Will Courts “Search, Forward” with eDiscovery Case Law on AI?

[EDRM Editor’s Note: This article was first published here on August 8, 2025, and EDRM is grateful to Trusted Partner HaystackID for permission to republish. EDRM is happy to amplify our Trusted Partners news and events.]

HaystackID Editor’s Note: In this contributed article, Phil Favro draws sharp parallels between today’s artificial intelligence (AI) adoption anxieties and the technology-assisted review (TAR) debates of the early 2010s, offering a measured perspective on where eDiscovery jurisprudence may head. His analysis cuts past the current hype cycle to examine substantive questions about defensibility, cooperation, and transparency that courts will inevitably confront. The article demonstrates how existing case law may guide courts facing novel AI discovery disputes, even as parties forge ahead without explicit judicial blessing. Favro’s measured assessment acknowledges both the promise and uncertainty surrounding generative AI’s role in document review workflows. His conclusion that practitioners should focus on defensible processes rather than waiting for definitive case law reflects hard-won experience in a field where technology often outpaces legal precedent.

Lawyers and other legal professionals are anxious for case law on the interplay between artificial intelligence (AI) and eDiscovery. To date, there have been few cases about AI, aside from decisions about hallucinations and other misdirected uses of AI. To be sure, there are a handful of cases addressing the use of AI in the context of discovery. Still, courts have yet to address fundamental issues regarding the use of generative AI (GenAI) in connection with the search, identification, and review process for discovery.

This period of anticipation is reminiscent of the early 2010s, when clients and counsel anxiously waited for cases that could offer judicial imprimatur of technology-assisted review (TAR). And just as the TAR case law landscape from that time, it’s anyone’s guess when the courts will rule on issues such as the reasonableness of search prompts, recall targets, validation procedures, and other matters impacting discovery and AI.

In the meantime, exploring past eDiscovery developments may yield guidance on how case law on AI and discovery issues may ultimately play out. One of the most significant pieces of thought leadership from eDiscovery past—Search, Forward, authored by the Honorable Andrew Peck (ret.) and published by Law Technology News in October 2011—is particularly noteworthy in this regard and may offer guidance regarding the future promulgation of AI case law.

Search, Forward on TAR

In Search, Forward, Judge Peck captured the eDiscovery field’s zeitgeist at the dawn of TAR. He recapped the ills of manual review and search terms while spotlighting the potential offered by TAR, which he generally referred to as “computer-assisted coding” or “predictive coding.” Citing academic scholarship and studies, Judge Peck suggested that TAR had the potential to improve ESI search and review, both in terms of identifying relevant documents and cost-effectiveness.

Judge Peck also explored possible challenges to using TAR and offered suggestions for better ensuring that use of the technology would be defensible. Presciently, he reasoned that proof of a “valid ‘process’” and “quality control testing” would be important factors in meeting challenges to defensibility. He also opined that cooperation and transparency would factor into the producing parties’ implementation of TAR.

Finally, Judge Peck observed that many attorneys were concerned about deploying TAR without a judicial opinion approving its use in discovery. He suggested that such apprehensions were misguided. By analogy, the judge mentioned that lawyers were using search terms in 2011, even though courts had not expressly approved them and had criticized their use. He also indicated that some lawyers had apparently been using TAR despite the lack of case law. To alleviate concerns over an absence of case law, Judge Peck concluded Search, Forward by encouraging counsel to “rely on this article as a sign of judicial approval” for the use of TAR in discovery.

Moore v. Publicis Groupe and TAR Use Issues

Within a few months of Search, Forward being published, Judge Peck issued the first formal judicial opinion—Moore v. Publicis Groupe, 287 F.R.D. 182 (S.D.N.Y. 2012)—endorsing the use of TAR in discovery. Moore was the first of several decisions that generally approved the use of TAR, followed by other cases that began addressing more nuanced questions about TAR implementation, such as transparency, cooperation, blending search terms with TAR, seed set disclosure, use protocols, recall targets, elusion testing, and other validation measures. Moore, which was plagued by satellite litigation after the court entered the parties’ stipulated TAR-use protocol, also previewed the contentiousness that has often unfolded in lawsuits where producing parties have signaled their intention to use TAR.

Disagreements over TAR use issues have frequently led to motion practice and judicial divergence over how parties should use the technology. A key dispute (among many others) focuses on the validity of the producing party’s process, i.e., the defensibility issue Judge Peck emphasized in Search, Forward. Process validity could refer to any number of items, including validation and other “quality control” measures. And yet, many quarrels regarding process validity involve questions over cooperation and transparency. Courts have not helped quell the contentiousness on this issue, with judges reaching different conclusions on the role of cooperation and transparency. The differences in the jurisprudence on this issue are best exemplified by Livingston v. City of Chicago, No. 16 CV 10156, 2020 WL 5253848 (N.D. Ill. Sept. 3, 2020) and In re Valsartan, Losartan, & Irbesartan Prods. Liab. Litig., 337 F.R.D. 610 (D.N.J. 2020).

In Livingston, the court declined to order the defendant to negotiate a transparent TAR-use protocol with plaintiffs. Reasoning that “Plaintiffs’ insistence that the City must collaborate with them to establish a review protocol and validation process has no foothold in the federal rules governing discovery,” the court found that the defendant had no obligation to disclose its TAR process to or otherwise negotiate with plaintiffs on its workflow, including validation measures and metrics. In contrast, Valsartan observed that “the backbone of TAR’s use is transparency and collaboration” and directed a defendant to produce 5,000 nonresponsive documents to plaintiffs for their review as part of the TAR validation process. Counterintuitively, Valsartan reached a completely different result from Livingston on the roles of cooperation and transparency, suggesting they were indispensable aspects of a TAR workflow. Livingston and Valsartan typify the different approaches courts have taken regarding cooperation and transparency, along with other TAR-use issues.

In summary, if Search, Forward captured the eDiscovery field’s desire to move forward with TAR, Moore and its progeny unleashed contentiousness over its implementation and then dissonance in subsequent TAR case law.

Search, Forward on AI?

In 2025, we are at a somewhat similar inflection point regarding eDiscovery technology to where things stood in 2011 when Judge Peck authored Search, Forward.

Like TAR in 2011, GenAI technologies have just burst on the scene, and there is a lot of hype regarding their potential. Just as in 2011, eDiscovery professionals are also waiting for courts to issue an opinion declaring that AI—like TAR in 2011—is “better than the available alternatives, and thus should be used in appropriate cases.” Moore, 287 F.R.D. at 191. Finally, it’s apparent from anecdotal reports—just like TAR in 2011—that lawyers are using AI to handle document identification and review, among other items in discovery. In certain instances, clients appear to be driving this trend. They are looking for a technological edge that AI might offer, and potentially at a more cost-effective rate.

But that’s where the similarities end. Unlike 2011, eDiscovery professionals today have TAR, an advanced technology that provides efficient and cost-effective methods to search, identify, and review ESI. It’s true that parties in many instances still rely exclusively on search terms, but many lawyers instead use TAR—particularly “TAR 2.0” active learning methods and workflows—which has generally fulfilled Search, Forward’s promise of being “cheaper, better, and faster.” Moreover, the promise that generative AI review will be able to match TAR on these three points remains uncertain. The answer will come in time, once we have the benefit of hindsight (as we now do with TAR).

Another key difference between 2011 and 2025 is on the case law front. In 2011, no TAR case law had been issued, and there were relatively few search term decisions. In contrast, counsel and the courts now have 14 additional years of ESI jurisprudence. Many of those cases have focused on TAR, search terms, and other aspects of ESI that may very well be analogous to search and review issues involving AI. For example, the different lines of authority regarding cooperation and transparency, as memorialized in Livingston and Valsartan, could be instructive regarding how courts may address similar issues with AI.

In addition, courts have issued case law addressing the interplay between AI and discovery. For example, the court in Concord Music Grp., Inc. v. Anthropic PBC, No. 24-CV-03811-EKL (SVK), 2025 WL 1482734 (N.D. Cal. May 23, 2025), protected certain lawyer prompts and outputs from discovery, holding they were protected as opinion work product. In that same order, Concord Music also found that disclosing certain prompts and outputs in discovery did not constitute a subject matter waiver. This is just one of the various decisions discussing the interplay between AI and eDiscovery.¹

The similarities and differences between 2011 and 2025 collectively offer insight into why lawyers and clients have ventured into AI and are willing to try the technology despite the lack of case law approving AI’s use in discovery. Given that trend, it’s apparent that a 2025 analogue to Search, Forward—essentially an advisory opinion—is unnecessary to ensure widespread adoption of AI.

Nevertheless, eDiscovery professionals would be well advised to follow Judge Peck’s past admonition to ensure their AI search and review process is defensible and consider how they wish to handle issues like cooperation and transparency. As Moore and its progeny make clear, parties should expect challenges to the defensibility of their process, and there is no guarantee how courts will ultimately determine the issues.

Read the original article here.

Notes

1. See, e.g., Ohio ex rel Yost v. Jones, No. 2:22-CV-2700, 2025 WL 1218088 (S.D. Ohio Apr. 28, 2025) (rejecting defendant’s assertion that his use of an AI “agent” constituted a “diligent search” and ordering defendant to to “review his records and produce all [responsive] documents in his possession, custody, or control.” (emphasis in original); Sheets v. Charlotte Cnty., No. 2:24-CV-958-JES-KCD, 2025 WL 1069271 (M.D. Fla. Apr. 9, 2025) (suggesting defendant could use AI to prepare discovery responses so long as he evaluated “any factual and legal references generated by AI.”). ↩︎

SOURCE: HaystackID
Assisted by GAI and LLM Technologies per EDRM GAI and LLM Policy.