People have been waiting for the “Da Silva Moore of Generative AI” – the case that approves the use of GenAI in discovery review. It’s already here.

And, at least in my opinion, it has been here for a long time. Over fourteen years…and counting.

In my opinion, the “Da Silva Moore of Generative AI” is…wait for it…Da Silva Moore. Same as it ever was (now you get the image above*).

For those who may be new to eDiscovery, the reference to “Da Silva Moore” is to the case Da Silva Moore v. Publicis Groupe & MSL Group, where (in February 2012) New York Magistrate Judge Andrew J. Peck (now retired) declared the following: “This judicial opinion now recognizes that computer-assisted review is an acceptable way to search for relevant ESI in appropriate cases.”

Side note: Judge Peck called it “computer-assisted review” instead of the term that has become standard in our industry: “technology-assisted review”. I recently heard Judge Peck state that he felt that “CAR” was a better acronym than “TAR” because a “CAR” goes places and we get stuck in “TAR”. Frankly, I think “CAR” would have been a better acronym too.

Over the years, there have been many more case rulings involving the use of TAR: at least 85 of them as of publication of The Sedona Conference TAR Case Law Primer, Second Edition in May 2023 (probably over 100 by now). One of those was Rio Tinto PLC v. Vale S.A. in May 2015, which was also authored by Judge Peck, where he stated: “In the three years since Da Silva Moore, the case law has developed to the point that it is now black letter law that where the producing party wants to utilize TAR for document review, courts will permit it.” That was over eleven years ago.

Great, but what does that have to do with GenAI? People have been saying that we need a ruling that specifically approves the use of GenAI for review.

Back in February, after a debate on LinkedIn about whether we need a case law ruling approving the use of GenAI for review, I posed this question to Judge Peck himself: “Do you consider your ruling in Da Silva Moore to be an approval of computer-assisted review methods in general? Or only predictive coding specifically?” I told him that I was thinking of writing a blog post to essentially state what I just stated above.

Judge Peck essentially shot down my proposal, saying: “I don’t think I even contemplated that. Sorry.”

So, I put aside my plan to write about it. Months went by.

Then, in July, the ruling in Schulte v. LinkedIn Corp. came out. In it, California Magistrate Judge Laurel Beeler denied plaintiffs request to prohibit LinkedIn from applying search strings or keywords to pre-cull the custodial documents reviewed by Relativity aiR (which uses GenAI) and denied plaintiffs request to compel LinkedIn to disclose various metrics about its use of Relativity aiR. Very notable ruling.

But I think the most notable part of the ruling was when Judge Beeler said this:

“LinkedIn has satisfied its obligations under the Interim ESI Order. Paragraph 5(a) of the Order requires the producing party to ‘disclose to the receiving party if they intend to use Technology Assisted Review (‘TAR’) to filter out non-responsive documents.’…On May 15, 2026, LinkedIn disclosed to the plaintiffs that it would use Relativity aiR — a form of Technology Assisted Review — to filter out non-responsive documents”.

“Relativity aiR — a form of Technology Assisted Review”. Exactly.

We tend to refer to the “old” technology as “TAR” and the “new” technology as “AI”. But it’s more precise to call them “predictive coding” and “generative AI” respectively. People like to speak in shorthand, even if it can sometimes lead to confusion – this is why many of us use the term “Kleenex” instead of “facial tissue”, even if the “facial tissue” is another brand. It’s simply easier to say.

Predictive coding and generative AI are both forms of AI. And they’re both forms of TAR as well.

So, far be it from me to disagree with Judge Peck about his own landmark case, but I’m going to anyway. 😉 My contention is that when Judge Peck effectively said in Da Silva Moore that “[TAR] is an acceptable way to search for relevant ESI in appropriate cases”, I think that applies to any form of TAR.

Even though Judge Peck in Da Silva Moore detailed the parties’ negotiation of an agreed protocol of what came to be known as a TAR 1.0 approach, I don’t think that just was a case just approving that particular approach – I see it as a case approving the use of technology to assist in the review process generally.

Case in point: “Da Silva Moore” may have involved TAR 1.0, but Judge Peck referenced the study Technology–Assisted Review in E–Discovery Can Be More Effective and More Efficient Than Exhaustive Manual Review (by Maura R. Grossman & Gordon V. Cormack) twice within that ruling, which advocated for a “continuous active learning” (CAL), aka TAR 2.0 approach, which had been in use at Grossman’s firm since 2009 (the term “CAL” wasn’t coined until about 2013, though four of the five systems tested in the study were a CAL-based approach).**

With his mentions of that study in Da Silva Moore (along with a mention of the study by Herb Roitblatt, Anne Kershaw, and Patrick Oot titled Document categorization in legal electronic discovery: computer classification vs. manual review), Judge Peck wasn’t limiting his general approval of TAR to just TAR 1.0. There are fundamental differences between the workflows associated with TAR 1.0 and those associated with CAL. Per the Sedona TAR Case Law Primer, Rio Tinto was the first case that discussed the use of CAL (by noting that the contents of the seed set in CAL “is much less significant”), but Judge Peck didn’t need to bless the use of CAL explicitly (he had already done so implicitly in Da Silva Moore). In fact, he said: “as in Da Silva Moore, the Court’s approval “does not mean … that the exact ESI protocol approved here will be appropriate in all [or any] future cases that utilize [TAR].”

That’s an important point. There’s a difference between approving a specific approach for a given case and a general acceptance of the use of technology to assist with review. Every case is different, with different circumstances and requirements, regardless of what approach – TAR 1.0, CAL, GenAI – is used. There is variability in workflows to each of those approaches – “one size does not fit all” with any of them. Courts have been ruling and will continue to rule on the specific approach in the case based on the circumstances of that case, either to accept the agreed approach between the parties or to resolve disagreements on the approach. That will never change.

But, in my opinion, the question of whether GenAI is generally approved for discovery review is settled. The “Da Silva Moore of Generative AI” is Da Silva Moore. Judge Peck approved the use of technology to assist in the review process. By referencing Relativity aiR as a form of TAR, Judge Beeler reinforced that approval applies to GenAI too.

So, what do you think? Do you agree that the “Da Silva Moore of Generative AI” is Da Silva Moore? Or are you still waiting for that case that approves the use of GenAI for review? Please share any comments you might have or if you’d like to know more about a particular topic.

*If you’re too young to get the image, this will help. 🙂

**Thanks to Maura R. Grossman for pointing out the inaccuracies in the timing of the adoption of CAL.

Disclaimer: The views represented herein are exclusively the views of the author, and do not necessarily represent the views held by my employer, my partners or my clients. eDiscovery Today is made available solely for educational purposes to provide general information about general eDiscovery principles and not to provide specific legal advice applicable to any particular circumstance. eDiscovery Today should not be used as a substitute for competent legal advice from a lawyer you have retained and who has agreed to represent you.