Lead story
Enterprise AI Procurement Creates New Preservation Demands
An ILTACON report says organizations are beginning to address the preservation of data generated by enterprise AI systems as those tools become relevant to litigation and investigations.
- What happened
- An ILTACON report says organizations are beginning to address the preservation of data generated by enterprise AI systems as those tools become relevant to litigation and investigations.
- Why it matters
- Legal and information-governance teams need to know which AI systems retain prompts, outputs, logs, and metadata; where that information resides; and whether it can be preserved and exported consistently. Those facts affect hold notices, collection scope, chain of custody, authentication, and the ability to explain AI-assisted business decisions during discovery.
- Watch
- Law.com published the report on August 24 from ILTACON 2026. The accessible article excerpt describes an emerging operational problem and vendor activity but does not establish a uniform preservation standard or test particular collection tools.
SourcesLaw.com
Briefing item
Data-Protection Rules Complicate Centralized Multinational eDiscovery
A multinational competition matter illustrates how data-protection restrictions can prevent a company from centralizing discovery data across simultaneous investigations in the United States, United Kingdom, and Europe.
- What happened
- A multinational competition matter illustrates how data-protection restrictions can prevent a company from centralizing discovery data across simultaneous investigations in the United States, United Kingdom, and Europe.
- Why it matters
- Cross-border investigations require an architecture that accounts for data location before collection and review begin. Teams may need regional processing, federated search, controlled transfers, consistent coding protocols, and coordination mechanisms that preserve defensibility without moving restricted data into a global workspace.
- Watch
- Published August 24 as a Law.com analysis based on an ongoing competition-enforcement matter. The accessible excerpt does not identify the company or jurisdictions beyond the United States, United Kingdom, and multiple European countries, and legal requirements remain matter- and jurisdiction-specific.
SourcesLaw.com
Briefing item
ILTACON Session Highlights Prompting Risks in Legal AI Use
ILTACON speakers addressed prompt-related failure modes that can be difficult for attorneys to detect, including model judgment calls and outputs that reinforce a user's assumptions.
- What happened
- ILTACON speakers addressed prompt-related failure modes that can be difficult for attorneys to detect, including model judgment calls and outputs that reinforce a user's assumptions.
- Why it matters
- When AI assists with document review, chronology building, issue analysis, or drafting, teams need written instructions, representative testing, source-linked outputs, escalation rules, and attorney verification. Recording material prompt changes also supports repeatability and helps explain how a result was produced.
- Watch
- Published August 24 from an ILTACON educational session. The publicly accessible portion identifies the risk categories but the full article requires an account, so this summary does not attribute additional techniques or performance claims beyond the accessible text.
SourcesLaw.com
Briefing item
Thomson Reuters Launches Proprietary Legal and Tax Language Model
Thomson Reuters launched Thomson, an in-house language model trained for legal, tax, and professional work, with its first deployment planned for structured document analysis in CoCounsel Legal.
- What happened
- Thomson Reuters launched Thomson, an in-house language model trained for legal, tax, and professional work, with its first deployment planned for structured document analysis in CoCounsel Legal.
- Why it matters
- A provider-controlled, domain-trained model gives legal teams another option for document analysis where model provenance, deployment control, citation behavior, privacy, and inference cost matter. Its place in a multi-model product also makes task routing and validation important: teams must know which model handled a job and what evidence supports the output.
- Watch
- Announced August 24 by Thomson Reuters. The $40 million training cost, benchmark comparisons, efficiency statements, and performance claims are company-reported; broader independent evaluations and the cited technical report will be important for validation.
Sourcesthomsonreuters.comsiliconangle.com
Briefing item
Newcode Series A Draws Investment From Relativity's Rel Labs
Newcode announced a Series A that brings its reported 2026 funding to $20 million, with participation from Relativity's Rel Labs and other legal-technology investors.
- What happened
- Newcode announced a Series A that brings its reported 2026 funding to $20 million, with participation from Relativity's Rel Labs and other legal-technology investors.
- Why it matters
- Rel Labs' participation links an established eDiscovery company to infrastructure intended to let firms connect and control their own AI workflows. For discovery and investigations teams, the relevant evaluation questions include connector permissions, retrieval boundaries, provenance, audit logs, data residency, and whether outputs remain traceable to underlying documents.
- Watch
- Published August 24. Newcode did not disclose the Series A amount separately; the $20 million figure is total reported 2026 funding, including a $6.5 million seed round. Customer count and revenue-growth statements are company-reported.
SourcesLegal IT Insidernewcode.ai
Educational only. Not legal advice. Review sources and matter-specific obligations before relying on any workflow decision.