Non-Party Waited Too Long to Request to Use A.I., and Did Not Provide Sufficient Details

Non-Party Waited Too Long to Request to Use A.I., and Did Not Provide Sufficient Details by Michael Berman, E-Discovery LLC.
Image: Holley Robinson, EDRM.

[EDRM Editor’s Note: The opinions and positions are those of Michael Berman.]


A non-party did not request authorization to use artificial intelligence to respond to a subpoena until it asked the court to reconsider a prior discovery order.  The court wrote: “This proposal comes too late and is far too generic to suffice.” Mi Familia Vota v. Fontes, 344 F.R.D. 496, 529 (D. Ariz. 2023). 

The Mi Familia Vota court addressed this request to use artificial intelligence that was first made—after months of disputes—as a request to reconsider an order to use keywords. The important dicta in Mi Familia Vota was: “Had the RPA [proposed to use AI] in a timely manner and described its anticipated AI-assisted search process in a sufficiently detailed manner, the Court would have been very open to it.” [emphasis added].  In my words, timely transparency could have led to a different outcome.1

In Mi Familia Vota, the Republican Party of Arizona (“RPA”), a non-party, was served with a subpoena by Plaintiffs that were challenging a State voter roll purging law. The court wrote that: “Since then—a period of over 21 months—the subpoena has generated no shortage of disputes between Plaintiffs and the RPA. The Court has repeatedly resolved those disputes in Plaintiffs’ favor, in large part due to the RPA’s failure to raise its objections in a timely or legally compliant manner.”  Id. at 501-02.

RPA’s search terms had located 61,298 documents. Id. at 508. In its brief in response to Plaintiffs’ motion to compel, the RPA argued that, “at some point after the August 7, 2023 production deadline, it “used artificial intelligence … to drastically cut down the number of responsive documents to approximately 5,000 documents. If the Court intends to grant Plaintiffs’ Motion to Compel, it should only order [the RPA] to produce the approximately 5,000 documents.” Id. at 527.

The Mi Familia Vota court found it important to provide the following context:

[O]n July 17, 2023, the Court held a lengthy discovery-dispute hearing… The main dispute was over how the RPA would conduct its search for responsive documents…. The parties had already met and conferred on that subject ad nauseam, including during the nine-month period after the issuance of the October 2022 order compelling the RPA to comply with the subpoena. At no point during those meet-and-confer discussions (or during the resulting discovery-dispute briefing and argument process) did the RPA raise the possibility of using AI tools to search for responsive documents. At the conclusion of the July 17, 2023 hearing, the Court ruled that the RPA “shall run the search terms as outlined in [the RPA’s counsel’s] May 10, 2023 email (except the search term CUR* shall be changed to cure, curing and cured) and shall produce the documents responsive to the search terms and/or disclose a privilege log.” … Thus, by now asking for permission to use AI tools to search for responsive documents, the RPA is effectively seeking to relitigate the issues raised and resolved during the July 17, 2023 hearing.

Id. at 527-28 (emphasis added).

The Mi Familia Vota court wrote that reconsideration was justified only to correct a manifest error, which was not shown here.  It wrote:

Applying these standards, the RPA has not established a valid basis for reconsideration. The RPA had ample time to work cooperatively with Plaintiffs in an attempt to formulate an agreed-to AI protocol for identifying responsive documents. Instead, the RPA initially refused to produce anything, then seemed to agree to use keyword searches in an attempt to locate responsive documents, then retracted that offer. When the parties eventually raised that dispute for judicial resolution, the RPA still did not suggest using AI tools to assist in the search. Had the RPA done so in a timely manner and described its anticipated AI-assisted search process in a sufficiently detailed manner, the Court would have been very open to it. Bridgestone Americas, Inc. v. Int’l Business Machines Corp., 2014 WL 4923014, *1 (M.D. Tenn. 2014) (allowing responding party “to switch horses in midstream” by “us[ing] predictive coding in reviewing something over two million documents for responsiveness,” where the adverse party “opposed this request as being an unwarranted change in the original case management order … and on the grounds that it is unfair to use predictive coding after an initial screening has been done with search terms,” but emphasizing that “openness and transparency in what Plaintiff is doing will be of critical importance,” that “Plaintiff has advised that they will provide the seed documents they are initially using to set up predictive coding,” and that “[t]he Magistrate Judge expects full openness in this matter”). Indeed, the Court has repeatedly expressed its interest in mechanisms for reducing the RPA’s compliance-related burdens. But that ship has now sailed. It is simply too late in the process for the RPA to belatedly, and without justification, identify an entirely different process for locating documents that are responsive to a subpoena served over 21 months ago.

Id. at 528 (emphasis added). 

The court noted that, during the “compliance period” following the July 17, 2023, discovery hearing, “the RPA again made no effort to communicate with Plaintiffs (or the Court) about the possibility of using AI as an alternative search method.  It was only in the days and weeks after the compliance deadline expired that the RPA first raised that possibility.” Id. at 528-29.  Further, the court added that “the RPA has still provided only cursory details about what the AI search process would involve.” Id. at 529.

[T]he RPA again made no effort to communicate with Plaintiffs (or the Court) about the possibility of using AI as an alternative search method.  It was only in the days and weeks after the compliance deadline expired that the RPA first raised that possibility.

Id. at 528-29.

Belatedly proposing a generic and undefined use of AI, after agreeing to other search methodologies that were incorporated into a court order, doomed the use of artificial intelligence in this case.

The Sedona Conference has long stated that responding parties are best situated to evaluate the procedures, methodologies, and technologies that are appropriate for producing their own ESI.  The Sedona Principles, Third Edition: Best Practices, Recommendations & Principles for Addressing Electronic Document Production, 19 Sedona Conf. J. 1, 118 (2018).  In Mi Familia Vota permission was required due to a prior court order.

At least since publication of the Sedona Conference’s “Cooperation Proclamation,” cooperation and transparency have been the hallmarks of electronic discovery.2 In 1850, Abraham Lincoln wrote: “Persuade your neighbors to compromise whenever you can…. As a peacemaker, the lawyer has a superior opportunity of being a good man [or woman].”  Cooperation as a requirement was recognized in Maryland in 1941.  J. Pike and J. Willis, The New Maryland Deposition and Discovery Procedure, 6 Md. L. Rev. 4, 32 (1941); Rodriguez v. Clarke, 400 Md. 39, 60 (2007).

Mi Familia Vota makes it clear that those principles apply equally to the use of A.I.


Notes

  1. The request to use AI came in response to a motion to compel.  Id. at 527. The court construed the request as a motion for reconsideration. ↩︎
  2. Failure to Cooperate Leads to Judicially-Imposed ESI Protocol (Jan. 15, 2025); Failure to Confer Regarding ESI Protocol Operates Against Failing Party (Dec.28, 2024); Parties Ordered to Cooperate in Good Faith and Develop a “Discovery Plan” (May 25, 2024); Ethical Contours of Discovery Sanctions and the Duty to Cooperate (Sep. 7, 2023); Be Careful What You Agree To, But the Duty to Cooperate May Provide an Escape Hatch (Part II) (Nov. 11, 2022); Historical ESI Highlights – Part VIII – Cooperation and The Cooperation Proclamation (May 31, 2022); Failure to Cooperate Costs Clients Money (Mar. 8, 2022). ↩︎

Assisted by GAI and LLM Technologies per EDRM’s GAI and LLM Policy.

Author

  • Michael Berman

    Michael Berman is a practicing lawyer and an adjunct faculty member at the University of Baltimore School of Law.  He has published extensively, including as the editor-in-chief and a contributing author in “Electronically Stored Information in Maryland Courts” (Maryland State Bar Ass’n. 2020), co-editing two American Bar Association books on electronic discovery, as well as co-authoring law review and other articles regarding electronically stored information (“ESI”).  He has presented widely in venues ranging from local to National events and served as a Court Appointed ESI Discovery Supervisor for ESI Protocol.

    View all posts Owner, E-Discovery LLC