Abstract
On 5 December 2024 the global e-discovery community marked the tenth anniversary of E-Discovery Day. More than 4 600 practitioners joined virtual panels while satellite meet-ups convened in nine U.S. cities. This report distils the technical take-aways, releases the first open-data bundle of mobile-forensics images, and outlines the 2025 challenge that invites teams to validate deep-fake detection tools against a crowd-sourced corpus.
Keynote Highlights
Mary Mack, Doug Austin and Mike Quartararo traced volume growth from 90 GB per matter in 2014 to 4.2 TB in 2024, attributing the 47-fold increase to ephemeral chat and sensor data. The speakers published an updated EDRM diagram that adds “Collaboration & IoT” as a ninth stage and released the SVG under CC-BY 4.0.
Deep-Fake Evidence Work-Group
Hon. Ralph Artigliere, Prof. Maura Grossman and Ralph Losey proposed a four-step authenticity test: cryptographic hash at capture, sensor attestation, chronograph hash chain and transformer-based artefact detection. A 1 020-file reference set containing synthetic audio and video is now hosted by the University of Waterloo; ground-truth labels are signed with CRYSTALS-Dilithium to ensure post-quantum integrity.
Hyperlinked Files Debate
Kelly Twigger presented empirical data showing that 38 % of “modern attachments” in O365 environments reside on third-party domains outside legal-hold scope. The group agreed on a working definition: “any URI referenced in a message body that is necessary to render the author’s intent.” A PowerShell script that expands bit.ly, t.co and SharePoint-generated links is available in the EDRM GitHub; it outputs a CSV suitable for Relativity or Nuix ingestion.
Mobile Collections Reality Check
Jerry Bui and Matthew Hamilton shared a 2024 Android image where 72 % of relevant chat artefacts lived in encrypted “App Storage” containers. A companion open-source wrapper for ADB bypasses screen-lock on Pixel devices running GrapheneOS 14 and exports a logical tarball that can be processed by the free Autopsy plugin. The wrapper is released under GPLv3; no commercial licence is required.
AI in Discovery—Old vs. New
Jakub Weberschinke demonstrated that transformer models fine-tuned on 1.2 million solicitor-reviewed documents reduce first-pass privilege miss-rate to 2.1 % compared with 8.9 % for legacy latent semantic analysis. The model weights are released in ONNX format and can be run locally on an M2 MacBook Air with 8 GB RAM, eliminating cloud-transfer concerns.
ReedSmith Debate Outcome
Therese Craparo and Anthony Diana sparred on whether technology-assisted review should be mandatory. Audience polling showed 61 % in favour of a presumptive rule, 24 % against and 15 % undecided. A draft rule text will be submitted to the US Judicial Conference advisory committee in January 2026.
Gayle O’Connor Spirit Award
The 2024 recipient, Maribel Rivera, was recognised for creating a mentorship network that has placed 312 newcomers into e-discovery roles since 2021. Rivera open-sourced the mentor-matching spreadsheet template; it is now translated into four languages.
Open-Data Bundle Released
To accelerate reproducible research, organisers published:
- 50 fully imaged mobile devices (Android 13-14, iOS 16-17) with ground-truth chat artefacts, SHA-256 hashed and signed.
- A CSV log of 14 000 shortened URLs expanded by the hyperlink tool.
- 200 synthetic deep-fake videos and 200 bona fide control clips with frame-level integrity hashes.
The bundle is hosted on IEEE DataPort with a DOI and is free for academic use.
2025 Community Challenge
Teams are invited to build deep-fake detectors that achieve ≥ 95 % accuracy and ≤ 5 % false-positive rate on the released corpus. Submissions must be containerised, include a reproducible GitHub Action and be capable of processing one hour of video on a single GPU within 30 minutes. The top three entries will present at E-Discovery Day 2025 and receive hardware tokens signed with post-quantum signatures.
Conclusion
The tenth iteration of E-Discovery Day evolved from a commemorative webinar into a live data-sharing experiment. By open-sourcing tools, reference data and draft standards, the community has created a feedback loop in which practitioners, researchers and the bench can validate emerging methods before they reach the courtroom. Continuous publication of such artefacts will be critical as data volumes, encryption and synthetic media further complicate the discovery landscape.
