ESI Protocol a/k/a Discovery Plans – E-Discovery LLC https://www.ediscoveryllc.com Mediation of E-Discovery Disputes Sat, 10 Oct 2026 23:59:42 +0000 en-US hourly 1 A Unique A.I. Protective Order https://www.ediscoveryllc.com/a-unique-a-i-protective-order/ Fri, 09 Oct 2026 20:52:43 +0000 https://www.ediscoveryllc.com/?p=5633 […]]]> A stipulated AI Protective Order in Dean & Deluca Brands, Inc. v. Fire Brands Innovation, LLC, 2026 WL 2905875 (S.D.N.Y.  Sep. 29, 2026)(emphasis added), stated:

A producing Party may utilize technology-assisted review, predictive coding, generative artificial intelligence, analytics, keyword searching, clustering, email threading, concept searching, active learning, or other automated or algorithmic tools in connection with the collection, processing, search, review, redaction, privilege review, or production of documents and ESI. The producing Party shall have sole discretion regarding the selection, configuration, implementation, and use of such tools and methodologies. No producing Party shall be required to disclose the identity, version, prompts, workflows, training materials, validation procedures, review protocols, quality control measures, precision rates, recall rates, elusion testing, or other details concerning the use of such tools absent a showing of good cause by the receiving Party or an order of the Court. The use of such technologies shall not, by itself, provide a basis for challenging the adequacy of a Party’s collection, review, privilege determinations, or production.

By comparison, please see Three More A.I. Protective Orders (Oct. 9, 2026), and the posts listed therein.

]]> Three More A.I. Protective Orders https://www.ediscoveryllc.com/three-more-a-i-protective-orders/ Fri, 09 Oct 2026 08:30:52 +0000 https://www.ediscoveryllc.com/?p=5617 […]]]> I have written a series of posts suggesting that AI protective orders are becoming the standard of care.  See A.I. Protective Order by Consent in Google Case (Sep. 13, 2026).

The jointly requested A.I. protective order in Hoffman v. Univ. of Nebraska Medical Ctr., 2026 WL 2969621 (D. Neb. Oct. 2, 2026), states:

Artificial Intelligence. Qualified Recipients shall not, without the Producing Party’s express written consent, upload, submit, or otherwise provide Confidential Discovery Material (or any information extracted therefrom) or any deposition testimony taken in this matter to any open-source artificial intelligence platform or service including, but not limited to, platforms or services that use or assemble uploaded information to train or improve machine-learning models. This prohibition applies regardless of whether the artificial intelligence platform or service represents it will maintain the confidentiality of information submitted by the user. The parties and their counsel are not prohibited from uploading, submitting, or otherwise providing Confidential Discovery Material (or information extracted therefrom) or any deposition testimony taken in this matter to a closed-source artificial intelligence platform or service, including Harvey, Westlaw, LexisNexis, Servient’s AI Canvas, or any other closed-source artificial intelligence platform or service agreed to by the parties.

In ANR Pipeline Co., LLC v. B&B Metal Processing Co., Inc., 2026 WL 2970050 (E.D. Wisc. Oct. 2, 2026), the court wrote:

On September 22, 2026, the parties filed a joint motion for the entry of a protective order, with a stipulated proposed order…. The parties request that the Court enter such an order to “prevent public disclosure of nonpublic confidential technical, commercial, or business information that may be exchanged in discovery.” … Federal Rule of Civil Procedure Rule 26(c) allows for an order “to protect a party or person from annoyance, embarrassment, oppression, or undue burden or expense” including “requiring that a trade secret or other confidential research, development, or commercial information … be revealed only in a specified way.” Fed. R. Civ. P. 26(c)(1)(G); see also Civ. L.R. 26(e).

Protective orders are an exception to the general rule that pretrial discovery must occur in the public eye…. Litigation must be conducted in public to the maximum extent consistent with respecting trade secrets … and other facts that should be held in confidence….

Nonetheless, the Court can enter a protective order if the parties have shown good cause and that the order is narrowly tailored to serve that cause…. The Court can find that even broad, blanket orders are narrowly tailored and permissible when it finds that two factors are satisfied: (1) that the parties will act in good faith in designating the portions of the record that should be subject to the protective order; and (2) that the order explicitly allows the parties to the case and other interested members of the public to challenge the sealing of documents….

The Court finds that the parties have requested the protective order in this action in good faith. The parties report that this case will entail the disclosure of confidential technical information about the parties…. Thus, the Court is satisfied that there exists a sufficient basis for the requested protective order. [cleaned up].

However, the court modified the parties’ request to establish parameters for the use of A.I. as it pertains to confidential documents.  Id. at *1.  The order states:

A receiving party may use AI tools in connection with discovery materials produced in this action only if the AI tool is operated in a secure environment and is subject to enforceable contractual terms that (a) prohibit the provider from retaining, using, disclosing, selling, or using the materials to train, improve, or develop any AI system; (b) prohibit the materials from being made publicly available or accessible to any other customer or third party; and (c) require commercially reasonable administrative, technical, and physical safeguards to protect the confidentiality of the materials. Discovery materials, whether designated CONFIDENTIAL or ATTORNEYS’ EYES ONLY, shall not be entered into any publicly available or consumer-based AI platform. Any use of AI tools must remain consistent with this Order and with the requirement that discovery materials be used solely for purposes of this litigation.

Id. at *5.

A protective order was also entered by consent in U.S.A. v. Amanuel, 2026 WL 2964210 (S.D.N.Y. Oct. 1, 2026):

Restrictions on Use of Artificial Intelligence Tools to Process Disclosure Materials.

a.  “AI tool” means any automated system that uses machine learning, or similar techniques to process inputs and generate outputs, recommendations, or predictions, including but not limited to large language models, generative AI services, or AI-assisted software tools, whether cloud-based or otherwise.

b.  The defense team will not submit Disclosure Materials to a publicly accessible version of an AI tool (e.g., ChatGPT, Claude, Grok, Gemini, Copilot, Perplexity, Midjourney, Canva). The defense team will not submit Disclosure Materials to an AI tool for which the AI tool provider or vendor (e.g., OpenAI, Anthropic, Microsoft, Google) uses submitted data for its own model training, allows submitted data to be publicly accessed, discloses submitted data to third parties, or, except as specified below, retains submitted data. Publicly accessible AI tools pose unique risks to the security and integrity of Disclosure Materials, given the practical inability to claw back or delete data once it has been incorporated into a model.

c. The defense may submit Disclosure Materials to an AI tool (i) that maintains the confidentiality of Disclosure Materials; (ii) that will not expose any Disclosure Materials to third parties not authorized under this Protective Order to receive such materials; (iii) that will not retain or use any Disclosure Materials for model training; and (iv) all Disclosure Materials will be removed from the AI tool(s) at the conclusion of this case.

Some prior posts are:

This blog was initially posted on  Electronic Discovery Reference Model.

UPDATE:

White v. Chime Financial, Inc., 2026 WL 3000904 (W.D. Wash. Oct. 6, 2026)(stipulated):

USE OF ARTIFICIAL INTELLIGENCE
Should a Party elect to utilize artificial intelligence (“AI”) to assist in reviewing, analyzing, or summarizing discovery material produced or exchanged in this litigation, the Party shall take all necessary steps to ensure that the application, service, or analytical software is fully containerized. “Fully containerized,” as used in this Order, means an AI tool that does not retain the substance of a prompt or documents reviewed, share the substance of a prompt or documents reviewed for training of large language models, or use the substance of a prompt or documents reviewed in any other matter or inquiry other than this litigation. For the avoidance of doubt, this restriction expressly applies to the use of non-containerized advanced large language models (such as publicly available large language models), generative AI tools, and other advanced AI systems, including, but not limited to, OpenAI, Gemini, LLAMA, MidJourney, and Stable Diffusion, but this provision does not limit the use of services leveraging the technology underlying these generative AI tools in a fully containerized environment, including, but not limited to, Copilot Harvey, Relativity aiR, DISCO Cecelia, Everlaw AI, Syllo, Alchemy, Lexis+ AI, Epiq AIDA, Lighthouse AI, Westlaw AI, ChatGPT Enterprise, Microsoft 365 Copilot Enterprise, Claude Enterprise, NotebookLM Pro and Ultra, and Gemini Pro and Ultra. Before submitting any Protected Material to a fully containerized AI tool, a receiving party shall ensure that it (or its vendor) can delete all such Protected Material from the AI tool at the conclusion of this matter, including any derivative information stored within the tool. The obligations and restrictions of this paragraph apply even where the Protected Material has been anonymized.

U.S.A.  v. Lockhart, 2026 WL 3003727 (S.D.N.Y.  Oct. 6, 2026)(by consent):

Restrictions on Use of Artificial Intelligence Tools to Process Disclosure Materials.

a. “AI tool” means any automated system that uses machine learning, or similar techniques to process inputs and generate outputs, recommendations, or predictions, including but not limited to large language models, generative AI services, or AI-assisted software tools, whether cloud-based or otherwise.

b. The defense team will not submit Disclosure Materials to a publicly accessible version of an AI tool (e.g., ChatGPT, Claude, Grok, Gemini, Copilot, Perplexity, Midjourney, Canva). The defense team will not submit Disclosure Materials to an AI tool for which the AI tool provider or vendor (e.g., OpenAI, Anthropic, Microsoft, Google) uses submitted data for its own model training, allows submitted data to be publicly accessed, discloses submitted data to third parties, or, except as specified below, retains submitted data. Publicly accessible AI tools pose unique risks to the security and integrity of Disclosure Materials, given the practical inability to claw back or delete data once it has been incorporated into a model.

*2 c. The defense may submit Disclosure Materials to an AI tool (i) that maintains the confidentiality of Disclosure Materials; (ii) that will not expose any Disclosure Materials to third parties not authorized under this Protective Order to receive such materials; (iii) that will not retain or use any Disclosure Materials for model training; and (iv) all Disclosure Materials will be removed from the AI tool(s) at the conclusion of this case.

d. The Government will not submit any materials disclosed by the defense pursuant to Federal Rules of Criminal Procedure 26.2, 32.1 or other defense disclosure obligations (“Defense Materials”) to a publicly accessible version of an AI tool. The Government will not submit Defense Materials to an AI tool for which the AI tool provider or vendor uses submitted data for its own model training, allows submitted data to be publicly accessed, discloses submitted data to third parties, or, except as specified below, retains submitted data.

e. The Government may submit Defense Materials to an AI tool (i) that maintains the confidentiality of Defense Materials; (ii) that will not expose any Defense Materials to third parties not authorized under this Protective Order to receive such materials; (iii) that will not retain or use any Defense Sealed Materials for model training; and (iv) all Defense Materials will be removed from the AI tool(s) at the conclusion of this case.

U.S.A. v. Santos, 2026 WL 3003653 (Oct. 6, 2026)(same, by consent); U.S.A. v. Robinson, 2026 WL 3003677 (Oct. 6, 2026)(same).

U.S.A. v. Peplow, 2026 WL 3026474 (E.D. Cal. Oct. 8, 2026)(stipulation):

Artificial Intelligence Tools
15. No person or entity authorized to have access to protected material under the terms of the requested order shall input, transmit, upload, submit, quote, feed, process, generate output from, or otherwise expose any protected material received pursuant to the requested order to any artificial intelligence (“AI”) tool absent the prior written consent of the government or an order of the Court following a properly noticed motion to which the government has an opportunity to respond, unless that AI tool—

a) is an enterprise-grade platform that counsel for the defendant has licensed;
b) is subject to a binding written agreement that (i) requires the provider to keep all user-supplied data strictly confidential, and (ii) expressly prohibits the provider from using such data for training, fine-tuning, product improvement, or any purpose other than providing the contracted-for services; and
c) employs technical and organizational security measures reasonably designed to prevent any unauthorized access, disclosure, or use of protected material.

16. The obligations and restrictions of this section apply even where the data or the protected material has been anonymized. “AI tool” means any automated system that uses statistical modeling, machine learning, or similar techniques to process inputs and generate outputs, recommendations, or predictions, including but not limited to large language models, generative AI services, or AI-assisted software tools, whether cloud-based or otherwise.

17. By participating in this agreement, counsel for the defendant certifies: (i) they have implemented reasonable measures to ensure, if using any AI tool, it will maintain the protected discovery’s confidentiality as required by this Agreement and (ii) they will ensure all protected discovery is deleted from the tool once this case concludes.

18. Under no circumstances will any member of the defense team submit protected material to a publicly accessible AI system that retains and uses submitted data to train models. Such AI tools pose unique risks to the security and integrity of protected material, given the practical inability to claw back or delete data once it has been incorporated into a model.

U.S.A. v. Sarris, 2026 WL 2937892, at *2 (S.D.N.Y.  Sep 30, 2026)(by consent).

]]> Court Ordered a Framework Where Parties Were Unable to Agree on an ESI Protocol https://www.ediscoveryllc.com/court-ordered-a-framework-where-parties-were-unable-to-agree-on-an-esi-protocol/ Thu, 24 Sep 2026 15:33:18 +0000 https://www.ediscoveryllc.com/?p=5591 […]]]> In Tudor v. State Farm General Ins. Co., 2026 WL 2668435 (N.D. Cal. Sep. 10, 2026), the parties were unable to agree on an ESI Protocol.  The court established a framework for resolving the dispute.  I suggest that, in doing so, the court provided a strong incentive for the parties to resolve their disputes.

The court entered the following order:

The Parties are ORDERED to continue to reasonably and promptly communicate and meet and confer to submit a Joint Proposed ESI Protocol by no later than September 23, 2026. See Litton v. Roblox Corp., 2026 WL 1483493, at *4 (N.D. Cal. May 27, 2026). To the extent the Parties have remaining disputes over the ESI Protocol, the Parties SHALL file by no later than September 23, 2026: (1) a Joint Redline of the Joint Proposed ESI Protocol (compiled jointly in a single document showing the Parties respective competing proposed language next to each other at each disputed location and comparing the proposed ESI Protocol against the Court’s Model ESI Protocol), (2) declarations from counsel explaining the substantive differences between the Proposed Joint ESI Protocol from the Model ESI Protocol, and (3) a Further Joint Discovery Letter Brief (complying with the Court’s Standing Discovery Order) which sets forth the Parties’ arguments on the issues remaining in dispute as to the specific competing disputed language (including why that Party’s proposed edits should be adopted and not the competing language from the other Party). Any additions the Parties do not agree upon SHALL be denoted in the Joint Redline with different assigned colors for the Court’s ease of reference; one Party’s edits shall all be in green, the other Party’s shall all be in red. To the extent any edits from either Party do not have competing language from the other Party, the Court will construe such edits as unopposed.

Id. at *4 (emphasis in original).

The court added:

  • “As noted at the hearing, the Court cautions against excessive or unnecessary deviations from the Model ESI Protocol.”
  • “The Court has broad authority and discretion to assist the Parties in resolving ESI disputes.”
  • If “the parties are unable to come to an agreement regarding all terms contained within an ESI discovery agreement, the Court will, in its discretion, assist the parties in doing so.”
  • “As part of our oversight responsibilities in discovery we have the discretion to modify ESI search term parameters.”

Id. (citations omitted).

]]> Unsupervised Self Collection—The Fox Guarding the Chicken Coop—Leads to Discovery on Discovery https://www.ediscoveryllc.com/unsupervised-self-collection-the-fox-guarding-the-chicken-coop-leads-to-discovery-on-discovery/ Sat, 19 Sep 2026 14:36:58 +0000 https://www.ediscoveryllc.com/?p=5569 […]]]> The plaintiff in EEOC v. Yas Caribe, Inc., 2026 WL 2593286 (D.P.R. Sep. 2, 2026), alleged “that defendants subjected Ms. Wilkerson to a hostile work environment and constructively discharged her after an incident involving Khalid Yassin, the pharmacy manager.”  Id. at *9.  Defendants denied the allegations.  After outlining the discovery disputes, the court wrote:

What follows is a somewhat lengthy Opinion, which the Court deemed prudent to write in order to reaffirm the duties incumbent on parties and their attorneys during the discovery process, especially in the context of electronic discovery.

In Yas Caribe, eight requests for production and four interrogatories were in dispute.  The issue that this blog post focuses on is unsupervised self-collection by Defendants.  “[T]he EEOC highlights that several of its document requests entailed the collection of ESI, but maintains that defendants’ collection efforts were unsupervised by counsel and demonstrably incomplete.”

WERE DEFENDANTS’ SELF-COLLECTION EFFORTS SUFFICIENT?

The Parties’ Contentions

EEOC challenged the sufficiency of Defendants’ search for documents.  Abdullah Yassin admitted that he failed to preserve ESI and failed to review his own the electronic devices, including email and texts, or those of Khalid Yassin, the alleged harasser.  Further, he had not directed any of his employees to do so.  And, the EEOC claimed “that deposition testimony revealed that relevant personnel files and employment-related documents may be stored in the offices of a non-party entity owned and operated by Abdullah Yassin, called Yassin Unlimited Drug Corp. d/b/a Droguería San Juan,” and none of this stored information had been disclosed. Id. at *5.

The EEOC sought two types of relief. First, it asked that “Defendants be required to amend all of their Responses to EEOC’s [document requests] to clearly state: (1) whether responsive documents exist; (2) what steps were taken to locate responsive documents; (3) whether any documents are being withheld.”  Second, it asked for forensic inspection of the cell phones of Abdullah Yassin and Khalid Yassin, and a re-opening of discovery.

Defendants responded that Abdullah Yassin testified that he had not deleted anything and had no responsive documents.  They asserted that the messages EEOC relied on to show insufficient production had eventually been produced.  Further “defendants produced a ‘Messages Search Certification’ subscribed by Abdullah Yassin to the effect that he personally supervised Khalid Yassin’s search of his cellphone for responsive messages exchanged with Ms. Wilkerson, and further produced three responsive pages.”  Finally, in what I suggest is an incorrect position:

[D]efendants’ attorney argued in response to the motion to compel that directly supervising any search for responsive documents in his client’s files would make him a “potential witness” and would be “ethically inappropriate.”

The EEOC challenged the sufficiency of the “Messages Search Certification” and it had obtained the messages independently from Ms. Wilkerson, not from Defendants. Id. at *10-11.

Discovery on Discovery

There was a threshold issue.  The court described the EEOC’s position as a request for “discovery on discovery.”  This is sometimes called “satellite discovery” or “culpability discovery.”  The court enunciated several principles governing that process.

“Discovery on discovery has been described broadly as ‘any discovery seeking information regarding a party’s preservation, collection, and retention efforts.’ …  Although some trial courts have considered these types of discovery requests to be outside the scope of Rule 26(b)(1), others have not.” Id. at *6 (citations omitted).  The court continued:

In any case, the general structure of the discovery process drawn into the Federal Rules of Civil Procedure strongly implies that such discovery should be the exception, not the rule. “The federal discovery system in some ways resembles an ‘honor system,’ ” where the party seeking discovery “must generally ‘rely on the representations of the producing party or its representative that it is producing all responsive, relevant, and non-privileged discovery….’  Casting doubt on a responding party’s efforts to provide full and complete discovery responses is by no means taboo, but anyone attempting to obtain a remedy on that ground must overcome the default assumption of good-faith compliance. This is even more so when the attempted second-order discovery targets a person’s electronic systems or devices….”

Id. (emphasis added; citations omitted).  Here, however, the threshold was crossed.  For more information about “discovery on discovery,” please see the citations at the end of this post.

Defendants’ Efforts Were Insufficient

The court wrote: “Defendants’ initially deficient response to the document request and Abdullah Yassin’s deposition testimony as to the inadequacy of his search raise substantial doubts as to defendants’ compliance with their discovery obligations.”

While a single missing text message does not normally justify discovery on discovery, id. at *11, here there was also testimony that Abdullah Yassin had failed to preserve and collect documents related to the alleged harassment, had failed to search his own email and text messages, had failed to search the electronic devices of relevant custodians, including the alleged harasser, and had not instructed anyone else to do so.  Id. at *11.  Further, while the EEOC had suggested search terms that the court viewed as reasonable, “Defendants did not describe what, if any, search terms they used to find responsive documents….” Id.

Self-Collection Should Be Supervised by Counsel

With limited exception, the requirement that an attorney supervise a client’s collection is well established:

There is contrary authority, which may be limited to a situation where there is no indication that relevant information has gone missing. Maker’s Mark: A Different View of Self Collection (Aug. 3, 2021).

Further, there are defensible ways to conduct self-collection.  “Self-Collection” May Be Reasonable Using Ralph Losey’s Dual-Protection System (Jan. 19, 2024). However, that method was not used in Yas Caribe.

Defendants’ Attorney’s Argument That Supervising a Search Would be Improper

Defendants’ attorney in Yas Caribe wrote:

It is my understanding that being with a witness while s/he searches for the requested records, explaining to the witness what is requested, verifying that the information is pertinent to the request, and asking the witness to produce everything he has, complies with a “reasonable inquiry” requirement. Otherwise, a counsel’s active search for those records will make a counsel a potential witness if a client misses a text message or for any other reason. I believe that my discussing every question or discovery request with my witnesses, while they are answering, producing documents, or giving instructions to their secretaries or assistants to look for and produce exactly what was asked for, is involvement that complies with Rule 26(g)…. And how would I remember what I saw? How would I know if they are showing me everything, or if messages were deleted, or if they have changed telephones? … Still, I believe it is inappropriate for me to get involved in the witnesses’ telephones, computers, etc., beyond what I already have.

As noted above, he argued that supervising his client’s collection would make him a “potential witness” and that it would be “ethically inappropriate.”  Id. at *6.  In my opinion—and more importantly, the court’s opinion—he was incorrect.

Supervising a Client’s Search Does Not Make an Attorney a Witness

In response to the argument that supervision would make the attorney a witness, the court wrote: “The Court first notes that neither supervising nor assisting a clients’ document search converts an attorney into a witness in a case. In fact, such supervision may be required in appropriate circumstances….  Indeed, the existence, relevance, and discoverability of ESI is one such circumstance that counsels in favor of attorneys exercising diligent supervision over their clients’ efforts to comply with discovery requests….”

Quoting a prior decision,[1] it added: “Attorneys have a duty to oversee their clients’ collection of information and documents, especially when ESI is involved, during the discovery process. Although clients can certainly be tasked with searching for, collecting, and producing discovery, it must be accomplished under the advice and supervision of counsel, or at least with counsel possessing sufficient knowledge of the process utilized by the client. Parties and clients, who are often lay persons, do not normally have the knowledge and expertise to understand their discovery obligations, to conduct appropriate searches, to collect responsive discovery, and then to fully produce it, especially when dealing with ESI, without counsel’s guiding hand.”

Of course, where there is a prima facie showing of spoliation, an attorney may be compelled to disclose how the litigation hold was implemented and how the search was conducted. The Hon. Paul W. Grimm, M. Berman, et al., “Discovery About Discovery Does the Attorney-Client Privilege Protect All Attorney-Client Communications Relating to the Preservation of Potentially Relevant Information?,” 37 U.Balt.L.Rev. 413 (2008).

Failure to Supervise May Run Afoul of Fed.R.Civ.P. 26(g)

The Yas Caribe court also looked to Fed.R.Civ.P.26(g), the so-called “stop and think” rule:

As to Attorney Mihailovich’s compliance with Rule 26(g)(1)’s reasonable inquiry requirement, his position warrants discussion. As mentioned above, Rule 26(g)(1) imposes on the signer of a discovery request, response, or objection a duty to perform a “reasonable inquiry.” Fed. R. Civ. P. 26(g)(1)….  In the case of a response to a document request, this duty falls squarely on the attorney, whereas interrogatory responses must be signed by both the attorney and the person who answers them. Fed. R Civ. P. 26(g)(1), 33(b)(5). [emphasis added].

The Yas Caribe court added: “An attorney’s failure to adequately guide and supervise a client’s search for responsive ESI may run afoul of Rule 26(g).”  Quoting a prior decision, it wrote: “[A]n attorney cannot abandon his professional and ethical duties imposed by the applicable rules and case law and permit an interested party or person to ‘self-collect’ discovery without any attorney advice, supervision, or knowledge of the process utilized.”

Application of Principles to Facts Presented

The court had little trouble rejecting the attorney’s argument on the facts presented:

In this case, such supervision was objectively warranted. Abdullah Yassin testified that he did not perform an adequate search for responsive documents, even when his attorney represented that he was “with a witness while s/he searches for the requested records, explaining to the witness what is requested, verifying that the information is pertinent to the request, and asking the witness to produce everything he has ….”

While, in making an inquiry, the court wrote that an attorney may rely on assertions of the client, that is permissible only “‘as long as that reliance is appropriate under the circumstances.’ 1983 Advisory Committee Notes to Rule 26(g).”  The court added:

A discovery request for ESI in a party’s cellphones and computer systems is a circumstance that triggers an attorney’s need to, at a minimum, understand how a client stores and uses ESI and how and where the client is searching for responsive documents.

As officers of the court, attorneys have a duty to make legally sound arguments and good-faith factual representations to the court. If an attorney does not have sufficient information to know whether his client is being accurate or not in a discovery response, particularly one involving ESI, this duty requires that he or she inquire deeper. “[T]he ethical rules now require attorneys to be competent with technologies such as ESI.” City of Rockford v. Mallinckrodt ARD Inc., 326 F.R.D. 489, 492 n.2 (N.D. Ill. 2018) (citing Model Rules of Pro. Conduct r. 1.1, comment 8 (A.B.A.)).[2] Attorney Mihailovich should have assumed a more active role in guiding defendants’ search beyond what he described in the opposition to the motion to compel. At the very least, he should have inquired into how and where the searches were being made so that he could give a reasonably informed certification of the thoroughness of what sources were and were not searched. Ideally, he would have discussed and run the keyword searches proposed by the EEOC or a mutually agreed upon version of them. Judging from the subsequent interactions of the parties that are on record, defendants’ position has been to engage in piecemeal corrections to their deficient responses and only when specific deficiencies are revealed by the EEOC’s efforts. This is not the way. [emphasis added].

The court provided “an admittedly extreme example: when attorneys receive documents that have been redacted by their clients and do not know what has been redacted, they cannot responsibly certify that the factual basis for a disclosure is compliant with Rule 26(g)(1).”

The court added that perfection is not the standard of review.  Instead, review is under an objective standard similar to Rule 11.

CONCLUSION

Applying principles of proportionality, the court considered privacy interests related to a forensic copy of the cell phones, and concluded that the burden would outweigh the benefits:

While there is certainly a well-founded suspicion that an adequate search was not carried out, and that defendants thus failed to comply with their discovery obligations, that does not necessarily mean that defendants are withholding responsive documents. After all, defendants represented that they had no messages between Abdullah and Khalid Yassin about Ms. Wilkerson or the alleged sexual harassment incident…, and the messages subsequently produced by defendants matched those in possession of the EEOC. In other words, there is no indication that, as of today, defendants have willfully withheld responsive documents or incurred in spoliation—only that they have not adequately searched for documents that may exist relating to the alleged sexual harassment incident at the center of the EEOC’s complaint.

As such, the request for a forensic inspection was taken “off the table for now” and the EEOC’s alternative request to compel Defendants to run supervised searches was granted.  For example, “defendants are ORDERED to perform an attorney-supervised search of both Abdullah Yassin and Khalid Yassin’s cellphones for responsive communications (email, text messages, call logs, etc.) across all relevant e-mail and messaging platforms/applications for [specified] periods….” As to one search, the response must include “the persons assisting in said searches….”

While the Yas Caribe court addressed several other issues, it stands for the principle that unsupervised self-collection is insufficient, at least where there is a factual predicate showing that the searches were insufficient.

For more information on “discovery on discovery,” please visit:

This blog was initially posted on  Electronic Discovery Reference Model and  JD Supra.

____

[1] For a discussion of the prior decision, please see Unsupervised Self-Collection Predictably Led to Problems (Jan. 2021).

[2] For an update on the ethical duty of “technological competence,” please see 40th State Adopts a Duty of Technological Competence – Is It a Good Idea? (Mar. 31, 2022); Proposed Changes to the MD Attorneys’ Rules of Professional Conduct Re: AI (Jul. 21, 2026).

]]> Thank You – Citation Re AI Protective Orders https://www.ediscoveryllc.com/thank-you-citation-re-ai-protective-orders/ Mon, 14 Sep 2026 19:25:57 +0000 https://www.ediscoveryllc.com/?p=5535 […]]]> Thank you to Cimplify for citing one of my blogs on protective orders.   Stepping Up to the Plate: Protective Orders to Restrict AI Training on Produced Data – Cimplifi

Cimplify discussed a number of cases and wrote:

As noted by Michael Berman in his blog for his site E-Discovery LLC, “Protective orders addressing the use of A.I. to review materials produced to an opponent in discovery are becoming routine.”

Cimplify concludes: “Given the risks associated with public AI platforms, protective orders that limit the ability of opposing parties to upload sensitive and confidential produced documents into a public AI tool are essential today. But the considerations associated with AI are changing rapidly. It’s important to keep that in mind when drafting your own protective order to protect your organization’s sensitive and confidential data.”

]]> A.I. Protective Order by Consent in Google Case https://www.ediscoveryllc.com/a-i-protective-order-by-consent-in-google-case/ Sun, 13 Sep 2026 09:19:26 +0000 https://www.ediscoveryllc.com/?p=5529 […]]]> The following AI Protective Order was stipulated in Skaf v. Google LLC, 2026 WL 2620563, at *10-11 (S.D.N.Y. Sep. 4, 2026):

Use of Generative Artificial Intelligence Tools and Large Language Models. Protected Material, and all information derived therefrom, shall not be inputted, uploaded, or submitted to any open, publicly accessible or freely available consumer-grade Large Language Model (“LLM”) or Artificial Intelligence (“AI”) tool (collectively, “AI Tools”). Protected Material may only be inputted, uploaded, or submitted to an AI Tool that is enterprise-grade and maintains confidential information using industry standard data security safeguards (“Confidential AI Tool”) after the Receiving Party first ensures that its contract with the AI Provider of the Confidential AI Tool:

(1) Prohibits the AI Provider from storing or using inputs to train, refine, or improve its model;

(2) Prohibits the AI Provider from disclosing inputs, metadata, and any information generated by the Confidential AI Tool to any third party, including integrated third party tools, except where such disclosure is essential to facilitating delivery of the service, in which case that third party shall be bound by obligations no less protective than those required by this Order; and

(3) Permits the Receiving Party to remove or delete the Protected Material, and any information generated by the Confidential AI Tool upon request.

A Receiving Party intending to use a Confidential AI Tool that it contends meets these requirements must retain written documentation of these contractual protections. In addition, the Protected Material, and any information generated by the Confidential AI Tool, including residual information stored in the Confidential AI Tool as a result of the submission of the Protected Material, must be deleted from the Confidential AI Tool at the close of the litigation, and in accordance with the Final Disposition section below. Any inputs, outputs, models, or other artifacts (e.g., embeddings, indexes, or logs) used in or created by a Confidential AI Tool that are derived from or contain Protected Material shall be subject to Section 7.1 and shall not be used for any purpose other than prosecuting or defending this litigation.

For the avoidance of doubt, submitting Protected Material, or excerpts therefrom, to any AI Tool that does not meet the requirements of this section is considered unauthorized disclosure to a third party under Section 10 of this Order. Nothing in this provision shall alter the obligations set forth in Section 1 (“Purposes and Limitations”) and Section 12.9 (“Data Security”) of this Order.

For more information on A.I. Protective Orders, please visit:

UPDATE: “Use of Large Language Model or Generative AI. A Party must not load, upload, input, import, submit, or otherwise transfer Confidential Information, Highly Confidential Information, or Restricted Highly Confidential Information to a publicly accessible Large Language Model (“LLM”) or generative artificial intelligence (“GenAI”) platform. Before using Confidential Information, Highly Confidential Information, or Restricted Highly Confidential Information in any non-publicly accessible LLM or GenAI platform, a Party must ensure that the LLM or GenAI platform cannot and will not utilize Confidential Information, Highly Confidential Information, or Restricted Highly Confidential Information to train public models or otherwise disclose Confidential Information, Highly Confidential Information, or Restricted Highly Confidential Information to other users of the LLM or GenAI platform not authorized to receive such materials.”  U.S.A. v. The New York and Presbyterian Hosp., 2026 WL 2547094, at *5 (S.D.N.Y. Aug. 28, 2026)(Amended Protective Order).

UPDATE: U.S.A. v. Jones, 2026 WL 2607719, at *2 (S.D.N.Y. Sep. 3, 2026).

UPDATE: Doe v. Google LLC, 2026 WL 2584519 (N.D. Cal. Sep. 1, 2026):

Any person in possession of Protected Material will maintain appropriate administrative, technical, and organizational safeguards (“Safeguards”) that protect the security and privacy of Protected Material, including restrictions regarding the disclosure or use of Protected Material in any large language models or any artificial intelligence services. The Safeguards will meet relevant industry standards and limit the collection, storage, disclosure, use of, or access to Protected Material solely to personnel and purposes authorized by this Order. As part of these Safeguards, each person will use a secure transfer method for all transfers or communication of Protected Material, and take reasonable measures to password protect and encrypt Protected Material. Each person will ensure that anyone acting on that person’s behalf is subject to the Safeguards or otherwise provides equivalent or greater protections for the security and privacy of Protected Material. [emphasis added].

“Such restrictions shall, at a minimum, prohibit the disclosure or use of Protected Material in any large language models or artificial intelligence services that are accessible to the public or will capture entered data in order to train what is accessible by the public.”

 

 

]]> No Violation of A.I. Protective Order https://www.ediscoveryllc.com/no-violation-of-a-i-protective-order/ Fri, 11 Sep 2026 20:30:43 +0000 https://www.ediscoveryllc.com/?p=5519 […]]]> Courts are routinely granting A.I. protective orders by consent.  U.S. v. Paula, 2026 WL 2665024, at *2 (S.D.N.Y.  Sep. 10, 2026).

In Elgin v. Wells Fargo Bank, N.A., 2026 WL 2664674, at *1 (N.D. Cal. Sep. 9, 2026), Wells Fargo asserted that Plaintiff violated such a protective order.  “Specifically, Wells Fargo contends that Plaintiff violated the stipulated protective order by uploading Wells Fargo’s discovery requests into an AI tool and then responding to those rewritten requests rather than the requests propounded.”  Id.  Two weeks later, Plaintiff served corrected responses.

The court wrote:

In opposition, Plaintiff clarifies that the stipulated protective order was not violated, because the material uploaded into the closed AI tool was not confidential and AI was not used to respond to the rewritten requests. Moreover, Plaintiff explains that Wells Fargo did not produce any documents marked “Confidential” until one week after the errant responses were served.…. Plaintiff’s points are well taken. Since Plaintiff provided corrected responses to all of the discovery devices within two weeks of the errant responses, and Plaintiff did not violate the stipulated protective order, the Court declines to find that Plaintiff waived all objections to the respective discovery devices or find that the requests for admission must be admitted.

Id. at *1 (emphasis added).

The court added: “Furthermore, the parties are advised that the Court has limited, judicial resources and does not appreciate unnecessary motion practice. Plaintiff served the corrected responses to all three discovery devices prior to the parties’ completion of the meet and confer process, which should have resolved this dispute informally. This is particularly true given the extreme relief sought by Wells Fargo. The Court trusts that, going forward, the parties will be more judicious in deciding which disputes require court intervention.”

For more information on AI Protective Orders, please visit:

 

]]> Use of an A.I. Recorder in Depositions?  In a “Meet and Confer”? https://www.ediscoveryllc.com/use-of-an-a-i-recorder-in-depositions-in-a-meet-and-confer/ Thu, 03 Sep 2026 10:37:51 +0000 https://www.ediscoveryllc.com/?p=5515 […]]]> In Greco v. Hon. Robert W. Lane, 2026 WL 2446747 (D. Nev. Aug. 19, 2026), the court held that there was no procedural defect in a request to use AI to record a deposition. Substantive questions were left open.

“Plaintiff served two Rule 30(b)(6) notices of deposition on Nye County. ECF No. 58 Exhibits A and B. Both notices (1) name Mark Gancarz as the person to testify on the County’s behalf, (2) request the production of documents, and (3) designate the recording method by stenographic means and/or video recording.”

Plaintiff moves to use the Mobvoi TicNote, an artificial-intelligence voice recorder, to maintain a secondary audio record of the depositions…. Rule 30(b)(3)(B) permits any party, with prior notice to the deponent and the other parties, to designate an additional method of recording the testimony. Plaintiff’s motion supplies the notice Rule 30(b)(3)(B) requires, so no procedural defect remains for this Court to resolve.

Because the depositions noticed in Exhibits A and B will not proceed on the current notices, the motion is moot. Should Plaintiff re-notice his 30(b)(6) deposition and designate this methos of recording, Defendants may file the appropriate motion to address any concerns they might have.

Fed.R.Civ.P. 30(b)(3) states:

(3) Method of Recording.

(A) Method Stated in the Notice. The party who notices the deposition must state in the notice the method for recording the testimony. Unless the court orders otherwise, testimony may be recorded by audio, audiovisual, or stenographic means. The noticing party bears the recording costs. Any party may arrange to transcribe a deposition.

(B) Additional Method. With prior notice to the deponent and other parties, any party may designate another method for recording the testimony in addition to that specified in the original notice. That party bears the expense of the additional record or transcript unless the court orders otherwise.

It is fairly easy to envision the same issue arising in, for example, a Rule 26(f) Conference of the Parties.

It will be interesting to see who bears the burden of persuasion to show that the specific AI tool is reliable.  Has it been tested?  Does it have a known error rate, for example? See Otter.ai “Transcript” Was Excluded From Consideration (Apr. 25, 2026); Felder: Part 1 of 4:  Court-Ordered Transcription of “Meet and Confer” Session (Aug. 13, 2024); Sometimes Discovery Disputes Do Not Bring Out the Best in Us – Part II (Jun. 20, 2024)(“courts have ordered that the parties record – by video or court reporter – their Local Rule 37.2 meet-and-confer sessions for the court’s review”).

]]> A Request for “Three-Strata Validation” Was Denied https://www.ediscoveryllc.com/a-request-for-three-strata-validation-was-denied/ Thu, 27 Aug 2026 15:32:36 +0000 https://www.ediscoveryllc.com/?p=5494 […]]]> Every search must be validated.  It is “entirely self-evident” that “ipse dixit pronouncements from lawyers unsupported by an affidavit or other showing that the search methodology was effective for its intended purpose are of little value to a trial judge.”  Victor Stanley, Inc. v. Creative Pipe, Inc., 250 F.R.D. 251, 261 (D. Md. 2008).

“Three-strata validation” was requested and rejected in In re Class Action Settlement Administration Litigation, 2026 WL 2444936 (D.D.C. Aug. 19, 2026). The parties in that case had narrowed their disputes and their “only remaining disagreement concerns how a party producing documents in discovery must validate that its productions are substantially complete.”

Plaintiffs asked for “three-strata validation.”  Defendants objected to the second and third strata.

The court explained the “three-strata” approach as:  “validation of documents coded as non-responsive, validation of documents coded as responsive, and validation of documents excluded from the universe of ESI material subject to discovery review.”

THE PARTIES AGREED TO NULL SET TESTING

The parties agreed that there is an obligation  “to confirm that responsive documents are not erroneously coded as non-responsive and omitted from their productions.” This involves review of a “statistically valid random sample” of documents coded as non-responsive. Id. at *1.

This standard practice is commonly called “elusion testing,” review of the “discard pile,” or, review of  the “null set.”  See Deal Genius, LLC v. O2COOL, LLC, 682 F. Supp. 3d 727, 734–35 (N.D. Ill. 2023) (“explaining that elusion testing, where the producing party reviews a random sample from the ‘null set’ of documents that did not hit on search terms and documents that were deemed non-responsive, is a ‘standard quality assurance practice’ that can ‘confirm the reasonableness and proportionality’ of the production”).

VALIDATION OF DOCUMENTS CODED AS RESPONSIVE – REJECTED

Defendants argued that validation of documents coded as responsive, to find false positives, was superfluous.  They asserted that Fed.R.Civ.P. 26(g) required that they certify that their response is complete and correct.  The court wrote: “Rephrased in statistical terms, producing parties must confirm that their false-negative rate falls within acceptable limits.”  In my less elegant words, a production may include nonresponsive documents; but, it cannot be a “dump truck.”

The court explained:

Review of documents marked responsive falls outside the bounds of litigants’ Rule 26(g) obligations because such review cannot identify whether documents are missing from a production, rendering it incomplete. Confirming that documents designated responsive are truly responsive reveals only the production’s false-positive rate, a proxy measure for whether the litigant has overproduced discovery material. But the Federal Rules do not prohibit parties from overproducing documents. Accordingly, this Court will not require defendants to incur additional costs re-reviewing documents that they have previously coded as responsive and intend to produce, barring claims of privilege. [emphasis in original].

VALIDATION OF DOCUMENTS EXCLUDED FROM THE UNIVERSE – REJECTED

The second strata consists of “ documents excluded from the universe of ESI material subject to discovery review.”  In other words, Plaintiffs asked for a review of documents that Defendants had not collected and that, presumably, were  not in Defendants’ litigation review platform.

The court wrote that examination of this set of documents, the “Excluded Set,” “could reveal false-negatives: responsive materials that should be produced but were erroneously omitted from the universe of materials collected for discovery review.”

However, the court responded that Plaintiffs’ request was duplicative and made too late, long after collection.  It wrote:  “Yet the Court is puzzled why plaintiffs request review of the Excluded Set after discovery review is largely complete, given that the parties also intend to review the Excluded Set for responsive material prior to beginning that review.” [emphasis in original].

The court explained that “ESI discovery generally proceeds in two phases: document collection and merits review.”  The parties had earlier “set forth agreed upon parameters for collecting the universe of documents that will be subject to merits review.”  That earlier agreement had required statistical sampling.

Insofar as the parties have agreed to statistically validate the completeness of the set of documents collected for merits review before even beginning that review, the Court finds that further validation of the Excluded Set after merits review is largely complete would be duplicative and unnecessary.

However, Plaintiffs made a different argument—so-called “substantial completeness evaluation.”  They argued that:

[T]heir proposed validation efforts “answer different questions at different points in time,” because search term validation “tests whether agreed terms are performing as intended before review begins,” while substantial completeness validation of the Excluded Set “tests whether the entire process — including all downstream coding decisions, review workflow, and production choices — resulted in a complete production at its conclusion.”

This argument was unsuccessful because the court held that “the Excluded Set is a collection of documents excluded before merits review begins and not subject to downstream coding decisions, review workflows, and production choices.”

Plaintiffs also contended that “the prevailing practice is for parties producing discovery to validate the Excluded Set both upon its creation and after merits review….”  The court noted that the authorities are split; however, it held that “the Court will not require producing parties to incur the cost and burden to re-validate the Excluded Set at the substantial completeness stage, provided that they statistically validate that documents excluded from merits review do not contain responsive material before beginning that review.”

I would add that “substantial completeness validation” of the Excluded Set could de-rail a case and mandate a “do over.”  If a sufficient number of false negatives was found in that set—and sufficiency is always debatable–presumably Defendants would have to engage in a new collection, review, and production process.

This blog was initially posted on  Electronic Discovery Reference Model

]]> A.I. Protective Orders Are Becoming Routine https://www.ediscoveryllc.com/a-i-protective-orders-are-becoming-routine-2/ Thu, 27 Aug 2026 09:22:39 +0000 https://www.ediscoveryllc.com/?p=5479 […]]]> I have been posting about A.I. protective orders since last year. Basically, these orders limit use of artificial intelligence by the recipient of discovery materials.  After I posted Another A.I. Protective Order – Part 10 (Aug. 21, 2026), at least three more were reported.

In Campbell v. TidalHealth, Inc., 2026 WL 2162749 (D. Md. Jul. 28, 2026)(Austin, J.), the Court wrote: “Campbell denies submitting documents into any generative artificial intelligence tools but admits to using unspecified ‘available tools solely to express [her] arguments in proper legal format, the same way any litigant uses available legal resources.’” Id. at *8.

After cautioning Campell about the use of GenAI, the Court wrote:

In addition, Campbell must ensure that any use of generative artificial intelligence does not violate the stipulated order governing confidential materials…. TidalHealth expresses concern that Campbell violated the February 27, 2026 Protective Order’s prohibition on submitting discovery material to public generative artificial intelligence tools. … see ECF 21 at ¶ 8 (imposing limitations on use of generative artificial intelligence). Campbell denies doing so…. If future filings suggest improper submission of discovery materials to generative artificial intelligence tools, the Court will be more inclined to issue a show cause order and consider sanctions.

Id. at *10 (emphasis added).

These orders are standard of care and are most often entered by consent. But not all orders are created equal and they are not always fair.  I have echoed Craig Ball’s novel post on the need for equity.  Another A.I. Protective Order – Part 10 quoting Craig Ball, The AI Protective Order Double Standard | Ball in your Court (Jul. 27, 2026).

UPDATE:

“Because the parties’ proposed protective order adequately complies with the standards set forth above, the Court will enter an order based on the parties’ submission. The Court makes four minor modifications to the parties’ protective order: … and (4) to establish parameters for the use of Artificial Intelligence as it pertains to confidential documents, see infra Paragraph (F).”  Zask v.Kohl’s Corp., 2026 WL 2531856 (E.D. Wisc. Aug. 27, 2026)(emphasis added). The order states:

F. USE OF ARTIFICIAL INTELLIGENCE (“AI”)
A receiving party may use AI tools in connection with discovery materials produced in this action only if the AI tool is operated in a secure environment and is subject to enforceable contractual terms that (a) prohibit the provider from retaining, using, disclosing, selling, or using the materials to train, improve, or develop any AI system; (b) prohibit the materials from being made publicly available or accessible to any other customer or third party; and (c) require commercially reasonable administrative, technical, and physical safeguards to protect the confidentiality of the materials. Discovery materials, whether designated CONFIDENTIAL or ATTORNEYS’ EYES ONLY, shall not be entered into any publicly available or consumer-based AI platform. Any use of AI tools must remain consistent with this Order and with the requirement that discovery materials be used solely for purposes of this litigation.

]]>