[Webcast Transcript] eDiscovery Lessons for 2026: Spotlighting the Top ESI Cases and Trends from 2025

Editor’s Note: The intersection of AI and electronic discovery has moved from theoretical concern to courtroom reality, and the legal profession’s response will define litigation practice for the next decade. HaystackID’s recent webcast “eDiscovery Lessons for 2026: Spotlighting the Top ESI Cases and Trends from 2025” examined how courts are adapting century-old discovery principles to AI-generated content, auto-deleting collaboration platforms, and datasets measured in the billions. The cases analyzed by Phil Favro, Judge Allison Goddard, Ruth Hauswirth, and Eric Stansell revealed a judiciary struggling to balance traditional proportionality standards with unprecedented data volumes, as evidenced by the OpenAI litigation, where 20 million ChatGPT logs represented merely 0.05% of available records yet still warranted production. What emerged from 2025’s landmark decisions is both reassuring and cautionary: while fundamental discovery principles remain applicable, organizations can no longer rely on default retention settings without strategic consideration of litigation readiness. For legal professionals navigating this landscape, ignorance is not just costly; it’s sanctionable. Read the full transcript to learn more.

Expert Panelists

+ Philip Favro (Moderator)
Founder of Favro Law PLLC

+ Judge Allison H. Goddard
Magistrate Judge, U.S. District Court, Southern District of California

+ Ruth C. Hauswirth
Special Counsel and Head of Litigation & eDiscovery, Cooley LLP

+ Eric Stansell
Senior Counsel, Discovery, Tyson Foods

[Webcast] eDiscovery Lessons for 2026: Spotlighting the Top ESI Cases and Trends from 2025

By HaystackID Staff

• Courts are sending clear signals that automatic deletion policies can be acceptable if they predate any duty to preserve, but organizations must be prepared to override these defaults once litigation is reasonably anticipated.
• The California Privacy Protection Agency is aggressively enforcing the CCPA, taking action against companies like Tractor Supply for continuing to share consumer data after receiving opt-out requests, demonstrating that privacy compliance is no longer optional.
• Traditional proportionality analysis still applies even when dealing with massive AI datasets, with one court ordering the production of 20 million ChatGPT logs as a reasonable sample, representing less than 0.05% of the total available records.

Transcript

Mary Mack

Thank you for joining today’s HaystackID webcast, eDiscovery Lessons for 2026, Spotlighting the Top ESI Cases and Trends from 2025. Today’s expert panel is led and moderated by Phil Favro, founder of Favro Law PLLC, and includes Judge Allison Goddard, Ruth Hauswirth, and Eric Stansell. We’re recording today’s webcast for future on-demand access, and slides will be available during and after the webinar. The webcast will be available on the EDRM Global Webinar Channel for the next quarter to help support your ongoing learning and reference needs. And before turning it over to Phil for the agenda, Holley Robinson of EDRM will share a few brief notes on the webinar console. Over to you, Holley.

Holley Robinson

Thanks, Mary. If you look at the top of your screen, you’ll see the HaystackID logo, which you can click on to learn more about HaystackID. You’ll also see an option to contact Team HaystackID directly, as well as speaker bios that you can pop open and learn more about today’s presenters. Moving down, you’ll see the Q&A box where you can type in your questions for today’s faculty, and we highly encourage you to do so. Below that, you’ll find today’s resources, including the slide deck, links to register for HaystackID’s upcoming webcasts on February 25th and March 25th, a link to Newsline by HaystackID, and links to keeping current with eDiscovery search trends, lessons from the OpenAI litigation on safeguarding privileged information, and information governance for enhanced legal hold process and improvement on HaystackID’s blog. Lastly, you’ll find some emojis down at the bottom of your screen. Please feel free to use them and react throughout the webcast. Back to you, Mary.

Mary Mack

Thanks, Holley. And our moderator, Phil Favro, is the founder and president of Favro Law PLLC, where he counsels clients on ESI, AI, and discovery issues, and he serves as a special master, mediator, and expert witness. As a special master, Phil is acclaimed for his collaborative approach, working with parties to find stipulated solutions to complex issues. For disputes that require adjudication, he’s renowned for the clarity and vigor of his written dispositions. And one of the things you might want to take a look at is Oakley versus MSG Networks Inc., featuring Phil. Judge Allison H. Goddard has served as a magistrate judge for the US District Court, Southern District of California since 2019. She presents regularly on eDiscovery and technology and the law for the Federal Judicial Center, also with The Sedona Conference, and at legal conferences throughout the country. She recently went overseas, spoke in Mongolia and Uzbekistan, and closer to home, Judge Goddard also coaches a local high school mock trial team. Ruth Hauswirth is special counsel and head of litigation and discovery at Cooley LLP, where she founded and leads the firm’s innovative eDiscovery and information governance practice. Ruth is a certified AI Governance Professional through the IAPP, and she’s also a Relativity AI Visionary, and she’s taught or led at San Diego School of Law and the Santa Clara University School of Law. Ruth also leads workshops for lawyers and legal professionals to integrate mindfulness into our practices for greater fulfillment and well-being. And from the in-house perspective, we’ve have Eric Stansell serving as Tyson Foods Senior Council for eDiscovery or for Discovery, excuse me, without the E. Eric is responsible for leading the development and implementation of eDiscovery best practices, management of eDiscovery vendor partner relationships, and evaluation of new eDiscovery oriented partners, technologies, processes, and policies. Prior to his current role, Eric led teams and initiatives focused on technology-assisted review for one of the country’s largest alternative legal services providers. And before his eDiscovery career, Eric worked as an insurance regulator and in the political sphere. And without further ado, Phil, please take it from here.

Phil Favro

Thanks, Mary. Thanks for your kind introductions. I want to thank HaystackID for sponsoring today’s program, and we’re all grateful to be here and look forward to the discussion of the issues. Before we get started, let me just go through the mandatory disclaimer. I know everybody knows this disclaimer, but none of us are here to create a lawyer-client relationship. The judge is not issuing any advisory opinions. Nothing we say is legal advice. This is all educational. If you think otherwise, then maybe you should drop from the program. Otherwise, carry on with us because we’re going to cover some very interesting decisions today. And let us kick off with some cases involving pre-litigation conduct that impacts litigation and discovery. And these are some of the more noteworthy cases from 2025. Now, there are so many cases that we could not be expected to cover them even in a week. There are that many.

So let’s kick off by highlighting both the Malhotra case and the Johnson case. So these decisions dealt with platforms, workplace collaboration platforms, where calls could be or were recorded, video calls. And in each of these cases, Malhotra, which dealt with Microsoft Teams, and Johnson, which dealt with Zoom, you had a retention policy. In the Johnson case, it was a 30-day active retention and then soft delete of any recordings that have been kept 30 more days and less, you could say active storage, but still around. So, effectively, 60 days until any recordings were hard deleted. In the Malhotra case, you had essentially only the Teams messages that were retained. If you wanted to have a video retained, a particular user had to enable the retention of that. In each case, pertinent video recordings and recordings of calls were not retained. And in each case, the plaintiffs sought to obtain sanctions against the defendants for their failure to retain the information.

Now, the court said, “Look, we’re not going to impose any sanctions because these recordings went away before the duty to preserve attached.” And it’s a very straightforward decision in each time. The recordings were eliminated pursuant to a retention policy, and then the duty to preserve attached. This is a lesson that we’ve known about for a long time in terms of information governance, Ruth. But I think the more interesting issue to talk about here deals with AI extrapolated information. In other words, what about the recordings of these… Excuse me. What about the transcripts from these calls? What about the summaries from these calls? What are companies doing with these things? While Johnson and Malhotra don’t speak to those particular issues, I think the cases are instructive on this particular issue. Ruth, give us your reactions on how you think or whether you think they’re instructive, and on the cases themselves.

Ruth Hauswirth

Sure. Yeah. So I think the cases are providing insight into something that we’ve been talking about for quite a while, which is that information does not need to be kept indefinitely and that there’s probably more risk to keeping too much than there is to having procedures and policies to manage what information is kept. And then to your question, specifically your point about AI related content, I do think that we’re getting into a time when these AI generated transcripts of meetings are particularly problematic because unless somebody is going and verifying the accuracy of these transcripts, they’re sitting in an organization’s, depending on if there’s no retention or disposition on these recordings, they’re just sitting in the company’s archive, storage.

And if an investigation or regulatory investigation, or a litigation kicks off, those transcripts are written records, but we don’t know if they’re actually accurate. I mean, the AIs are doing… They’re getting better, but I think we’ve all had experiences where we read something, and it’s, “Well, that’s not accurate.” So I think that’s particularly problematic because I don’t think people are going to take time to do that, to verify the transcripts or the summaries. The summaries are also very concerning, and there’s a lot of exposure around. Did everyone on the call consent to the recording, the type of information that’s in the recording? There could be privileged information, there could be private information. There are a lot of concerns with that information just hanging around without any management.

Phil Favro

Yeah, I agree with you, Ruth. And I’m curious to hear what Eric has to say from an in-house perspective because Eric, we hear about companies all the time wanting to generate all this type of data, not just, “Well, we want to have some sort of note-taking and note-keeping and some sort of record-keeping associated with these calls.” AI provides that. This is great, but what’s the value of this information over the long term? Even if you ingest it into your tools to analyze what happened, does the value stay high, or does it drop, and how soon? Eric, what’s your reaction to these particular issues?

Eric Stansell

Yeah, I think you’re hitting in the right direction here. It’s definitely a double-edged sword. From an in-house perspective, the business units definitely love the ability to do this. It is productive in many ways, right? When you have a note taker that can do that, and you do have a transcript, it is helpful, but I think Ruth hit on my biggest concern, which is that the transcript is bad enough; keeping that indefinitely is terrible. But my real concern, and I think what we all should be concerned about, is that derivative of that, the summary. The summary is what terrifies me, because in some cases, you’re going to have terms of art or whatever that are summarized in a way that doesn’t make a lot of sense. And if you read that summary, it can come across wrong, especially if it’s not noted that that particular summary was AI-generated.

So I think that’s a big problem. And look, I think one of the fixes here is I think what we’ve known about for a long time and what we continue to talk about is keep information that you need as long as you need to keep it, but no longer than that. Instituting, for instance, an email governance policy, which a lot of us now are just starting to go down. Some people don’t have it, some have had it for a long time, but the value of an email and the value of a transcript, whatever, has 100 points on day one. It may have 90 points on day two and three points on day 30. I mean, it follows in a logarithmic fashion.

Phil Favro

Yes.

Eric Stansell

So yeah, I think it’s really important to be very aggressive on those policies. And again, what Ruth said earlier is exactly right. I think there’s over-preserving, whether particularly by choice or just because you don’t have a policy, which is really dangerous, and there’s frankly more value in getting rid of data than keeping it too long.

Phil Favro

Judge, what’s your reaction here? You’re looking at this from a little bit different perspective, although no less important, especially given your views on AI and how you are from the outside perspective as a power user. Tell us about your perspective on these cases and the issues we’re discussing.

Judge Allison Goddard

Well, I think we can all agree that this is nightmare fuel for anyone who’s working in-house. So this is why I describe information governance as an up-all-night job. How do you sleep when you think about all the different ways that these things can happen? Because beyond discovery, you have to worry about breaches and people being able to reach in and get this information. And in these cases, though, I do think that the organizations did have clear policies that helped them in a spoliation issue. So because they had policies in place and they were following their policies, they didn’t get sanctioned. But beyond just discovery sanctions, you have to ask, if you have a policy where you can selectively decide, an employee can selectively decide whether to retain the video from a meeting, why didn’t they retain the video from this meeting? This is where you discussed what was going on with the contract, and the allegations are that you had some sort of tortious breach or that you fraudulently induced the other side into the contract, but you saved certain meetings, but you didn’t save all.

So that can be problematic. Even if it’s not going to get you a discovery sanction, it’s going to be problematic at trial, and why did you save this but not that? But overall, it’s a no-brainer to have a policy in place that allows for automatic deletion of certain information. I think all corporations should be looking at doing that, and then you have to make sure it’s actually applied across the organization, just like the policy is written, because the policy’s no good if it’s not being actually applied. So I think in both of these cases, I think they’re good examples of how it may cause problems to you down the road at trial, but at least you’re not going to get sanctioned for deleting evidence if you have a clear policy that’s followed.

Phil Favro

Thanks, Judge. And thanks, Eric and Ruth. Let’s transition to our next case. It’s not really a case, necessarily. It’s a regulatory decision and a settlement. And this pertains to the California Consumer Protection Act. And this is another pre-litigation issue. And this is a fairly noteworthy settlement that was reached by the California Privacy Protection Agency, the CPPA, and Tractor Supply. Tractor Supply is a fairly notable retailer throughout the country. And apparently, according to the settlement that was reached between the CPPA and Tractor Supply, Tractor Supply had an opt-out form that allowed consumers to opt out of the sale of their personal information, but apparently it gave, according to the agency, a false sense that they had actually accomplished this when in fact the Tractor Supply form did not provide and do that. And so there was an enforcement action filed by the CPPA, and you ended up having this particular settlement.

And Eric, I mean, we’re not talking about a lot of money here. I think it was a $1.35 million fine. I can’t remember. Well, maybe it’s, what is it? $5 billion or something like that would be Tractor Supply’s global revenue. So 1.35 million. I don’t want to say it’s laughable, but it’s approaching that. The issue really would be the continuing oversight.

Eric Stansell

That’s correct.

Phil Favro

That’s where I see this being an issue, where you don’t want to get in the CPPA’s regulatory sites. What’s your reaction, Eric?

Eric Stansell

No, I think you hit the nail on the head. The number itself was not draconian. It was, I mean, eminently reasonable, especially in a lot of their revenues, but I think there’s oversight until I want to say something like 2030, something along those lines, where they have to… A mandatory annual review of their website apps and how they deal with third parties on that, as well as a written statement of compliance. That’s where the breadth of it is, that really is. And there’s a lot of… It shows what you’re going to have to do is just to move on and to look and see, is your client going to be subject to the CRPA or the CPRA, I should say. And if you do, do you have the systems in place to do that?

And then, as Judge Goddard was saying a second ago, if you set this up, then you have to make sure it’s working, and there are no gaps. That’s the biggest thing. In this case, they had a web form that people could fill out, but it didn’t do anything. They were still being tracked. And so yeah, the CCPA is being enforced, and I think fairly, I wouldn’t say broadly, but for a length of time. And that’s what people need to be watching over because that’s where the cost of that’s going to be more than 1.35 million.

Phil Favro

Yes, agreed. And Eric, I think we’re seeing not just enforcement actions by the CPPA, but also by the California Attorney General’s office. We’re seeing private enforcement actions. And Judge Goddard, are you seeing any of these come before you, the private enforcement actions?

Judge Allison Goddard

I haven’t seen any of these cases… I haven’t had one in front of me, but I have to tell you, first of all, it seems to be basic. If you’re telling customers you’re not going to sell their information and then you do it through a third party, I won’t comment on whether this regulatory scheme is the most effective, but it’s not a great practice for a company in the long term anyway. And there’s a lot. I mean, maybe it was a little accidental, maybe they didn’t have the systems in place, but you need to be careful about what you’re promising to the people who are making your business run. And it’s an interesting regulatory scheme that you would have a penalty that is, in effect, nothing more than the cost of doing business, but then have on the back end this threat of ongoing interference with the way the company does their business. It’ll be interesting to see how that plays out.

Phil Favro

Agreed. Agreed. So we’ll be on the sidelines and hopefully not in this particular match. Nobody wants to be in that game, be regulated, and deal with these. So we’ll be watching hopefully from the sidelines to see how these things play out. Let’s skip to our next case dealing with lawyer-client privilege. And this particular case, the FirstEnergy case, I feel like is super important. Every 10 years or so, we get a new court of appeal decision reaffirming that the attorney-client privilege, and in this case also the work product doctrine, they mean what they say. And in this particular case, apparently FirstEnergy had… There were certain allegations leveled against it that would’ve placed it in criminal and civil liability. There would’ve been civil liability, and there could have been criminal liability as well. And so, apparently, FirstEnergy engaged counsel for advice on how to handle these particular issues.

And the district court had said, “Look, this was business advice that you got. This was nothing more than business advice. Maybe there were some legal considerations mixed in, but principally business advice.” The Sixth Circuit Court of Appeals last fall disagreed and said, “Look, when we’re talking about criminal and civil liability here and of a significant nature, the engagement of counsel is done principally for legal purposes and for rendering legal advice. The business advice may be given as a collateral issue to the principal issue of criminal and civil jeopardy.” And so the court found that the privilege applied and also that work product applied to certain documents that were generated.

And judge, I don’t know how many privilege disputes are coming before you, but they do come in the context of discovery and in other areas. What’s your reaction to FirstEnergy? Because I see this as the most noteworthy decision on this issue since the Kellogg, Brown, and Root cases that came out of the DC Circuit 10, 11, 12 years ago.

Judge Allison Goddard

I do think it’s interesting. I’m always, when I read an opinion like this, I’m always looking for the story behind the opinion because I’m looking for the tea, for example. And I think that the Circuit Court seemed really displeased overall with the district judge’s order on a number of levels. There’s an interesting side step where the district judge refused to accept a declaration that was submitted by FirstEnergy because it said it was under penalty of perjury. It didn’t say it was true under a penalty of perjury. It didn’t have that word true. So I think overall, the Circuit Court wanted to express its displeasure in a subtle way with that order overall. I think this is a really tough issue. I don’t know that it comes out the same way in different circuits. You really have to look at what the… What we don’t have by just reading the opinion is what the investigation report looked like and the information that came out of it.

But I think that at first blush, it doesn’t seem like it’s… I wouldn’t say it’s wrong. I think it just depends on what’s actually in the document. So I think that even though it is… Certainly, if you’re a corporate defendant, you really like this opinion, and you’re going to cite it when you can to protect the information. But I don’t think it totally solves the issue that companies face when they are doing an investigation, and they want to be sure that what they’re doing is protected by the attorney-client privilege and work product doctrine. So I think you still have to be careful, but it does give you some insights into how you can structure your investigation reports and how you conduct the investigation so that you can protect their attorney-client privilege and the work product doctrine to the fullest extent possible.

Phil Favro

Thank you. Thank you, Judge. Maybe Eric and Ruth, if you can just give us some brief insights. I mean, Eric, imagine you engage Ruth and her firm just hypothetically to assist you with allegations of internal corruption. What do you learn from this? And you’re working for hypothetical company XYZ. What do you learn from this in terms of trying to protect privilege and safeguard the company’s interests? Maybe Eric, if you could speak to that. And Ruth, from the outside perspective, if you could briefly speak to that as well.

Eric Stansell

Yeah. Well, I’ll say two things. One, a different result here would be very unsettling to me, period, but that’s neither here nor there. But I would say the great thing is when we do have any questions on anything, we would go to Ruth, right? We would make sure there are no questions whatsoever. But this decision going a different way would be troubling because almost all of our attorneys do have a business purpose. It’s going to be primarily for legal advice, but they’re not giving it for just no reason at all. And so, especially on that advisory piece, that’s going to exist. But yeah, I think in the context of investigations, even these kinds of questions coming up are going to push us to utilize Ruth and outside counsel to just make sure that we have the belt and suspenders taken care of.

Phil Favro

Thanks, Eric. Ruth.

Ruth Hauswirth

Yeah. And just real quick on the… I think it is, we’re seeing this more and more in terms of making sure that… I think the judge, there was in the opinion, an important clarifying point about, and Eric hit on this, there are always going to be business decisions and implications of the legal advice. And so parsing what business advice is and what legal advice is is challenging. And there’s a lot of case law that addresses that issue. So I think it underscores the need to raise awareness about how these communications are happening, how they’re being documented and therefore protected because the point is to make sure that there’s free-flowing information so that the issue can be addressed, and you don’t want to have… I feel I agree with Eric that if it had gone a different way, that would be problematic because you can’t always separate the two things, whether it’s business or legal advice. And I think the judge balanced it well. And then, to Judge Goddard’s point, how are you documenting this advice so that it meets the elements of the privilege?

Phil Favro

Thanks, Ruth. Thanks, everyone, for your comments. Let’s skip to the next slide, and we’re going to talk about the OpenAI litigation that is before Judge Ona Wang. And I believe it’s Judge Sidney Stein, or is it… I can’t … I think it’s Judge Sidney Stein in federal court in Manhattan, the Southern District of New York. We actually have a number of decisions from this particular case, copyright or allegations of copyright infringement by OpenAI with its use of ChatGPT. The Times, New York Times is just one of many litigants in this case, but in this particular decision issued, I think it’s on October 1st. So just a few months ago, OpenAI had taken the position that they had set up a de facto Slack channel for its internal stakeholders to get legal advice from their in-house counsel. And the court had said, and they had said, “Look, all these communications should be considered privileged because they’re in our Slack legal counsel channel, and the judge wasn’t going to just sign off on that.”

Judge Wang had said, “Well, that’s interesting, but I’m actually going to look at these one by one.” And the judge said, “Some of these are privileged, a lot of them aren’t.” And so it’s an interesting issue, but I think the one that’s more fascinating in this case is the Excel spreadsheet reflecting prompts and outputs that, according to OpenAI, its counsel directed internal stakeholders to prepare. And so OpenAI had taken the position that the spreadsheet itself was privileged because it contained these prompts and outputs prepared by the client and internal stakeholders, at the direction of counsel. And Judge Wang disagreed, saying, “Look, as a principle matter, there isn’t even a legend on the document indicating it’s privileged. And then second of all, there’s nothing about these prompts and outputs that suggests that it would be privileged.”

And so I think these are interesting issues because we see judges, a lot of the time, Ruth and Eric, saying, “You can’t just put a legend of privilege on the document and expect it’s going to be protected.” And you see judges criticizing the overuse of legends, do they not? And yet in this case, we have a judge that says, “Look, there’s no legend on this document. There’s nothing about it that suggests that it’s privileged, and there’s no explanation.” So, Ruth, why don’t you start us out on this? And then Eric, if you can also give your thoughts on this, too, please.

Ruth Hauswirth

So I think this is a really interesting order to read. Judge Wang parsed very closely all of the particular issues that were brought up in this privileged dispute. And I think there’s a lot to be gleaned from looking at the particulars. And I think that it also underscores how difficult a privilege review of ESI in particular is because it isn’t just in the old… It’s not easy to just slap a legend on some of the types of data that we use now, but it really highlights for me some of the broad misunderstandings about how the elements of attorney-client privilege are applied and how it really works in practice. And I think, and Eric will probably have a lot to say about this in the in-house context, but there’s a lot… I feel like we see this often, where there’s a lot of misunderstanding about how the privilege actually works.

And so it’s a really difficult and dicey area. We’re trying to do the best we can when we’re doing a privilege review. There are all kinds of constraints, but what the judge showed here is that there are going to be a lot of documents and types of information that are going to have multiple purposes, as we talked about in the last session. And so you have to look at the predominant purpose of the document when you’re making privileged designations and determining whether an entire document is going to be privileged or maybe part of it. I think it was interesting that the judge said that the spreadsheet didn’t have any indication that there was any actual legal advice being exchanged. There was no legend, and there was just no indication of what the legal advice was.

And so I think it’s like going back to law school and we’re keeping those elements of the attorney-client privilege, a communication between privileged parties for the purpose of seeking legal advice and ensuring that we’re looking at those communications at the time of creation and making sure that we’re appropriately protecting those communications as best as possible, but not just de facto.

Phil Favro

Let me jump in really quickly. We got a question from the audience for you. Does this mean that courts are going to be holding clients and lawyers to a higher standard now, with AI-driven discovery tools becoming more common? I don’t know that I see it as a higher standard. I see it as just holding them up to the normal standard, but in the context of technology, what’s your reaction, briefly, to that? And then Eric, we’ll segue to you.

Ruth Hauswirth

No, I agree with you, Phil. I don’t think it’s a higher standard at all. I think it’s a… When you read these opinions and orders, and Judge Goddard probably would have something to say, it basically looks at the elements and makes sure you’re applying the elements as they exist. And it’s not easy in practice, but no, I don’t believe that there’s any higher standard for AI-generated content. We may have to approach it a little differently in terms of how we’re… It’s not just AI-generated content. It’s all of the Slack channel, all the things that were talked about in order to present unique issues.

Eric Stansell

Yeah. I mean, look, I can’t add much to what you said, Ruth, because I think this is much simpler than it looks on its surface. It really gets back to the elements, in my opinion. I agree wholeheartedly. I think that in-house, you do have to, for those that haven’t been around the block a few times and have been told this a number of times, you do have to let folks know, “Hey, just because you put a label on something does not make it attorney-client. Just because you do X, Y, or Z doesn’t make it… Or work product doctrine, too.” I mean, we forget sometimes that folks get the idea that they’ve heard work product, and then they think, oh, well, this must be work product then if I can’t do attorney-client. It has elements, and we have to follow those. And so I can’t add much to what you said.

I can tell you that, and I won’t say this for where I work now, because I work… But I have seen before where people do think that if I have this particular, in this particular case, a Slack channel that was opened by the general counsel, or if I slap an attorney’s name on it, everything’s great. It’s not that way. And I don’t think that AI content is going to be any different. I mean, why should it be? It gets back to the elements.

Phil Favro

Thanks, Eric. Judge, you get 30 seconds on this one, if you don’t mind me putting you on a time clock. You’re a federal judge. You put people on time clocks all the time.

Judge Allison Goddard

That’s because I’m the judge, Phil. I don’t have much to add. I think it’s really interesting to hear the in-house perspective about it. I think these are really difficult issues. And I’ll tell you, it’s really miserable for judges to have to review this stuff in camera and try to guess at whether or not it’s privileged or not based on the party’s arguments. I just think it’s a really tough… It can be a tough call. And one thing to keep in mind is the nature of the document, and it’s really hard to put a legend on a spreadsheet. I mean, I guess you could put it up at the header, but that might affect how you use the data and how it gets sorted as far as… You could sort it without the header, but how people are more likely to get it sorted. So it’s just not something we necessarily see all the time. And sometimes you need to be thinking about the nature of the information and whether it lends itself to that type of label.

Phil Favro

Thanks, Judge. And Eric and Ruth, thank you as well for your insights here. So let’s look at decisions addressing preservation and production rules applying to AI. And these decisions, again, arise from the In Re: OpenAI Copyright Litigation. So we’re going to cover these two holistically. The first one is a decision from mid-September. So last fall, and you had, in this particular instance, OpenAI is seeking to compel information from the New York Times. The New York Times has licensed OpenAI’s one of its chat tools, one of its AI tools, to be able to conduct journalistic research. And OpenAI said, “We want to see the prompts and outputs in those from that particular instance that you’ve licensed from us.” And the New York Times calls these ChatExplorer logs.

The Times said, “We’ve given you some information. We think the information you’re asking for isn’t relevant, nor is it proportional to the needs of the case, and nor do we think it’s appropriate that you get a four-hour 30(b)(6) deposition to explore preservation issues.” And Judge Wang agreed with the Times and rejected the motion. And the discussion went along these lines: “What the Times does with its copyrighted content is not pertinent to your fair use defense of the Times’ copyrighted content.”

So that’s the first issue. The second issue, the court had said, “There are no grounds that you put forward to show that it’s proportional, OpenAI. And in contrast, the Times to put forward specific metrics showing why it’s disproportionate to the needs of the case.” So contrast that decision with a series of decisions requiring OpenAI to produce a sample of 20 million logs, essentially prompts and outputs that consumers have used of its product around the world, with certain privacy protections that have been put into place. So many litigants and commentators have looked at that order and said, “Well, look, that’s exceptionally broad. I mean, where do our preservation and production obligations go?” But then you contrast it with the earlier order that we talked about, and we say, “Well, maybe it’s just like ordinary discovery.” Judge, why don’t you start us off here? What’s your reaction to these particular decisions?

Judge Allison Goddard

Well, I have thoughts and feelings, and I will tell you that I know Judge Wang really well from doing a lot of stuff with her on the FJC, and I don’t envy her at all in this case. We would all, as judges, love to have a big case that gets a lot of attention, but we wouldn’t want one that goes on for years and years and years and has discovery disputes about every other day. And so she knows these lawyers really well. And I think she nailed it on the copyright defense and the relevance of the information. And so one of the things, and I think, I don’t know even who the lawyers are in this case, but I will tell you, the better the lawyers, the more sophisticated. They have this tactic sometimes of they go into and like, “Judge, I’ve never seen any judge ever say that something like this is not relevant in a copyright case.” And make you feel like you would be such an outlier. And I don’t know if that happened here, but I would assume it may have.

And Judge Wang cut through all that, and she went right to first principles, and I just was really proud of her for that relevance and proportionality. That’s what we need to look at. And there’s a little bit of, “But we’re having to produce so much on this side, and they’re not having to produce anything over there.” Well, that’s what the law says is relevant, and that’s the rule that we’re going to follow. And when we talk about the 20 million de-identified logs, if you look at it, I’m not sure if it’s in this order or in one of the other orders that we highlighted for this webinar. OpenAI offered that. They offered that as a sample size. So it’s not like the New York Times just said, “We want 20 million user logs.” There are billions of user prompt logs. And OpenAI offered that up as a reasonable sample.

So I think that’s getting lost in some of the reports about this case. And so I don’t think… Discovery is not always tit for tat, especially in asymmetric cases. It may be frustrating to a defendant that the plaintiff does not have to do nearly as much work as they do, but that’s just the way the law works as far as how it defines relevance. So those are my thoughts and feelings.

Phil Favro

Thanks, Judge. Eric and Ruth, what are your reactions on these particular cases?

Eric Stansell

Yeah. I mean, real quickly, I think one of the… And I think, Judge, what you said was exactly right. I think the relevancy issue is there. And I can tell you from the in-house perspective, we often deal with asymmetrical matters. That’s part of being a corporation. That’s just it. But in this particular instance, I found it to be very interesting that, yeah, there are billions, and these numbers get really lost because they seem so big, but the tools that we can use in eDiscovery are equal to the task in many cases. And I think that’s where the judge in this case saw that, is that 20 million is a number for most folks that seems big. But when you look at the tools that we have and the way that these things can be looked at, it’s not terrible.

And I think also, for good, bad, or indifferent, for this particular motion, one of the problems that OpenAI had is that they had essentially completed their review of everything. They were basically ready to go. They were just complaining about the number. And so therefore, look, you had this, you’re essentially done. What’s the burden at this point to go ahead and finish it?

Phil Favro

Thanks, Eric. Ruth?

Ruth Hauswirth

Really quick, I just underscore what Judge Goddard was saying about relevance and proportionality. I mean, the one thing that stood out for me, and I feel like I say this all the time, whether something is discoverable doesn’t turn on the form of the information, it’s the content. And so applying the rules of relevancy and proportionality is really how we navigate these kinds of things, regardless of what the type of information is. And so that was one of my just general takeaways of reading these very interesting opinions.

Phil Favro

Thanks, Ruth, Eric, and Judge. And I agree with the judge, these are tricky decisions, but I think you’ve got great counsel, you’ve got great judges involved. I think that we will enjoy, again, from the sideline, looking and seeing how these particular cases play out and how this, I should say, particular case, with all of its discovery decisions, plays out. And we will just have to see where things go because there isn’t a lot of case law on these particular issues quite yet.

So with that, let’s segue now to our last section dealing with ESI search, production, and evidence cases. So we’ve got a number of decisions we want to cover here. Let’s go ahead and segue to our first one on possession, custody, or control. So we’ve got just a little background feed. So if you hear any background, just go ahead, and we’ll try to make it so that the audience can hear us a little bit better.

So for possession, custody, or control, there is no clear set of rules that are applied across every state and across every single government, or excuse me, the federal space. Instead, you have a patchwork of rules. Most of them involve a legal right test. Do you have a legal right to get information? And we see in the Allergan versus Revance case that the special master applies essentially a legal right test and holds that, looking at the BYOD policy that Revance had in this particular instance, which disclaimed any sort of ownership of personal devices, that Revance was not going to have to preserve and produce information from its employee phones.

Essentially, that Allergan was going to have to go out and subpoena the information if it wanted to get it. So that makes things more difficult for Allergan, and at some level, it makes things more difficult for Revance employees. And so if you’re, as a company, if that’s the way you’re going, then there could be some challenges there, but that’s not how every single possession, custody, or control case comes out. And I think, Ruth, I know you were going to talk about this. On the one hand, there’s Allergan versus Revance, but on the other hand, there are cases like the Miramontes case. We don’t have it up on the screen, but it’s from three years ago, where the judge said, “I’m just going to cut through all of this stuff about formalities with legal right, and I’m going to see what the company’s doing.”

And in that particular case, the Miramontes case, the judges said, “The company is implicitly requiring the use of cell phones. Therefore, they have possession, custody, or control of these particular items, even if there isn’t a specific policy addressing these issues.” So it’s a moving target, Ruth. How do you deal with these different possession, custody, or control rules and rulings that seem to be all over the map?

Ruth Hauswirth

Yeah. So we could all transport back just about 45 minutes ago when we started talking about retention policies and other kinds of policies. And this area, as we can see from the two cases that you just described, the wording of the policy, the intent of the policy, how the policy is enforced and understood at a company, is critical. And so this is an example, the issue of how possession, custody, and control is applied, looking at real-world examples of how that front end of how you’re managing information, how you’re managing electronic communications, and what the employees of an organization, what they understand and what type of information is being relayed to them is critical. And so, before anything happens, is when to be thinking about these issues, because we do have courts going different ways depending on what the policy actually says.

And I think that all I can say about this area, possession, custody, and control, is one of those areas that is very fact-specific and instance-determinative. Have clear policies, understand how those policies are being understood and applied within the company before there’s an issue, and make sure that it aligns with the company’s understanding of what it intended with the policy. And have a clear device use policy, have a clear policy around messaging applications on end phones, on what’s allowed, what isn’t allowed. It takes a little bit of work, but the payoff, the dividends of doing that work on the front end, really pay off when you’re in an instance where you have to respond to certain government investigation demands or civil litigation demands.

Phil Favro

Thanks, Ruth. Any comments, Eric or Allison, about possession, custody, or control? Judge, sorry.

Judge Allison Goddard

No, that’s fine. So my only comment would be going back to the policies because the policies were so critical here. This is a great example of how AI can’t replace lawyers. You could probably ask any AI tool to draft, bring a BYOD policy for your company, and think, okay, well, this is good. We’ll throw it in. But if you haven’t thought through how it works or talked to a skilled litigator about what the pros and cons of that policy are, you need… Sure, use it for a first draft, but the experience people have from the cases that they’ve litigated or from the cases they know have been litigated is essential for your company. So this is just, to me, an example for me to plug. You’re not going away yet, lawyers. You still have jobs for now.

Phil Favro

Oh, that’s good. Thank you. Eric, you get the final word here. 30 seconds on this.

Eric Stansell

I hate mobile phones. I hate mobile phones, and I hate mobile phones. And I won’t say much more than that, other than I can assure you that what this pushes you to is having to figure out what the most restrictive policy is in some ways. I do think that what Ruth and the judge said is exactly right: a policy upfront is what you absolutely have to have, and you have to enforce that policy. And I won’t even get into company-owned phones versus BYOD. There’s an entire layer of complexity that’s there, but this is a very important area, and it’s probably that, per unit of value of relevant information, it has the most burden of any other data source.

Phil Favro

Yeah. Yes. And imagine now with the consumer-grade AI tools and apps being placed on phones, that’s going to increase the complexity. So Eric, you can keep your long and great hate for these phones and the complexities-

Eric Stansell

It’s there. Will be for a long time. Yes.

Phil Favro

So let’s segue now to a discussion of ESI search. And the Mosaic case that we’ve got up on the screen is, I guess, not particularly noteworthy. There’s squabbling between the parties over probably about like a hundred search terms or something like that, 70 search terms. And ultimately, the judge in that particular case makes a decision on which search terms are proper, which ones aren’t, and so forth. The point of raising Mosaic is not to talk about another search term case that we’ve been reading for the last 20-plus years. It’s to highlight whether it’s a trend really to use AI to conduct search, or, like in the Mosaic case, which is again, another copyright infringement case dealing with artificial intelligence, are our search terms continuing to predominate? We hear a lot of buzz about AI for search, but are lawyers just sticking with search terms, except for the most sophisticated outside counsel like Cooley and Ruth?

Ruth, I mean, you’re surely using AI at some level, but what do you think about this? Is it a trend? Is it not? Or is it just a mixed bag? We’re going to continue to see both.

Ruth Hauswirth

So what I’m seeing and what I suspect will continue is that we will continue using search terms along with other methods in a multimodal approach to how we search, review, and analyze information, which will best serve how we practice. And I don’t see us going yet, and I don’t have a crystal ball, but we’re still going to be using search terms, we’re still going to be using TAR 2.0, continuous active learning, and we’re still using, and we are now incorporating generative AI into the process.

And what I’m seeing is real benefits to it, because there’s just still so much information that we have to, even in relatively cases that aren’t that complicated, just the amount of data that organizations are creating, the potential universe is so large, we can’t just have all of that data put into a generative AI tool at this point. It’s cost-prohibitive, and it’s not practical at these early stages. And so the tried and true traditional methods that we’ve been using, I think, are going to partner with the newer emerging tools, and that’s what I’m seeing. And I think that the… I don’t see search terms going away. I think we’re going to continue to have this multimodal approach.

Phil Favro

Yeah, very good. Very good. Any comments, Judge and Eric, on this particular decision and this issue of trends?

Eric Stansell

The only thing I would say is that I think Ruth hit it on the nail on the head. The structure of the way that hosting costs and everything right now, I can tell you that we have options and we use both terms and AI, but we are nowhere close to right now to saying, “Well, we’re just going to run AI prompts over it, over this whole set of documents.” That we’re not ready for that right now at all, from, again, almost wholly from a cost perspective.

Phil Favro

Thanks, Eric. Judge, final word on this one.

Judge Allison Goddard

Yeah, I was just going to say, I still see search terms all the time, but can I just say, if there’s an eDiscovery hall of fame for judges, Judge Cisneros would be an early entry. I mean, she has this case, she has the Uber Sexual Assault MDL, and for lawyers to ask a judge to go through search firms one through 99 and make a call on them, I think, is almost malpractice. You should figure out a way to compromise on that. It’s really hard for judges to just step into a case. Search term disputes are really, really difficult overall. So I just hats off to Judge Cisneros for keep on keeping on.

Phil Favro

Thanks, Judge Goddard. And thanks, Eric and Ruth, for that. Another interesting decision, a different context deals with the production of contextual messages. And we’ve seen other decisions on this, particularly in the most prominent case, the Lubrizol case, the Northern District of Ohio, from about three years ago, where a judge ordered IBM to produce contextual messages surrounding its Slack chats and then for Lubrizol to produce contextual messages surrounding its Teams chats. And the Soqui case here basically renders a similar decision following the Lubrizol case and saying to the defendant, “You’ve got to produce contextual messages. You can’t just produce only the messages that hit on search terms.” And it’s interesting, I find it a little bit surprising, maybe on the one hand, but maybe on the other no, that the parties couldn’t work this out. I mean, I’m working as a special master in a particular case. This issue came up before me in my case, and I directed the parties to work it out.

I didn’t want to put my thumb on the scale. I wanted the parties to come up with the solution that best worked for each of them, since each of them was using collaboration platforms and would have to produce the information and discovery. And where it’s symmetrical, it seems like a no-brainer, even though in the Lubrizol case, it wasn’t. I tell you, Eric, you’ve read this decision. What do you think? What’s the right way to go about handling these issues?

Eric Stansell

Yeah. I mean, I think you hit the nail on the head when you said that about symmetrical earlier. In this case, it is asymmetrical. So there is some value, I guess, from the plaintiff’s perspective in going in this direction. I can tell you that I would find it very difficult when working with our outside counsel to not counsel them to go ahead and let’s not fight this fight. Let’s not fight a context fight. Yeah. If it hit on search terms, we’re going to obviously, and it’s relevant, we’re going to produce that, but we also are going to have to produce around that whatever provides context to that particular hit, that conversation. Now, does that mean that a three-month back and forth between two particular custodians was going to produce all of that? No. But-… relevant to that conversation, I think that that’s not a fight I want to fight. That’s not a hill I want to die on. That’s not somewhere I want to lose political capital with the court, with opposing counsel. I cannot believe that England would even consider going in the direction they went in. So this is one that surprised me. It was so, to me, it’s so straightforward.

Phil Favro

Yeah. And maybe Ruth, you could comment here just very briefly. I think it actually could be symmetrical because the plaintiff’s got to produce text messages almost certainly, and you’re going to be producing contextual messages surrounding text conversations. So, a single plaintiff is not going to have a chat tool; they are going to have those text tools. So I don’t know, Ruth, any comments here?

Ruth Hauswirth

Just that it’s fascinating to me as we continue to see these different types of ways that people are communicating now, how we as lawyers have to define what a document is. And we’re in a territory where we’re going to have to expand how we think about that. And so that’s what I think these kinds of issues raise. And I also agree that with the symmetrical, everyone carries a phone, and everyone uses text. So, almost in every situation, if it’s relevant, text messages are probably going to be involved.

Phil Favro

Thanks, Ruth. Let’s cover our last case. And Judge, I want to get your feedback on this. This is the Ninth Circuit’s recent decision in Allen, dealing with the admission of evidence. It’s under a criminal case, but it’s worth spotlighting because parties spend all this time and money and emotion and effort fighting to get the so-called smoking gun ESI. But what if you can’t use it at the hearing on the preliminary injunction, summary judgment, or trial? What good was all of that time and money? And I think the Allen case spotlights one way to go about doing things, an effective way in terms of using Rule 90211 and the corresponding analog, or, should say the corresponding rule for the business records exception, to get Facebook activity admitted to show that Allen was the perpetrator and the organizer of this drug ring and this weapons ring in the Spokane area.

And a hat tip to one of my students at Gonzaga Law, who pointed this case out to me last spring, saying it was under appeal, and the case ended up coming down. I just think it’s fascinating to see this particular decision and to see how the evidence was admitted and used to get a conviction, and then it was affirmed on appeal. You’re dealing with evidentiary issues all the time, Judge. What do you take from Allen and, more generally, from this proposition that you’d better hope you can use the evidence if you’re going to try and get it in discovery?

Judge Allison Goddard

The best lawyers are thinking about that at every step of the process. And I can’t tell you how many cases I’m able to get the parties who don’t want to settle to turn to change their opinion, especially if we do like a settlement conference right before trial and they’ll tell me all the great information they have and they’re going to win and I’ll say, “How are you going to get that in?” And it’s crickets. So you need to be thinking about how you’re going to lay a foundation to get that evidence in at the end of the day. And we even put in our Rule 26 guidelines in the Southern District, if it’s going to be AI-generated evidence, you need to be talking to the parties about that at every step, because we need to know if we need to set any dates for motion practice on whether or not it’s admissible.

So the best lawyers are thinking about that, and they’re thinking about that from the day they ask for discovery to the day they get it to the day they’re approaching trial.

Phil Favro

Thanks, Judge. Eric and Ruth, 20 seconds for any closing comments on this particular issue.

Ruth Hauswirth

I’ll just say, I couldn’t have said it better than what Judge Goddard said. And it is not usually on folks’ minds about how the particular information is going to be authenticated and admissible. So it’s a really great case. I’m glad you highlighted it.

Phil Favro

Thank you.

Judge Allison Goddard

Kudos to your student.

Phil Favro

Yes. Yeah. My student did an excellent job spotlighting this, and it’s going to be in my curriculum for this upcoming semester, which we’re in right now, when we talk about evidence. So anyway, thank you, Judge. And you can see for the audience, the resources up on screen, some of the items that you can look at that cover these particular issues, and I’d like to turn the time back over to Mary Mack, now, who’s going to finish up things. But let me just say thank you, Eric, Ruth, and Judge Goddard. It’s been terrific, great insights on the issues. Thank you for your efforts in reviewing all the cases and for your insights. Thank you very much.

Judge Allison Goddard

Thank you, Phil.

Mary Mack

Indeed. Thank you, Phil, for wonderful moderating, and thank all of you for joining today’s HaystackID webcast. Thank you to all our panelists for sharing their expertise. And before closing, please mark your calendar for HaystackID’s next webcast, Real Benefits, Real Constraints, a Practical Guide to Copilot Rollout, happening on Wednesday, February 25th at 12:00 PM Eastern. You can find the registration link in today’s resources, and we hope to see you there. And on behalf of EDRM, sincere appreciation is extended for your participation today, and we wish everybody a productive day. Thank you.

Expert Panelists

+ Philip Favro (Moderator)

Founder of Favro Law PLLC

Philip Favro is the founder and president of Favro Law PLLC, where he counsels clients on ESI, AI, and discovery issues and serves as a special master, mediator, and expert witness. Phil is nationally recognized for his expertise on ESI, discovery, and information governance, with courts acknowledging his credentials. See, e.g., Oakley v. MSG Networks, Inc., No. 17-CV-6903 (RJS), 2025 WL 2061665 (S.D.N.Y. July 23, 2025). This background makes Phil particularly well-suited to counsel clients and advise courts on information-related issues. As a special master, Phil is acclaimed for his collaborative approach, working with parties to find stipulated solutions to complex issues. For disputes that require adjudication, he is renowned for the clarity and vigor of his written dispositions.

+ Judge Allison H. Goddard

Magistrate Judge, U.S. District Court, Southern District of California

Judge Allison H. Goddard has served as a Magistrate Judge for the U.S. District Court, Southern District of California, since 2019. She presents regularly on e-discovery and technology and the law for the Federal Judicial Center, The Sedona Conference, and at legal conferences throughout the country. She recently spoke on “AI and the Courts in the Digital Era” at the Congressional Office for International Leadership Conference in Mongolia and has traveled to Uzbekistan to speak to its Supreme Judicial Council about judicial independence. Judge Goddard also coaches a local high school mock trial team.

+ Ruth C. Hauswirth

Special Counsel and Head of Litigation & eDiscovery, Cooley LLP

Ruth C. Hauswirth bridges the gap between legal strategy and emerging technology. As Special Counsel and Head of Litigation & eDiscovery at Cooley LLP, she founded and leads the firm’s innovative eDiscovery and Information Governance practice. With over 30 years of experience navigating complex litigation and the evolving landscape of electronically stored information, Ruth delivers strategic, defensible solutions that integrate emerging technologies, including Generative AI, to optimize legal outcomes. Ruth is also a Certified AI Governance Professional through the International Association of Privacy Professionals. She is also a trusted advisor to clients on litigation readiness and enterprise information governance, with work spanning policy development, implementation, and regulatory compliance across global jurisdictions.  In 2023, Ruth was named a Relativity AI Visionary and is a sought-after and frequent speaker on eDiscovery, GenAI, and the intersection of law and technology. Ruth served as an adjunct professor teaching eDiscovery Law at the University of San Diego School of Law between 2015–2023, and also previously served as Assistant Dean for Law and Technology at Santa Clara University School of Law. Ruth holds a J.D. summa cum laude from Santa Clara University School of Law and is a member of the California State Bar since 1994.  Ruth is also a long-time meditator and trained mindfulness meditation facilitator through UCLA. Ruth regularly leads workshops for lawyers and legal professionals to integrate mindfulness into their practice for greater fulfillment and well-being. Full Bio: cooley.com/people/ruth-hauswirth

+ Eric Stansell

Senior Counsel, Discovery, Tyson Foods

Eric Stansell serves as Tyson Foods’ Senior Counsel for Discovery. In this role, Eric works with Tyson’s legal team and outside counsel to ensure that Tyson’s Discovery obligations and interests are properly addressed. He is also responsible for leading the development and implementation of eDiscovery best practices, management of eDiscovery vendor-partner relationships, and evaluation of new eDiscovery-oriented partners and technologies. Additionally, Eric also works with internal partners on strategic, enterprise-wide data retention and deletion policies, and collaborates with internal stakeholders to streamline collection processes. Prior to his current role, Eric led teams and initiatives focused on technology-assisted review for one of the country’s largest alternative legal service providers. Over his 15-year career in E-Discovery, he has worked on hundreds of matters, saving his clients millions of dollars and helping them to reduce their legal risks. Before his E-Discovery career, Eric worked as an insurance regulator and in the political sphere. Eric received his BA, MBA, and JD from the University of Tennessee at Knoxville. He is married to Kristina, and they have an eight-year-old son, Alex, and a six-year-old daughter, Katie, all of whom assist him in his second job of wrangling Ozzy the Bernadoodle, Eevee the German Shepherd, and Callie the Calico cat.

About EDRM

Empowering the global leaders of e-discovery, the Electronic Discovery Reference Model (EDRM) creates practical global resources to improve e-discovery, privacy, security, and information governance. Since 2005, EDRM has delivered leadership, standards, tools, guides, and test datasets to strengthen best practices throughout the world. EDRM has an international presence in 145 countries, spanning six continents. EDRM provides an innovative support infrastructure for individuals, law firms, corporations, and government organizations seeking to improve the practice and provision of data and legal discovery with 19 active projects.

About HaystackID®

HaystackID® solves complex data challenges related to legal, compliance, regulatory, and cyber requirements. Core offerings include Global Advisory, Cybersecurity, Core Intelligence AI™, and ReviewRight® Global Managed Review, supported by its unified CoreFlex™ service interface. Recognized globally by industry leaders, including Chambers, Gartner, IDC, and Legaltech News, HaystackID helps corporations and legal practices manage data gravity, where information demands action, and workflow gravity, where critical requirements demand coordinated expertise, delivering innovative solutions with a continual focus on security, privacy, and integrity. Learn more at HaystackID.com.

Assisted by GAI and LLM technologies.

SOURCE: HaystackID