The American duty to preserve electronically stored information does not begin when a complaint is filed. It begins, in Judge Shira Scheindlin's now-canonical formulation, the moment a party "reasonably anticipates litigation." That standard was articulated in Zubulake v. UBS Warburg LLC, 220 F.R.D. 212, 218 (S.D.N.Y. 2003) (Zubulake IV), which held that "[o]nce a party reasonably anticipates litigation, it must suspend its routine document retention/destruction policy and put in place a 'litigation hold' to ensure the preservation of relevant documents." Anticipation is an objective inquiry, not a subjective one, and it can be triggered well before any pleading is on file.
What concretely triggers anticipation? In Zubulake itself, the duty attached when the plaintiff filed her EEOC charge in August 2001 — long before suit was filed in federal court. More broadly, courts following the Zubulake framework treat demand letters, pre-suit correspondence threatening litigation, internal investigations into wrongdoing that could foreseeably lead to claims, and government subpoenas as classic triggers. The common thread is foreseeability: if a reasonable party in the same position would expect a dispute, the duty attaches even if no lawsuit has yet been threatened in writing.
The consequence of getting the trigger date wrong is severe. Zubulake IV set the elements for an adverse inference instruction: "(1) that the party having control over the evidence had an obligation to preserve it at the time it was destroyed; (2) that the records were destroyed with a 'culpable state of mind' and (3) that the destroyed evidence was 'relevant' to the party's claim or defense." 220 F.R.D. at 220. Element one turns entirely on when the duty attached. A trigger date set conservatively early is the cheapest insurance an organization can buy against a future spoliation motion; a date set late, or never set at all, exposes the producing party to the full sanctions ladder of Rule 37(e).
If Zubulake defined the trigger, Pension Committee of the University of Montreal Pension Plan v. Banc of America Securities, LLC, No. 05 Civ. 9016 (SAS), 2010 WL 184312 (S.D.N.Y. Jan. 15, 2010) defined the floor. Often called "Zubulake Revisited," Pension Committee was Judge Scheindlin's seven-years-later answer to the question of what counsel must actually do once the duty attaches. Her answer was unambiguous: an oral instruction is not a litigation hold. A litigation hold must be written, it must reach the right people, and it must be paired with a mechanism to actually collect the preserved material.
The court reviewed the conduct of thirteen plaintiffs who had relied on informal, ad hoc preservation instructions and on custodial self-collection — the practice of letting individual employees decide which of their own documents were responsive. Judge Scheindlin found that this fell below the standard of care: "[t]his instruction does not meet the standard for a litigation hold. It does not direct employees to preserve all relevant records — both paper and electronic — nor does it create a mechanism for collecting the preserved records so that they can be searched by someone other than the employee." 2010 WL 184312, at *7. After 2004, she concluded, "the failure to adhere to contemporary standards can be considered gross negligence."
The opinion identified three specific failures that, on their own, support a finding of gross negligence: failure to issue a written hold, failure to identify and notify key players, and failure to suspend ordinary email deletion. Judge Scheindlin later issued a clarifying order on May 28, 2010 narrowing the per-se reach of the gross-negligence finding to employees actually involved in the subject matter, but the core holding stands. The takeaway for in-house counsel is procedural and unforgiving: if you cannot produce, after the fact, the written hold notice, the list of recipients, and the acknowledgment log, you are arguing about culpability from the back foot.
It is also worth noting Zubulake V's companion holding: counsel cannot simply send a hold notice and walk away. Zubulake v. UBS Warburg LLC, 229 F.R.D. 422 (S.D.N.Y. 2004) held that "it is not sufficient to notify all employees of a litigation hold and expect that the party will then retain and produce all relevant information. Counsel must take affirmative steps to monitor compliance." A defensible hold program is a continuous obligation, not a one-time email.
A defensible hold notice does five things in plain English. First, it identifies the matter and the trigger date, so that a court reviewing the notice years later can reconstruct when the duty attached. Second, it defines the scope of relevant material — the topics, custodians, time period, and data sources covered — with enough specificity that a recipient knows what to keep, but with enough breadth that no foreseeably-relevant category is excluded. Third, it expressly suspends routine deletion: auto-delete on email, document-management retention purges, BYOD wipe policies, and any backup-tape recycling that might overwrite responsive ESI. Zubulake IV's indictment of UBS turned in part on the failure to halt backup-tape recycling.
Fourth, the notice must require an affirmative acknowledgment from each recipient — not silent receipt. The acknowledgment is the evidentiary record that satisfies the Pension Committee requirement of a written hold reaching identified key players, and it is the document opposing counsel will subpoena first if a spoliation motion is filed. Fifth, the notice must commit to a refresh cadence: every six months at minimum, and more frequently when custodians change roles, leave the company, or when the scope of the dispute expands. Zubulake V's monitoring requirement is satisfied by a documented refresh process, not by hope.
Practical drafting points repeat across well-run hold programs. Identify data sources by name, not by category — "your Outlook mailbox, your OneDrive folder, your Slack DMs, your iPhone text messages" beats "all electronic communications." Name the custodians individually rather than listing "the marketing department." Give a contact for questions, because recipients with questions are recipients who are trying to comply. And keep the notice short: a two-page hold that gets read is more defensible than a ten-page hold that gets archived unread.
The white paper accompanying this article includes a two-page annotated litigation hold notice template — a real, drop-in usable sample with margin annotations explaining why each clause exists and which specific holding it satisfies. Readers can adapt it to their own matters.
Before December 2015, federal courts were divided on the culpability required for the most severe ESI sanctions. Some circuits permitted adverse inference instructions on a finding of mere negligence; others required intent. The 2015 amendment to Federal Rule of Civil Procedure 37(e) resolved the split by imposing a uniform two-tier framework. Tier one, codified in Rule 37(e)(1), applies when ESI that should have been preserved is lost because a party "failed to take reasonable steps to preserve it" and the loss cannot be cured through additional discovery. The court "may order measures no greater than necessary to cure the prejudice." Tier two, Rule 37(e)(2), is triggered only on a finding that the party "acted with the intent to deprive another party of the information's use in the litigation," and only then may the court impose adverse-inference instructions, dismissal, or default judgment.
The Advisory Committee notes are blunt about the policy: "negligent or even grossly negligent behavior does not logically support that inference" of unfavorable evidence. The amendment was a deliberate rejection of the broader culpability standards used by some pre-2015 cases. Negligent loss of ESI now produces curative measures proportionate to the prejudice; intentional spoliation produces the full sanctions ladder, including case-terminating sanctions in egregious cases.
The leading cases applying the amended rule track this distinction precisely. In CAT3, LLC v. Black Lineage, Inc., 164 F. Supp. 3d 488, 498 (S.D.N.Y. 2016), one of the earliest reported applications of the amended rule, the court found that the plaintiffs had failed to take reasonable steps to preserve relevant emails and that the defendants had been prejudiced — but it declined to impose an adverse inference because there was no finding of intent to deprive. Instead, the court precluded the plaintiffs from relying on their version of the disputed emails. In Matthew Enterprise, Inc. v. Chrysler Group LLC, 2016 WL 2957132, at *6 (N.D. Cal. May 23, 2016), the court applied Rule 37(e)(1) to a party that had failed to halt automatic deletion of customer communications by a third-party vendor after receiving a litigation threat letter, again limiting the remedy to a curative measure rather than an adverse inference.
The intent-to-deprive cases sit at the other end of the spectrum. In Klipsch Group, Inc. v. ePRO E-Commerce Ltd., 880 F.3d 620, 627 (2d Cir. 2018), the Second Circuit affirmed substantial monetary sanctions after multiple custodians used data-wiping software on their devices during an active litigation hold — conduct the court treated as willful spoliation under Rule 37(e)(2). The line between "we should have done more" and "we deliberately destroyed evidence" is the line between a curative ruling and a case-terminating one, and it is drawn by the contemporaneous record of the hold program.
For litigators with cross-border exposure, the United Kingdom's Disclosure Pilot Scheme — originally Practice Direction 51U (in force from January 2019 in the Business and Property Courts) and now Practice Direction 57AD (permanent since October 2022) — imposes preservation duties that are doctrinally distinct from the American framework but functionally overlapping. PD 57AD imposes a duty on parties and their legal representatives to take reasonable steps to preserve documents in their control that may be relevant to any issue in the proceedings, and to disclose known adverse documents as an ongoing obligation regardless of which disclosure model the court orders.
The Disclosure Pilot replaces a single default disclosure standard with a menu of five disclosure models (Models A through E), ranging from "no disclosure" through "narrow search-based disclosure" to the broadest train-of-inquiry searches reserved for exceptional cases. For any of the search-based models (C, D, or E), parties must complete a Disclosure Review Document (DRD) which identifies the Issues for Disclosure, the custodians, the data sources, and the proposed search and processing methodology. The DRD is the procedural cousin of the American litigation hold scope memo, and it forces the same conversations — custodians, sources, scope — into a court-supervised written record at an early stage.
The English courts have been quick to police the boundaries. In McParland & Partners Ltd v Whitehead [2020] EWHC 298 (Ch) at [29], the Chancellor held that "Issues for Disclosure" under the Pilot are confined to "key issues in dispute, which the parties consider will need to be determined by the court with some reference to contemporaneous documents" — not every issue raised in the pleadings. In Lonestar Communications Corporation LLC v Kaye [2020] EWHC 1890 (Comm) at [73], the Commercial Court emphasized that Model C requests must be "narrowly focussed" so that the disclosing party can actually run defensible searches. Both decisions track the same proportionality instinct that animates the American Rule 26(b)(1) framework, even though the procedural mechanics are different.
For practitioners managing parallel US and UK proceedings, the practical convergence is this: a written, scoped, custodian-specific preservation regime — documented in a US litigation hold notice and reflected in a UK Disclosure Review Document — will satisfy both regimes without doubling the work. The DRD's certification of compliance and the US hold notice's acknowledgment log serve the same evidentiary purpose: a paper trail showing that counsel understood the duty, scoped it intelligently, and supervised its execution.
A litigation hold is only as defensible as the record of its execution. DecoverAI is built around the assumption that the contemporaneous record — who was notified when, who acknowledged, what data sources were preserved, what was collected, what was reviewed — should live in one auditable place rather than scattered across email threads, spreadsheets, and a shared drive. The platform's chronology viewer reconstructs the complete timeline of a matter from collection forward, which is the first thing a litigator wants when defending a Rule 37(e) motion or a Pension Committee-style sanctions motion.
Once data is collected under a hold, DecoverAI's AI-powered relevance and privilege workflows process it under the same defensibility constraints courts apply to traditional review. Reasonable steps under Rule 37(e)(1) are not just about preservation — they extend to validated, documented search methodology on the back end, and to the ability to demonstrate that the workflow used was the workflow disclosed. Our pricing is published and predictable: $60 per gigabyte per month, all-in, with no per-user fees and no contracts. That structure removes the financial incentive to under-preserve in marginal matters and makes early case assessment a default rather than a luxury.
For a worked example of how this plays out in practice, see our Tax Credit Investigation case study: 30,000 documents collected under a litigation hold, processed in three days, with full privilege log generation and a 98% cost reduction relative to the traditional vendor estimate. For broader workflow context, the early case assessment page walks through how preservation, collection, and ECA fit together as a single defensible motion. And for adjacent reading on the cost mechanics that make over-preservation so expensive on legacy platforms, see The Real Cost of Document Review: A 2026 Pricing Benchmark.