The Privilege Log Is Due Tomorrow: A Triage Plan

A privilege log due tomorrow: group the documents first, then log the groups.

Federal Rule of Civil Procedure 26(f) puts privilege and work-product protection claims on the agenda of the very first meet-and-confer, right beside preservation and the form of production. Charming, if you were there. If you are reading this at nine in the evening with a privilege log due tomorrow, that conference is a distant memory.

Still, the rule tells you something useful tonight. Privilege logging is a scoped, negotiable obligation. It is not a ritual of one row per email.

A privilege log is the list a withholding party serves to identify each document, or each category of documents, kept back as privileged, with enough description for the other side to test the claim.

Its format, rather absurdly, is seldom fixed by statute. In California it comes mainly from case law. So the format is often open to negotiation, and one short call to opposing counsel may settle more than a long night of review.

So here is my position, stated first. A log due tomorrow is salvageable. Not by coding every document overnight. The research section below walks through what real log fights show... so begin tonight by asking which documents even belong in the set.

What does a privilege log actually have to cover?

A privilege log must account for every document withheld on privilege grounds, whatever its date, and give opposing counsel enough detail to test each claim.

Before you open a single file tonight, I would settle three things:

1. Scope: which documents you are withholding, and from which date range.
2. Format: whether your court or local rule allows a categorical log.
3. Cutoff: whether communications created after the complaint fall inside the obligation at all.

A categorical privilege log is one that groups withheld documents by shared traits (participants, date range, privilege basis) and logs each group once. Rather a relief, at midnight.

The common assumption is that the only safe log has one row per email. Not quite. In the Lively and Wayfarer litigation in the Southern District of New York, the court's August 29, 2025 order allowed a categorical log under Local Rule 26.2(c) for a defined window of documents. When the fight came back, Judge Liman treated the Wayfarer Parties' proposal as a reasonable compromise that would suffice, for now.

For now. Two charming words... and a warning.

The rules themselves are oddly unwritten. In California, privilege log format comes mainly from case law, not from the Code of Civil Procedure or the Rules of Court. The burden sits with whoever asserts the privilege. And the log reaches every withheld document, regardless of when it was created or whether it was previously produced.

Then there is the dreadful alternative: no log at all. In Camacho v. Rose-Mary, Johanna Graselli Rehab., Inc., decided by Ohio's Eighth District Court of Appeals on July 25, 2024, a rehabilitation facility refused to produce any privilege log and refused redactions too. Its theory was that even reviewing residents' records to build a log would violate their privacy. The court noted that other litigants asserting the same physician-patient privilege over third-party records had still produced logs under Civ.R. 26(B). It added, rather pointedly, that it was "not at all clear" the facility even knew whether it held privileged material.

Refusal is not a strategy. It is simply a second motion, waiting.

Read together, the 4 legal sources here point one way. Ask opposing counsel one plain question tonight: which date window and which custodians must the log reach? In practice, the danger tonight is not brevity. It is a log nobody can check.

That is where verification earns its keep. At Relevant e-Discovery, every answer the platform gives traces to its source, linking back to the exact exhibit it came from, so an attorney can verify each privilege call with one click before filing. For the structural side of the same idea, see how teams go about designing a privilege gate for ESI before review even begins.

So the first triage question is not how many documents remain. It is what minimum, defensible log this court will actually accept...

Does an AI shortcut on the log create a new privilege problem?

Yes, when the tool retains or trains on what you feed it. Processing that stays single-tenant, or inside your own cloud account under your own keys, avoids that trap.

The worry is old; only the tool is new. The traditional rule holds that once a third party is privy to a privileged conversation, it is no longer privileged. Pour a midnight batch of attorney emails into a consumer AI service whose terms allow retention or model training, and you may have pulled up a chair for exactly that third party.

Oh dear.

So the shortcut carries a price. Before a single document leaves your hands tonight, I would put three questions to the tool:

1. Where does processing run, and does anyone else's data share that environment?
2. Does the vendor retain anything, or train a model on it?
3. Whose encryption keys protect the collection?

Those questions have plain answers on the right architecture. At Relevant e-Discovery, processing runs single-tenant, or inside your own AWS account under your own keys, with no vendor retention and no model training, so nothing you feed it trains anyone's model or waives privilege. In our view, in-account deployment under the client's own keys is a trust posture no consumer AI tool, and almost no small-firm tool, can match. Read the vendor paper closely all the same; it is worth checking whether your AI review vendor can swap models mid-matter before you hand over a privilege set at two in the morning.

The second tension is quieter, and rather more dreadful for being quiet. A rushed log will be tested. Perhaps next week, perhaps after discovery closes. What survives that test is not the speed of the drafting but the record behind it: immutable originals with content hashing, append-only audit trails, documented chain of custody, and a fail-closed privilege gate on production. A fail-closed gate is one that blocks a document from production by default whenever its privilege flag is unresolved. Our approach treats that record as the thing that holds up if opposing counsel challenges the process.

Speed without a trail. Quite useless, really.

Nor does exposure risk excuse skipping the work. The Ohio appellate court in Camacho was unmoved by a facility's claim that merely reviewing records to build a log would itself be the privacy harm. Other parties holding the same kind of sensitive third-party records had managed a compliant log. The answer to exposure is a safer review, not no review at all.

The tool, then, is a choice about where the documents sit and what remembers them afterward. The clock is a different matter, and it is ticking rather loudly now, which is why the order of the next few hours counts for more than their pace...

What is the step-by-step triage order for finishing a privilege log in under 24 hours?

Work in a fixed order: narrow the collection, screen for attorney names, group by family, log by category, then verify. Test any tool on your own messy collection first.

EDRM's review guide makes a quietly delicious point about deadlines: a DOJ second request with a 15-day turnaround might be handled differently than ordinary litigation on a longer schedule. Fifteen days. You have one. So the order of work is everything, and I would run it like this:

1. Narrow first. Pull the key custodians and the disputed date range, and set the rest aside. The same guide favors strategic collection of key custodian data over broad collection, plus filtering by date limits or keywords.
2. Collapse and group. Remove duplicates, then keep each email with its attachments and each thread intact. A privilege call made on a whole family is one call, not six.
3. Screen for lawyers. Search for the names of attorneys and law firms. EDRM notes such hits carry "a high probability" of attorney/client communications, and suggests routing them to one attorney whose task is to review for privilege and mark them.
4. Draft categories. Sort the flagged families into groups that share participants, date window and privilege basis, with a count for each.
5. Verify and record. Spot-check each category against its documents, write down how the screen was run, then serve.

Step one is where most midnight logs go wrong. Running it all from the mail client feels quick, and it is rather a trap; there are sound reasons searching Outlook isn't a defensible document review when the other side later asks how you found what you withheld.

Notice what the order does. Nobody codes a document line by line until the set has been cut, grouped and screened. The attorney's hours land on the documents most likely to be privileged, which is precisely where a judge will look.

Now, the economics. Relevant e-Discovery's position is blunt: AI-assisted review runs cents per document, against the dollars per document and roughly $19K/GB of manual review, a gap a firm can see on its own matter. That gap answers the perennial question of how to cut the cost of document review in litigation. It does not, on its own, tell you how many hours a categorical log saves on a deadline night; nothing on record measures that, and I would not pretend otherwise.

What I would insist on is a proper trial. Bring a messy collection and a hard question, and watch the tool read, code and cite your own evidence rather than reading a case study about someone else's.

Too sensible, perhaps. But consider the alternative.

Enterprise platforms such as Relativity and Everlaw are priced for larger shops and expect an administrator to run them. A solo or small firm working a single messy matter has no administrator at midnight. It has one attorney, one paralegal if lucky, and a clock that has stopped being polite... which is exactly when the categories themselves have to be written, and written well.

How can I reduce the cost of document review in litigation when the log is due tomorrow?

Cut the set before anyone reads it, log by category rather than by row, and spend attorney hours only on documents a name screen flags as likely privileged.

That is the whole economy of the thing. I expect categorical logs to become the ordinary compromise, with format settled at the first meet-and-confer instead of at midnight.

A candid aside. No public record I can point to measures how many hours a categorical log saves on a deadline night. The case for it rests on structure, not on a stopwatch.

And structure matters later. A log served tonight may be challenged again, possibly after discovery closes, and more likely through a request for in camera review than a demand for sanctions. What answers that request is your record of how each privilege call was made...