Lead story
Court Orders Perplexity to Produce Another RAG Snapshot and Six Months of Activity Logs
A Southern District of New York magistrate judge ordered Perplexity to produce one additional retrieval-augmented-generation database snapshot and host six months of additional user activity log data, with plaintiffs contributing up to $6,000 per month toward hosting.
- What happened
- A Southern District of New York magistrate judge ordered Perplexity to produce one additional retrieval-augmented-generation database snapshot and host six months of additional user activity log data, with plaintiffs contributing up to $6,000 per month toward hosting.
- Why it matters
- The order treats operational AI records as discoverable evidence and applies familiar relevance, proportionality, and cost-allocation principles to unusually large and technically complex data stores. Litigation teams should identify AI-system logs and retrieval stores early, document retention and snapshot practices, and develop defensible estimates for extraction, secure hosting, and review.
- Watch
- The September 2 order resolves a discovery dispute, not the underlying copyright or trademark claims. The cited case-law page is hosted by Minerva26, and the eDiscovery Today article notes Minerva26 is an affinity partner.
SourceseDiscovery Todayapp.minerva26.com
Briefing item
Husch Blackwell Launches AI-Based Platform for Complex Torts Defense
Husch Blackwell launched CXT, an AI-focused platform intended to provide shared infrastructure, market intelligence, and pricing tools for the firm's complex-torts work.
- What happened
- Husch Blackwell launched CXT, an AI-focused platform intended to provide shared infrastructure, market intelligence, and pricing tools for the firm's complex-torts work.
- Why it matters
- Complex-torts matters generate large, recurring evidence and workflow demands across cases, teams, and jurisdictions. A shared platform can standardize matter data, discovery processes, staffing, and pricing, but legal teams will need clear controls for source data, work-product validation, privilege, model use, and measurement of promised efficiencies.
- Watch
- The article provides limited technical detail and is available through Law.com's Legaltech News. Platform capabilities and efficiency claims are based on the firm's launch description and were not independently tested in the cited report.
SourcesLaw.com
Briefing item
UniCourt Adds Court Dockets and Documents to ChatGPT Enterprise
UniCourt launched a beta ChatGPT Enterprise plugin that lets subscribed users search dockets and documents from more than 5,000 state and federal courts through natural-language requests.
- What happened
- UniCourt launched a beta ChatGPT Enterprise plugin that lets subscribed users search dockets and documents from more than 5,000 state and federal courts through natural-language requests.
- Why it matters
- Conversational access to broad docket and document collections could accelerate early case assessment, precedent research, party and counsel analysis, and litigation monitoring. Teams should test court coverage, update latency, citation accuracy, source-document access, query logging, and whether results can be reproduced outside the conversational interface.
- Watch
- Coverage, document counts, performance, and cost-efficiency statements are vendor-provided through a company announcement published on LawNext. The plugin is in beta and requires a UniCourt enterprise subscription; no independent accuracy benchmark is cited.
Sourceslawnext.com
Briefing item
Legal AI Differentiation Shifts Toward Data Quality, Provenance, and Workflow
A new analysis argues that as foundation-model capabilities converge, legal AI products will increasingly be distinguished by current, structured, connected data and verifiable paths back to source authority.
- What happened
- A new analysis argues that as foundation-model capabilities converge, legal AI products will increasingly be distinguished by current, structured, connected data and verifiable paths back to source authority.
- Why it matters
- Discovery and litigation teams evaluating AI systems need to assess more than the underlying model. Procurement and validation should examine corpus completeness, update practices, metadata and citation relationships, source traceability, matter-level permissions, reproducibility, and the controls applied when tools act across multiple workflow steps.
- Watch
- This is an analysis by the chief executive of Manupatra, which offers the legal AI products discussed in the article. Its conclusions are directional and do not report a comparative product test or independent benchmark.
Sourcesartificiallawyer.com
Briefing item
Deloitte Survey Finds Workplace GenAI Use Widespread but Shallow
Deloitte's survey of 25,000 UK employees found widespread generative-AI experimentation but limited daily use, with common applications concentrated in email writing, information searches, and summaries.
- What happened
- Deloitte's survey of 25,000 UK employees found widespread generative-AI experimentation but limited daily use, with common applications concentrated in email writing, information searches, and summaries.
- Why it matters
- Document review, investigation, and legal research are plausible candidates for deeper workflow integration, but adoption alone is not evidence of return. Discovery leaders should measure matter-level time, quality, rework, defensibility, training, and exception handling while monitoring unsanctioned tool use and data exposure.
- Watch
- The Deloitte survey covers UK workers generally rather than legal or eDiscovery professionals. The detailed time-saving figures cited by Artificial Lawyer come from a finance, banking, and insurance segment, so they should not be treated as legal-industry results.
Sourcesdeloitte.comartificiallawyer.com
Educational only. Not legal advice. Review sources and matter-specific obligations before relying on any workflow decision.