in these Rules, including though judicial notice where appropriate.”); Gold, supra note 81, § 7114 (“Where the evidence in question is the result of a process or system that is generally known and accepted as accurate, a court may take judicial notice of the foundational facts required by Rule 901(b)(9).”). ↑
  • . See Fed. R. Evid. 701 (stating that lay opinion testimony is limited to opinions “not based on . . . specialized knowledge within the scope of Rule 702”); Fed. R. Evid. 902(13) advisory committee’s note to 2017 amendment (“A challenge to the authenticity of electronic evidence may require technical information about the system or process at issue, including possibly retaining a forensic technical expert.”); Gold, supra note 81, § 7114 (“But where a process or system is less well known, the party offering the results produced by that process or system must describe that process or system and show that it produces an accurate result. Because Rule 901(b)(9) evidence commonly is the product of some sophisticated machine or process, this foundation often is established through expert testimony.”). ↑
  • . See United States v. Pancholi, No. CR 19-20639, 2023 WL 5706197, at *4 (E.D. Mich. Sept. 5, 2023) (discussing authentication of emails obtained from government email system, specifically to confirm whether the emails were sent by defendant); United States v. Sandoval-Rodriguez, No. 22-4330, 2025 WL 1098843, at *3 (4th Cir. Apr. 14, 2025) (discussing authentication of video from laundromat’s surveillance system by witness without personal knowledge of the events recorded); United States v. Gafford, 149 F.4th 1002, 1006 (8th Cir. 2025) (discussing authentication of data collected from Apple AirTag through testimony of witness “not familiar with the internal operations of the device”); Wiseman v. Tastefully Better, No. 19-CV-1441, 2023 WL 3200229, at *3 (N.D. Ill. May 2, 2023) (discussing authentication of screenshot of computer file data through testimony of lay witness). ↑
  • . Fed. R. Evid. 901(b)(9). ↑
  • . 2025 Proposed Amendments, supra note 2, at 109. ↑
  • . 2025 Reporter’s Memo, supra note 3, at 140–42. See also 2026 Reporter’s Memo, supra note 70, at 116–17 (discussing Gafford, which involved evidence about the use of an Apple AirTag, and suggesting that the prior version of proposed Rule 702 would have applied there). ↑
  • . For example, one group suggested a revision to proposed Rule 707 because its current language is “unclear when applied to the output of computers, computer software, web-based services, and the like, which are not simple scientific instruments but can produce both non-inferential and inferential outputs.” Fed. Cts. Comm. of the N.Y.C. Bar Ass’n, Comments on Proposed Federal Rule of Evidence 707 and Amendments to Rule 609, at 3 (February 13, 2026), https://www.nycbar.org/wp-content/uploads/2026/02/20221587-CmmtLttrOnFedRule707and609_FINAL.pdf [https://perma.cc/JX2M-EGP2]. Similarly, another commenter observed:

    Originally, the Evidence Committee seemed focused on developing a rule to address evidentiary challenges related to machine learning. However, the use of the term “machine-generated” in the proposed rule text implies a significantly broader application, sweeping in every type of machine imaginable—from appliances, common-place tools, and machinery, to software, mobile devices, and industry-specific equipment—most of which is readily accepted as reliable and frequently presented by lay witnesses or to lay a foundation in courts nationwide.

    Am. Ass’n for Just., Comment on Proposed Rule 707 Machine Generated Evidence 1–2 (February 16, 2026) (footnote omitted), available at: https://www.justice.org/advocacy/legal-affairs/federal-rules/fc—fre-707 [https://perma.cc/C3TW-VDJW]. ↑

  • . See Appendix. ↑
  • . See May 2026 Committee Report, supra note 32, at 531. ↑
  • . The Committee may well determine that the existing reliability scheme in Article 7 of the Federal Rules of Evidence is more than sufficient. In cases where a layperson testifies to authenticate non-AI-generated evidence obtained from an electronic device, that evidence could still be excluded under Rule 702 if specialized knowledge was required to understand how the evidence was created. Indeed, this would result in the proper exclusion of the evidence on reliability grounds, rather than under the misplaced authentication framework currently in use. In addition, as the Reporter has noted, the Federal Rules of Evidence do not contemplate a general reliability check for all evidence; they impose these barriers only for particular categories of evidence that carry special reliability concerns. See 2026 Reporter’s Memo, supra note 70, at 117–18, 157. Perhaps these cases involving layperson testimony about digital data simply fall outside the reliability rules altogether. ↑

Download PDF