eDiscovery Reference

The EDRM (Electronic Discovery Reference Model) consists of: (1) Information Governance, (2) Identification, (3) Preservation, (4) Collection, (5) Processing, (6) Review, (7) Analysis, (8) Production, and (9) Presentation. The stages are sequential but can be iterative, with data volume decreasing at each stage in a funnel shape. Review typically accounts for 60-80% of total eDiscovery costs.

TAR 1.0 uses a seed set coded by a subject matter expert, trains a model in batches, and iterates until predictions stabilize. TAR 2.0 (Continuous Active Learning / CAL) starts with random documents, the algorithm continuously presents the most likely relevant documents for coding, and the model updates in real-time after each decision. CAL eliminates seed set bias, achieves higher Recall, and is the current mainstream approach approved since Da Silva Moore v. Publicis Groupe (2012).

Under the 2015 amendment to FRCP Rule 37(e), there are two tiers of sanctions. If a party fails to take reasonable steps to preserve ESI and it cannot be restored: (1) the court may order additional discovery or require the party to pay costs. If the party acted with intent to deprive the other side of information: (2) the court may give an adverse inference instruction, dismiss the action, or enter default judgment.

A litigation hold (legal hold) is a directive to preserve all potentially relevant ESI when litigation is reasonably anticipated. The notice must include: an overview of the litigation, scope of documents to preserve, preservation period, instruction to suspend auto-deletion, and acknowledgment request. Failure to issue a timely hold can result in spoliation sanctions as established in Zubulake v. UBS Warburg.

Under FRCP Rule 26(b)(1), discovery must be proportional to the needs of the case. Courts consider six factors: importance of the issues, amount in controversy, parties' relative access to information, parties' resources, importance of discovery in resolving issues, and whether burden/expense outweighs likely benefit. Proportionality arguments can limit custodian counts, date ranges, data sources, and support cost-shifting requests.

The three main production formats are: (1) TIFF/PDF with load files, which includes single-page TIFF images at 300 DPI with Bates stamps, accompanied by DAT files (metadata), OPT files (image mapping), and extracted text files; (2) Native format preserving original files with full metadata, best for Excel and PowerPoint; and (3) Hybrid combining both. Bates numbering uses a PREFIX-000001 format to assign unique identifiers to every page.

International eDiscovery faces multiple challenges: EU GDPR requires lawful processing basis and restricts cross-border transfers (Standard Contractual Clauses or adequacy decisions needed); blocking statutes in France, Switzerland, and China restrict domestic data export; data localization laws in Russia, China, and India require in-country storage. Practical strategies include local counsel, in-country review, anonymization/pseudonymization, and Data Protection Impact Assessments.

Generative AI is adding capabilities beyond traditional TAR: automated document summarization, privilege review assistance, timeline generation, search term suggestions, and coding consistency checks. Key tools include Relativity aiR, DISCO Cecilia, Everlaw EvA, and Reveal Ask. Adoption considerations include hallucination risk, confidential information leakage prevention, explainability requirements, and maintaining attorney supervision obligations.